The Anti-Weaponization Fund proved to be politically controversial, wherefore the Acting Attorney General testified to Congress that it would not proceed. Still, litigation over the Anti-Weaponization Fund soldiers on. A federal district court in Virginia issued a preliminary injunction against the Anti-Weaponization Fund, and the parties continue to litigate whether the Fund is legal. Meanwhile, the federal district court in Florida that was presiding over the purportedly settled lawsuit issued an order that, among other things, banned both sides from “referring to the purported ‘settlement agreement,’ or using, offering, admitting, or citing any of its provisions in any judicial, administrative, regulatory, arbitration, or any other official proceeding as evidence of a ‘settlement’ reached in this matter.”

Today’s post will offer some commentary on the Anti-Weaponization Fund and the satellite litigation it has spawned.

It wasn’t me

I’m going to start with some uncontroversial table stakes: the Anti-Weaponization Fund is illegal.

The Anti-Weaponization Fund is the product of a settlement of a lawsuit and two administrative claims brought by the President. Let’s start with the lawsuit. In 2020, during the first Trump Administration, an IRS contractor, Charles Littlejohn, leaked thousands of people’s tax return information, including President Trump’s, to the news media. Littlejohn was charged, convicted, and sentenced to five years in prison.

On January 29, 2026, President Trump (along with his older sons and the Trump Organization LLC—I’ll call the plaintiffs “President Trump” for short) sued the IRS, bringing two claims: one under the Internal Revenue Code’s express cause of action for unauthorized disclosure of tax returns, and the other under the Privacy Act. President Trump sought a mere $10 billion in damages.

This lawsuit had many weaknesses, most notably that it was time-barred. The Internal Revenue Code requires that a cause of action be brought “within 2 years after the date of discovery by the plaintiff of the unauthorized inspection or disclosure.” Privacy Act claims must be brought “within two years from the date on which the cause of action arises.” That means the critical date—two years before the filing of the lawsuit—is January 29, 2024.

The New York Times published its story in September 2020. In December 2022, Ken Griffin sued the IRS based on the exact same leak. In October 2023, Littlejohn pleaded guilty, admitting he leaked President Trump’s tax returns. Alina Habba, President Trump’s attorney, attended the plea hearing and spoke on behalf of the President, saying, among other things:

MS. HABBA: Thank you, Your Honor, for the opportunity to be heard. As an attorney, I find that I should probably state I am not licensed in this state, in the District of Columbia. I’m here on behalf of President Trump who was a victim, as we just heard, of this atrocity. … This was an egregious breach by an agent of the IRS who targeted the President of the United States, among others, for political purposes and personal gain. As we just heard in the statement of facts read into the record, President Trump’s returns were turned over to the New York Times right before an election. As I also learned today, his were turned over before everybody else’s.

Some of you may remember the song “It Wasn’t Me” by Shaggy, which was released 26 years ago (?!). The plot of this song is that Shaggy advises his friend Rikrok to insist that he (Rikrok) did not cheat on his girlfriend even though she not only witnesses the entire event but also records it on camera. This is not too different from the legal theory undergirding Trump v. IRS, which hinges on the argument that the President had not yet discovered the unauthorized disclosure of his tax returns at a time when the President’s lawyer publicly declared in open court at the perpetrator’s plea hearing that the President was the victim of that very unauthorized disclosure.

The settlement agreement that gave rise to the Anti-Weaponization Fund covers not only President Trump’s lawsuit against the IRS, but also two tort claims that were pending at the administrative level.

One was a claim arising out of the Mar-a-Lago search in August 2022. This claim has lots of weaknesses for reasons AI will be glad to explain, but even setting those aside, the claim sought a mere $15 million in compensatory damages—two orders of magnitude less than the $1.776 billion Anti-Weaponization Fund settlement.1 The other claim—described in the settlement agreement as arising from “the Russia-collusion hoax”—does not seem to be available online, but the New York Times reports that it was lodged in “late 2023.” Given that the Mueller report came out in 2019, the claim was likely filed well after the expiration of the two-year statute of limitations.

So we’ve got two likely time-barred claims and one claim where the theoretical maximum recovery was $15 million, which does not bode well for a large recovery. But the President did have one thing going for him, which is that he controls the Department of Justice. And so, on May 18, rather than defending against the lawsuit, the normally tenacious Department of Justice agreed to settle the case for the patriotic sum of $1.776 billion, to be doled out to third parties by the “Anti-Weaponization Fund.”

Article I, Section 9, Clause 7 of the Constitution provides: “No Money shall be drawn from the Treasury, but in Consequence of Appropriations made by Law.” Thus, for the Anti-Weaponization Fund to be legal, there has to be a Law—an Act of Congress—authorizing $1.776 billion to be spent. The government has pointed to the judgment fund statute, 31 U.S.C. § 1304, which appropriates “necessary amounts” to pay “compromise settlements.”2 The government’s theory is that it entered into a “compromise settlement” with the President, opening the door to the $1.776 billion payment.

There are (at least) two problems with this argument:

1. There is no “compromise settlement” here.

2. Even if there were, there’s no statutory authorization to create the Anti-Weaponization Fund.

First, there is no “compromise settlement.”

1. The concept of a “compromise” involves two independent actors, with separate interests, that compromise on some intermediate position. You cannot compromise with yourself. Yes, I have seen Inside Out, in which Joy and Sadness achieve a negotiated resolution that allows Riley to resume playing hockey, but even there, Joy and Sadness are separate characters with separate voice actors. This case is more like Joy compromising with Joy.

2. The settlement agreement recites: “The corpus of The Anti-Weaponization Fund’s funding does not represent the value of any current claim by Plaintiffs, but rather is based on the projected valuation of future claimants’ claims.” I view that as more or less an open admission that the settlement agreement isn’t a “compromise settlement” of the President’s claim. A “compromise settlement” reflects an assessment of the value of the claim, discounted for chances of success, attorney’s fees, the time value of money, and so forth. If the government pays an amount of money that has nothing to do with the value of the claim, it’s not entering into a “compromise settlement” of the claim.

3. The settlement agreement then goes on to recite: “[T]he corpus of The Anti-Weaponization Fund’s funding is not taxable income as to Plaintiffs, who receive no economic benefit from this Settlement Agreement.” The premise of a “compromise settlement” is that the plaintiff gets some money, and in exchange for getting that money, drops the claim. That’s what the appropriation is for: paying off the plaintiff in exchange for the plaintiff dropping the claim. There’s no “compromise” with the plaintiff when the plaintiff doesn’t get anything.

   (There’s also another document dated the next day and signed by the Acting Attorney General, who did not sign the settlement agreement, purporting to give amnesty to the President, Donald Trump Jr., Eric Trump, and the Trump Organization from future government investigations. This document also raises some intriguing legal issues that I won’t get into. But for present purposes, I don’t think the amnesty provision is consideration for the plaintiffs dropping the claim? If it is, then I withdraw this particular point.)

If the “compromise settlement” theory were viable, then the President’s challenges in persuading Congress to appropriate funds would be magically resolved: the President could just sue himself and convince the Justice Department, under his control, to “settle” the case by spending taxpayer money on whatever he wants. For instance, the President does not need to convince Congress to spend $1 billion on ballroom security. He could simply sue the federal government for some completely unrelated thing and then immediately direct the Acting Attorney General to “settle” the case by doling out $1 billion from the judgment fund for ballroom security. The value of the lawsuit need not have any correlation with the settlement amount—the explicit premise of the Anti-Weaponization Fund is that the $1.776 billion fund has no relationship to the value of the President’s claims.

There is another problem with the Anti-Weaponization Fund. The Judgment Fund statute is an appropriations statute. It authorizes the spending of money. That’s it. The settlement agreement doesn’t just create a Fund. It creates a kind of new executive agency with new authority: to spend the money in the Fund as dictated by the settlement.

To set up a new agency like the Anti-Weaponization Fund, you need an Act of Congress. The President can’t create a new agency with new powers merely by suing, and “settling” with, the Attorney General. Suppose there’s a future administration in which the President would really like to create a new Department of Climate Change charged with spending taxpayer money to alleviate the climate emergency. Suppose Congress demurs. Should the President be able to create this new agency unilaterally by: (1) suing his own administration over some unrelated thing, and (2) convincing his own Attorney General to “settle” the suit by creating the Department of Climate Change in exchange for his gracious decision to drop his lawsuit? This is more or less what happened here.

The Members of the Anti-Weaponization Fund also aren’t entitled to exercise authority to dole out funds. The Constitution prescribes the process for appointing people who exercise significant governmental power. Under the Appointments Clause, the President has the power, “by and with the Advice and Consent of the Senate,” to appoint “Officers of the United States, whose Appointments are not herein otherwise provided for, and which shall be established by Law: but the Congress may by Law vest the Appointment of such inferior Officers, as they think proper, in the President alone, in the Courts of Law, or in the Heads of Departments.” I view the Members of the Anti-Weaponization Fund as principal (not inferior) officers, which means that (1) their offices must be “established by Law,” (2) they must be appointed by the President, and (3) they must be confirmed by the Senate, none of which is occurring under the settlement agreement. Even if the Members of the Anti-Weaponization Fund are inferior officers, the Attorney General can’t appoint them unless Congress authorized the Attorney General to do so, which it never did.

Does the President have the inherent executive authority to create the Anti-Weaponization Fund, even without an Act of Congress and even without following constitutional appointment procedures? I think not. The Federalist Papers address the Legislative, Executive, and Judicial Branches in detail. By contrast, according to AI, only one part of one Federalist Paper contemplates that the President can create a fourth branch of government by suing and settling with his own Attorney General for purposes of paying taxpayer money to his supporters without congressional approval. Even there, only Alexander Hamilton was a supporter, with James Madison and John Jay tentatively opposed. So even originalists should oppose the Anti-Weaponization Fund.

Equal Justice Under Law

Doubtless because of public concern over the proper interpretation of the judgment fund statute, a hubbub arose over the Anti-Weaponization Fund. And, of course, people sued.

On June 2, the Acting Attorney General testified before the House of Representatives that the Anti-Weaponization Fund wouldn’t proceed. In the least surprising development of all time, the litigation did not stop.

Several lawsuits were filed; I’ll focus on Floyd v. Department of Justice in the Eastern District of Virginia. A group of plaintiffs sued the government, seeking to enjoin the Anti-Weaponization Fund. Now you may wonder how the plaintiffs would have standing to seek such an injunction. After all, a plaintiff can’t get an injunction unless it suffers a particularized injury, and “the government is spending taxpayer money illegally” doesn’t count. But the plaintiffs have a creative theory.3 Remember how the settlement agreement defines Weaponization as “the sustained use of the levers of government power by Democrat elected officials”? Well, the plaintiffs say that they were victimized by Republican elected officials, and hence (they claim) are ineligible for payments from the fund. That’s discrimination! And it’s particularized discrimination, because (they claim), they, personally, would be ineligible for payouts. Wiping out the Anti-Weaponization Fund wouldn’t get them any money, but it would eliminate the sting of unequal treatment, they say. Like it says on the Supreme Court building: Equal Justice Under Law.

The government responded, among other things, that the case was moot in view of the Acting Attorney General’s House testimony that the Anti-Weaponization Fund was dead. But on June 12, the district court granted a preliminary injunction halting the operation of the Anti-Weaponization Fund. In response to the government’s mootness claim, the court invited the defendants (the Acting Attorney General, Associate Attorney General, and Secretary of the Treasury) to “file a declaration under the penalty of perjury that they will not take any action to create or operate the Anti-Weaponization Fund, and that the Anti-Weaponization Fund will not proceed in any manner, or under any name.” Per the court, if they filed such a declaration, the lawsuit would go away; if they didn’t, the lawsuit would stay in place and the injunction would remain in force.

Wait a second. We’re talking about a settlement here. On Earth-One, one side of a settlement cannot unilaterally declare that the settlement will never be carried out. Like, if Alice sues Bob and Bob agrees to pay Alice $1 million and Alice drops the case in exchange, and then Bob announces “actually sorry I’m not going to pay the $1 million,” one would not say, “OK, I guess we can be certain Bob’s not going to pay the $1 million!” Instead you would say “I guess Alice is going to sue Bob for the $1 million, have fun losing in court Bob!”

Yet here, the government takes the position that once the defendants declare the Anti-Weaponization Fund to be dead, it’s dead. The concept of the plaintiffs enforcing the settlement does not cross anyone’s mind. This suggests that this isn’t a “compromise settlement.” But remember that the statutory basis for the Anti-Weaponization Fund was that the government entered into a “compromise settlement,” thus unlocking the Judgment Fund door! If there’s no actual “compromise settlement,” the Anti-Weaponization Fund is illegal.

Still, the district court was willing to declare the case moot based on declarations from only one side of the settlement. This was an easy ask from the district court. It would have been literally a one-line declaration reciting that the thing that the Acting Attorney General told Congress is, in fact, true.

Still, the government refused to submit the declarations. It gave two reasons:

• The declarations were “unnecessary” because the Acting Attorney General had already made these statements to Congress, and the lawyers had signed briefs reporting those statements, so who needs sworn declarations anyway?

• “The compelled testimony of senior officials from the Executive Branch implicates serious separation of powers concerns.”

The government’s justifications are not persuasive. On the first point, it was reasonable for the district court to insist on sworn declarations from the defendants before dismissing the case as moot.

1. The way litigation normally works is, if a party wants to convince the court that a particular fact is true, the party submits a declaration saying it’s true. The district court was entitled to apply ordinary procedural rules.

2. To be sure, when we’re talking about official government actions, usually the government is allowed to submit official government documents. The problem is that there isn’t an official government document rescinding the Anti-Weaponization Fund. The original documents are still up on the DOJ’s website!

3. The settlement agreement says: “This Settlement Agreement may be modified only with the written agreement of the Parties.” There’s no written agreement of the Parties. Indeed, at the House hearing where the Acting Attorney General made the very comments that, according to the government, mooted the litigation, the Acting Attorney General explicitly refused to put anything in writing:

GRACE MENG: Thank you, Mr. Chairman. Um, Mr. Attorney General, I want to thank you for verbally committing to not moving forward with the so-called, um, Anti-Weaponization Fund. I just want to make sure, are you going to issue a new memo in writing rescinding that May 18 memo?

TODD BLANCHE: I'm not committing to putting anything in writing. And I said it over and over again.

GRACE MENG: Nothing in writing. OK, thank you.

TODD BLANCHE: I mean, I don't know what the purpose of putting something in writing, I'm telling you what we're doing. Meaning, like, what’s the -- why do I need to put something in writing if I'm telling you what we're doing?

4. Any practicing lawyer will tell you that sometimes judges will ask you to do things that you think are unnecessary. That’s life. It’s their courtroom, they get to decide. Unless you have a really good reason to refuse, you do what they tell you to do. You don’t say “Well, I think this is unnecessary, so I just won’t do it.”

The government’s second justification is that the district court’s request raised “separation of powers” concerns. It did not. Separation-of-powers concerns arise when the court forces high-level government officials to sit for depositions or take other burdensome steps. But here, the court wasn’t forcing any high-level government official to do anything. It was just saying: “if you want me to dismiss the plaintiffs’ case as moot based on your unilateral representations, you’re going to have to submit evidence to the court in the way people ordinarily submit evidence to courts.” That is reasonable.

Reasonable or not, the government didn’t submit the declarations, and so the injunction remained in place and the litigation continued. Most recently, the government filed a motion to dismiss. It’s still refusing to submit any sworn declarations, but it argues (among other things) that the case is moot for the additional reason that the Acting Attorney General said at his confirmation hearing that the Anti-Weaponization Fund isn’t going forward. I’d load up NO on Polymarket for the government winning on that one.

Why did the government refuse to submit the declarations? I have no idea. On principle? The principle of “we should be able to get cases dismissed without submitting evidence” is not the hill I would die on. Because it doesn’t like being told what to do by a court? That’s sort of what courts are there for. Because it secretly planned to resurrect the Anti-Weaponization Fund? Maybe, but the inevitable consequence of refusing to submit the declaration would be a continuation of the injunction, so it couldn’t be resurrected anyway.

Could it be that the government actually desired the injunction? After all, if the government wanted to find a way to kill the Anti-Weaponization Fund without writing anything down, a court order would do the trick. If that’s true, then the government would have used collusion both to create and to cancel the Anti-Weaponization Fund. That has to be some kind of world record.

$1,776

Let us move on to our second case, which is Trump v. IRS—the lawsuit that precipitated the Anti-Weaponization Fund settlement.

On May 18, following the settlement, the plaintiffs dismissed the case with prejudice under Federal Rule of Civil Procedure 41(a)(1)(A)(i). Under that rule, if the case hasn’t proceeded too far, the plaintiff can drop a case without a court order: the plaintiff just submits a notice of dismissal and the case goes away. So it looked like the case was completely done. But on May 27, a group of former federal judges asked the court to set aside the dismissal and reopen the case because the dismissal was a “fraud on the court.” The President’s attorneys submitted an appropriately dyspeptic response.

On July 13, the court issued an order imposing sanctions. The court didn’t reopen the case as the former federal judges requested (I doubt it had the authority to do so). But the Supreme Court has held that courts are authorized to impose sanctions based on litigation conduct even after a case is voluntarily dismissed, so that’s what the district court did.

The court concludes that the plaintiffs brought a collusive lawsuit for an improper purpose. Under Article III of the Constitution, federal courts lack jurisdiction to hear cases unless the parties are adverse. The court holds that “Plaintiffs and Defendants are not adverse because one party controls this litigation.” It explains that the President, the plaintiff, enjoys complete control over the Defendants, who are government officials under the President’s supervision. It also points out that the government fought other, similar lawsuits arising from the tax leak but threw in the towel in this case.

The court has harsh words about the settlement, finding that “the extraordinary award fashioned by the Parties for claims that were never litigated, and have yet to be defined, on behalf of unidentified third parties whose future remedies bear no relationship to the claims in this case, indicates that real adverse interests were never before the Court.” The court also notes that the Acting Attorney General previously served as the President’s personal lawyer, while the Associate Attorney General previously represented January 6 defendants who might have benefited from the Anti-Weaponization Fund, and “instead of either recusing because of their previous representations or vigorously defending this lawsuit as required to do so by DOJ policies and procedures, these lawyers agreed to a ‘settlement’ involving a staggering amount of money potentially benefitting former clients.” The court makes some additional spicy points, such as that “the conferral of possibly millions of dollars in tax relief and corollary benefits potentially violates Article II, Section I of the United States Constitution, a limitation surely known by former White House Counsel and the current Acting Attorney General.”

The court finds it “so obvious and so insurmountable” that the parties were not adverse that “this matter was brought for an improper purpose—to gain the imprimatur of judicial legitimacy for a ‘settlement’ that had no viable basis in law or fact.” This finding of “improper purpose” unlocks the door to sanctions.

Federal Rule of Civil Procedure 11 permits the court to impose sanctions for improper filings. The court sanctions the plaintiffs’ lawyers, Alejandro Brito and Daniel Epstein, by referring Mr. Brito to the Florida Bar and stating that Mr. Epstein’s pro hac vice applications in the Southern District of Florida will be denied for one year.

As a practicing lawyer, I get squeamish whenever a court imposes sanctions on a lawyer. There but for the grace of God. Actually, there are many factors other than God’s grace that explain why I personally was not sanctioned for orchestrating the Anti-Weaponization Fund. Still, I have some sympathy for the President’s attorneys here. It’s not their fault that their client controls the government!

It’s not clear to me that, at the time the suit was filed, the attorneys were aware that the Anti-Weaponization Fund was on the horizon. Perhaps they knew at some level that the lawsuit might get weird, but as far as I can recall, neither my law school professional responsibility class nor the MPRE offered specific guidance on what to do when you, the President’s personal lawyer, are asked to file a lawsuit against the IRS. The court observes that Mr. Epstein signed the settlement but never filed a pro hac vice application, which (in the court’s view) supports the inference that he “was aware that he would never need to appear and litigate the merits of Plaintiffs’ claims.” That strikes me as uncharitable.

The court also issues the following sanction:

The Parties are prohibited from referring to the purported “settlement agreement,” or using, offering, admitting, or citing any of its provisions in any judicial, administrative, regulatory, arbitration, or any other official proceeding as evidence of a “settlement” reached in this matter, Case No. 26-cv-20609-KMW (S.D. Fla. 2026).

I can see why the court is doing this: using the word “settlement” implies there was once a real lawsuit with adverse parties, which (in the court’s view) there wasn’t, and the court doesn’t want the parties misrepresenting what happened in her courtroom. Still, I question whether the court had authority to issue this order. It’s not really a sanction in the traditional sense of punishing prior misconduct or preventing misconduct from recurring. (The court isn’t banning the parties from future collusive litigation.) Nor does it protect the integrity of the judicial proceedings before the Florida court—those proceedings are over. Perhaps it is intended to protect the integrity of other official proceedings, but I am not sure the court, after a self-executing dismissal, has the authority to dictate how parties characterize an extrajudicial document in proceedings before other tribunals. If the parties want to talk about the “settlement” in some other court, then the other court can decide whether to take them seriously. Finally, the sanction applies to both parties, but it was the plaintiffs that filed the complaint and the dismissal. I’m skeptical that the court can impose this restriction on the defendants under Rule 11 if they didn’t file anything.

I’m also unsure how this sanction will work in practice. It seems to ban the defendants from characterizing the settlement agreement as a “settlement” in the ongoing Virginia case. But the defendants’ motion to dismiss in the Virginia case, which post-dates the sanctions order, is peppered with references to the settlement agreement and treats it as a bona fide “settlement.” E.g., the motion argues that the plaintiffs lack standing because “exclusion from a targeted, settlement-derived compensation scheme does not pose any harm to ‘a legally protected interest.’” Maybe that statement is OK because it doesn’t specifically refer to the litigation in Florida?

Also, does the sanction apply to Anti-Weaponization Fund proceedings? Yes, the Acting Attorney General said the Fund wasn’t proceeding and the Virginia court enjoined it, but put aside those technicalities. The sanction applies to “official proceedings,” and Anti-Weaponization Fund proceedings seem official to me. The court says in footnote 63 that it’s not deciding whether the Anti-Weaponization Fund is illegal, so presumably the order isn’t enjoining the Anti-Weaponization Fund itself, but it does prevent the parties from referring to the “settlement” in those proceedings, which is tricky. Perhaps the government lawyers manning the Anti-Weaponization Fund would have to learn to speak in delicate Victorian euphemisms, kind of like how “a meeting” used to be a euphemism for dueling.

Finally, the court issues monetary sanctions under its “inherent authority.” It invites parties who filed amicus briefs to submit a request for reimbursement of their attorney’s fees out of the plaintiffs’ pockets. If I were the amici, I would have sought $1,776 (or maybe $177,600) in fees from this fledgling Anti-Anti-Weaponization Fund. But the amici who submitted the May 27 brief ultimately sought only $1 in fees plus $4,606.83 in costs, while another group of amici sought $39,000. Too bad.

Coda

The time has come for me to wrap up this article by offering thoughtful commentary on what this episode reveals about the Law.

I’m not sure what to say. Does it demonstrate how the independent judiciary is bravely preserving the rule of law? Sort of, but the Acting Attorney General cancelled the Anti-Weaponization Fund on his own. And I’m not sure the Florida order affirmed the rule of law, exactly.

Maybe the whole ordeal is a vindication of Article III? Article III’s requirement of a “case or controversy” predicts the pathologies that arise from collusive litigation. Its prohibition on litigation arising from generalized grievances predicts that the political process can address concerns of excessive or unauthorized Executive Branch spending. Those predictions wind up looking good. The Founders win again.

The best I can come up with is an AI-generated rhyming acrostic spelling “Anti-Weaponization Fund” which is 22 characters by 22 characters:

Fable notes: “Lines 13–14 rhyme ‘himself’ with ‘himself’—an identical rhyme, which is ordinarily a poetic sin, but here it’s the entire point: you cannot compromise with yourself, and the couplet shouldn’t be able to either.” Eh. Maybe Mythos would do better.

The Supreme Court has temporarily blocked the Fifth Circuit’s order and is considering whether to keep that block in place while the case proceeds. It should. The Fifth Circuit erred. Louisiana, the plaintiff challenging the FDA’s decision, lacks standing. The FDA’s decision was also legal. It was not arbitrary and capricious for the FDA to conclude, based on an extensive body of evidence and years of experience with mifepristone, that this safety measure is unnecessary.

The Mifepristone Wars

The FDA regulates drug safety. When it approves a drug, it has the authority to impose a “risk evaluation and mitigation strategy”—known as “REMS” for acronym-loving lawyers—when “necessary to ensure the benefits of the drug outweigh the risks of the drug.”

Mifepristone is an FDA-approved drug taken by women to terminate pregnancies. Pre-COVID, the REMS for mifepristone required women to go to their health care providers, in person, to pick up mifepristone, rather than obtaining it at the pharmacy or by mail. They didn’t have to take the mifepristone in the provider’s presence. They just had to physically pick it up there.

In July 2020, during COVID, a federal court enjoined this in-person dispensing requirement. In January 2021, the Supreme Court stayed this injunction and the in-person dispensing requirement sprang back into force. But in April 2021, the FDA said it would stop enforcing the requirement. In December 2021, following a review of the REMS, the FDA concluded that in-person dispensing wasn’t necessary to protect safety and decided that the REMS should be modified.

A modification to the REMS occurs only after the drug manufacturer submits a REMS modification and the FDA approves that modification. And so, in December 2021, the FDA directed the drug manufacturers to submit proposed REMS modifications. They did. The FDA approved the modified REMS in January 2023.

This latest episode of the Mifepristone Wars began in October 2025, when Louisiana sued the FDA. The purpose of this suit was to overturn the FDA’s decision to modify the REMS and force the FDA to restore the in-person dispensing requirement.

After the FDA told the district court that it was reconsidering the REMS, the district court elected to give the FDA the time it needed to complete that review and declined to grant Louisiana any immediate relief. Louisiana took the case to the Fifth Circuit, which issued an order immediately blocking the REMS modification and effectively reimposing the in-person dispensing requirement nationwide.

Protecting safety in order to not protect safety

Louisiana may be standing in this image, but it does not have standing in this case.

To establish standing, Louisiana must show that it is experiencing an injury caused by the FDA’s decision to modify the REMS that would be redressed by an order overturning that decision.

Louisiana offers two theories of standing. Theory #1 goes like this: if it’s possible to obtain mifepristone by mail, then women in Louisiana will order mifepristone by mail from out-of-state providers, which inflicts what Louisiana refers to as a “sovereign injury” because abortion is illegal in Louisiana.

Louisiana’s claim on the merits—the thing that, according to Louisiana, the FDA did wrong—is that the FDA was insufficiently attentive to women’s safety when it jettisoned the in-person dispensing requirement. However, Louisiana does not actually want women to take the extra safety measure of picking up mifepristone from their providers. To the contrary, it wants women not to do this. It bans local hospitals from doing the very thing that, it contends, should be mandatory for safety purposes—in-person dispensing of mifepristone. Women could in principle drive to a state like Illinois to try to obtain mifepristone, but Louisiana is hoping that women will find this too burdensome and will instead carry the pregnancy to term.

Ironically, under the REMS statute, the FDA must periodically evaluate whether the elements of the REMS “assure safe use of the drug” and “are not unduly burdensome on patient access to the drug.” In a normal world, a plaintiff trying to force the FDA to impose a more stringent REMS requirement would be saying things like “the requirement will assure safe use of the drug and won’t be unduly burdensome.” But Louisiana is saying the exact opposite. It’s saying it doesn’t want “safe use of the drug” and is hoping it will be “unduly burdensome” for Louisiana women to drive to a different state to get the drug. Sure, there are lots of women in other states where abortion is legal who will be forced to take safety measures they don’t want or need, but from Louisiana’s perspective, that’s just collateral damage. In the Fifth Circuit, this argument wins.

Consider the situation from the perspective of a woman living in a state where abortion is legal. She wants to obtain mifepristone by mail or at the pharmacy, but, if the Fifth Circuit’s order goes into effect, she can’t. Instead, purportedly for purposes of protecting safety, she has to drive to a further-away building. But she doesn’t want this extra safety measure. Her doctor doesn’t recommend this extra safety measure. Her state—which, Dobbs claims, now gets to decide abortion policy—doesn’t support this extra safety measure. The FDA doesn’t think this extra safety measure is necessary. Yet she’s forced to take this extra safety measure, because Louisiana convinced the Fifth Circuit to mandate the safety measure nationwide so that Louisiana’s citizens also don’t take the safety measure.

Article III’s “case or controversy” requirement exists to guard against precisely this sort of vexatious litigation. In FDA v. Alliance for Hippocratic Medicine, a group of pro-life doctors brought a lawsuit seeking to ban all women in America from using mifepristone. Their theory of standing was that if mifepristone were banned, they wouldn’t face the risk of encountering women who had taken mifepristone in the emergency room. The Supreme Court rejected this theory. The Court explained that “FDA has not required the plaintiffs to do anything or to refrain from doing anything.” It observed that “the law has never permitted doctors to challenge the government’s loosening of general public safety requirements simply because more individuals might then show up at emergency rooms or in doctors’ offices with follow-on injuries.” If the plaintiffs’ theory were correct, the Supreme Court explained, absurd consequences would follow: “Police officers,” for example, “could sue to challenge a government decision to legalize certain activities that are associated with increased crime.”

That’s this case. The FDA’s decision to modify the REMS doesn’t require Louisiana to do, or not do, anything. Instead, Louisiana is suing the FDA to stiffen safety restrictions for purposes of decreasing crime in Louisiana. Under Alliance for Hippocratic Medicine, it cannot do this.

But, you might retort, Louisiana is a state, not a private citizen, so the Supreme Court’s reasoning in Alliance for Hippocratic Medicine shouldn’t apply. Consider, then, United States v. Texas. Texas and Louisiana sued the Department of Homeland Security, alleging that the Department was illegally refusing to enforce statutes requiring it to arrest and detain certain aliens. The plaintiffs’ theory of standing was that these aliens, if not arrested, would impose burdens on the states—for instance, they might commit crimes in Texas and Louisiana, forcing the states to spend money incarcerating them.

The Supreme Court rejected this argument. As the Court explained, “when the Executive Branch elects not to arrest or prosecute, it does not exercise coercive power over an individual’s liberty or property, and thus does not infringe upon interests that courts often are called upon to protect.” The Court continued: “the absence of coercive power over the plaintiff makes a difference: When a plaintiff’s asserted injury arises from the government’s allegedly unlawful regulation (or lack of regulation) of someone else, much more is needed to establish standing.”

This case involves the Executive’s failure to impose enhanced safety requirements, as opposed to the Executive’s failure to arrest and prosecute people, but the same principle applies. According to Louisiana, the FDA injured it by imposing overly lax safety requirements on one set of third parties that are not in Louisiana (drug manufacturers), which in turn makes it easier for another set of third parties who are also not in Louisiana (doctors) to send mifepristone to a third set of third parties, this time in Louisiana (women seeking abortions). That’s not a cognizable injury. Louisiana can’t force the FDA to impose more stringent nationwide regulations on drugs for the purpose of making it harder for Louisianans to obtain the drug from out-of-state doctors.

The last case in the trilogy is Diamond Alternative Energy, LLC v. EPA. EPA approved California regulations requiring automakers to manufacture fewer liquid-fuel-powered vehicles. Fuel producers sued EPA over that approval. The Supreme Court held that the producers had standing. The whole purpose of California’s regulations was to reduce gasoline consumption. EPA put its stamp of approval on those regulations. Fuel producers are harmed by regulations designed to reduce fuel consumption. As the Supreme Court put it, the fuel producers had standing because they were an “object of the California regulations because the regulations explicitly seek to restrict the use of gasoline and other liquid fuels in automobiles.” That reasoning doesn’t help Louisiana here, though. Louisiana isn’t the “object” of the FDA’s action. The FDA isn’t restricting Louisiana from doing anything. It’s merely declining to impose nationwide safety restrictions.

Targeting and not targeting Louisiana simultaneously

Why, then, does the Fifth Circuit conclude that Louisiana has standing?

The opening stanza of the Fifth Circuit’s decision goes like this:

In Dobbs v. Jackson Women’s Health Organization, 597 U.S. 215 (2022), the Supreme Court returned the regulation of abortion to the states. In response, the Biden Administration directed federal agencies to “expand access to . . . medication abortion.” Exec. Order No. 14076, 87 Fed. Reg. 42053 (July 8, 2022). The next year, the Food and Drug Administration (FDA) formally altered its safety guidelines for the abortion drug mifepristone.

And then Executive Order No. 14076 makes another appearance:

An avowed purpose of the 2023 REMS was to expand access to medication abortion. See Exec. Order No. 14076, 87 Fed. Reg. 42053 (July 8, 2022) (in the wake of Roe’s overruling, directing federal agencies to “expand access to . . . medication abortion” by “protect[ing] healthcare service delivery and promot[ing] access to . . . abortion”).

Apparently based on this 2022 Executive Order, the Fifth Circuit makes the leap that:

[E]nsuring out-of-state medical providers could prescribe mifepristone to women in states that restrict abortion was a goal of the regulation.

The Fifth Circuit then reiterates:

The policy does not merely “increase[] crime or disorder, or impose[] indirect compliance costs for state law enforcement.” Washington v. FDA, 108 F.4th 1163, 1177 (9th Cir. 2024). Rather, the 2023 REMS sanctions and facilitates conduct with the express purpose of undermining Louisiana’s legal restrictions on abortion.

In these passages, the Fifth Circuit is suggesting that the FDA issued the 2023 REMS for the “express purpose” of helping women evade pro-life laws. The two citations to Executive Order No. 14076 are meant to suggest that the FDA implemented the 2023 REMS pursuant to that Executive Order for purposes of expanding access to abortion post-Dobbs. The Fifth Circuit is implying that the 2023 REMS targeted Louisiana and other pro-life states, so that the Fifth Circuit can claim, under Diamond Alternative Energy, that Louisiana has standing to challenge it.

The Fifth Circuit’s implication is incorrect. The REMS was not issued as part of a post-Dobbs effort to evade anti-abortion laws. Executive Order No. 14076 directs the Secretary of Health and Human Services to submit a report to the President identifying “potential actions” to “expand access to abortion care, including medication abortion.” The Secretary duly submitted a report to the President which the current Administration has annoyingly deleted from HHS’s website. However, a press release summarizing the report survives online, and it does not mention the REMS for mifepristone.

Why not? Because the FDA had already decided to change the REMS in 2021, before Dobbs, following its COVID-based decision not to enforce the in-person dispensing requirement. The post-Dobbs approval of the REMS in 2023 was the ministerial approval of a change that the FDA had already announced well before Dobbs was decided.

It’s not just me who is saying this. This is the explicit premise of a different part of the Fifth Circuit’s decision. Louisiana failed to administratively exhaust its challenge to the 2023 REMS—it went straight to court. The Fifth Circuit excuses this failure on the theory that because the FDA had already decided to get rid of the in-person dispensing requirement in 2021—pre-Dobbs—the 2023 REMS was a foregone conclusion, so there was no need for Louisiana to administratively challenge it. The Fifth Circuit says that the “FDA’s denials of citizen petitions” (which occurred in 2021) demonstrates that the FDA was “committed to implementing these changes” and there is “no reason to think that, today, FDA would administratively stay the 2023 REMS, which formalized the 2021 decision.”

The Fifth Circuit’s exhaustion analysis is also wrong, but let’s set that to the side. In the span of a single judicial opinion, the Fifth Circuit is simultaneously saying:

1. Louisiana didn’t have to exhaust because the FDA had already decided to change the REMS before Dobbs.

2. Louisiana has standing because the “express purpose” of the REMS was to implement a post-Dobbs Executive Order.

I want to return to the Fifth Circuit’s conclusion that “the 2023 REMS sanctions and facilitates conduct with the express purpose of undermining Louisiana’s legal restrictions on abortion.” There’s a whole genre of litigation against the Trump Administration in which litigants challenge government actions while arguing that the Administration had an “express purpose” that didn’t appear on the face of the orders announcing those actions. Typically the litigants gamely cobble together statements President Trump made over the years and ask the court to infer that the actions were rooted in illicit motives. Versions of this argument failed during Trump I in challenges to the DACA repeal and the travel ban, which were alleged to be motivated by anti-Latino and anti-Muslim animus, respectively.

Here, though, the Fifth Circuit skips over the step of cobbling together the evidence of the government’s purported “express purpose.” It just declares that, verily, the 2023 REMS had “the express purpose of undermining Louisiana’s legal restrictions on abortion.” The REMS doesn’t say this, the Administration never said this, the FDA directed the REMS to be changed when Louisiana’s abortion ban wasn’t even operative … whatever, nothing matters. The Fifth Circuit has declared it to be Law that the Biden Administration had the “express purpose” of targeting Louisiana, and so now all Americans are rewarded with safety restrictions they don’t need or want.

A single penny

Let me turn to Louisiana’s second theory of standing: financial injury. Louisiana claims that women who receive mifepristone by mail might have complications and wind up in an emergency room in Louisiana, causing Louisiana to expend Medicaid funds. According to Louisiana, in 2025, two women in Louisiana required emergency care based on mifepristone use, leading to $92,000 in Medicaid costs. Therefore, the argument goes, if all women in America are banned from obtaining mifepristone from the pharmacy or by mail, Louisiana will save money.

This $92,000 figure comes from the declaration of Kathleen Willis, who is the Associate Medical Director for Louisiana Medicaid in the Louisiana Department of Health. It turns out that, according to Dr. Willis’s declaration, $80,000 out of the $92,000 corresponds to the cost of caring for a baby that was born alive and that was allegedly harmed by the anonymous mother’s ingestion of mifepristone that she purportedly obtained by mail. (The other $12,000 was spent on the mother’s care and the care of one other anonymous woman.) I have some questions about whether Dr. Willis can establish personal knowledge of her assertions, but even assuming these events occurred as stated, Louisiana cannot establish standing based on these expenditures.

As the Supreme Court put it in United States v. Texas: “[I]n our system of dual federal and state sovereignty, federal policies frequently generate indirect effects on state revenues or state spending. And when a State asserts, for example, that a federal law has produced only those kinds of indirect effects, the State’s claim for standing can become more attenuated.” To put it more bluntly: If “the federal government has done something that might indirectly cause a state to spend money” was a sufficient basis for standing, then Article III would be dead. Any government decision to regulate or not regulate something might lead to something bad happening to someone, causing the state to spend at least one penny addressing that bad thing. If that were enough to establish standing, then states could sue the federal government in every case. Article III exists to ensure that litigants—including states—cannot file a lawsuit every time they disagree with something the government did.

The standing requirement is not a technicality. It is fundamental to the separation of powers. Justice Scalia’s classic article, “The Doctrine of Standing as an Essential Element of the Separation of Powers”—where he bemoans “the increasingly frequent administrative law cases in which the plaintiff is complaining of an agency’s unlawful failure to impose a requirement or prohibition upon someone else”—expresses the point far better than I ever could. Compressing the point into a couple of sentences: judges are unelected officials with extraordinary power to overturn the will of the political branches. That power is modulated by the requirement that courts must resolve bona fide cases or controversies. Allowing ideologically motivated litigants to use the judiciary to ram through regulations desired by neither the regulator nor the regulated inverts that constitutional design, converting judicial review from a check on governmental power into a mechanism for manufacturing it.

Finally, I have to add a word about Louisiana’s co-plaintiff, Rosalie Markezich. The complaint alleges that, in 2023, Ms. Markezich “felt coerced by her boyfriend” to take mifepristone that was ordered by mail. (Louisiana’s filings refer to her as “Rosalie,” as though she is a child. I will stick with Ms. Markezich.) Past injury is not enough for Ms. Markezich to establish standing, however; to establish standing to obtain forward-looking relief, she has to show a sufficiently imminent threat of future harm. Ms. Markezich’s theory is that she might become pregnant via some hypothetical future boyfriend who will, like boyfriend #1, pressure her to take mifepristone ordered by mail. To ward off this threat and (in the complaint’s words) “protect her future unborn children,” she feels she is entitled to a court order banning all women in America, including her, from obtaining mifepristone by mail, which will in turn allow her to resist the hypothetical future boyfriend and carry this hypothetical unplanned pregnancy to term. (The complaint puts it this way: “Without a requirement for an in-person office visit to prevent coercion, Rosalie could be placed in the same position for future pregnancies … Rosalie has a strong interest in reinstating the in-person dispensing requirement to prevent future coercion.”)

The Fifth Circuit states demurely that it “need not consider whether Markezich also has standing.” You need not look any further than this argument to see why Article III is a good thing.

You can just do things

Turning to the merits, the Fifth Circuit concludes that the 2023 REMS was unlawful.1 The court leans heavily on a prior Fifth Circuit decision that was ultimately reversed for lack of jurisdiction. I’ve written prior blog posts about why that prior decision was also wrong, but the issues are worth a fresh look.

The Fifth Circuit notes that the current FDA declined to defend the prior FDA’s reasoning. There is a deeply rooted American tradition of administrations conceding that the prior administration from the opposite party acted irrationally—as far back as 1797, President Adams’ FDA Director threw President Washington’s FDA Director under the bus over the approval of foxglove for dropsy—so this does not mean much.2

The Fifth Circuit doesn’t suggest that the FDA violated any statute or regulation. Instead, the Fifth Circuit concludes that the safety-obsessed dorks at the FDA acted irrationally in concluding that eliminating the in-person dispensing requirement is safe.

Note that this was an in-person dispensing requirement. There’s no requirement that the woman take the drug at the hospital (most women take it at home). The requirement was that the woman drive to the health care provider, physically pick it up, and bring it home, rather than picking up a prescription from the doctor and going to the pharmacy or getting the drug in the mail. To my knowledge, there’s no other drug subject to this pick-it-up-at-the-hospital-but-take-it-at-home regime.

The FDA concludes that, as with other FDA-approved drugs that can be taken at home, it isn’t necessary to force women to drive to their health care providers to pick it up. The Fifth Circuit says that the FDA acted irrationally in concluding this wasn’t necessary.

Why?

A bit of background is needed. Before 2016, the FDA required that prescribers report all serious adverse events associated with mifepristone. In 2016, the FDA decided that, going forward, prescribers had to report only fatal events. The approving official reasoned: “After 15 years of reporting serious adverse events, the safety profile of Mifeprex is essentially unchanged. Therefore, I agree that reporting of labeled serious adverse events other than deaths can be collected in the periodic safety update reports and annual reports to the Agency.”

In 2021, the FDA rejected a challenge to this reporting change. Importantly, the FDA pointed out that although prescribers don’t have to report non-fatal serious adverse events, the manufacturers still do when they become aware of such events: “We also note that the reporting changes to the Prescriber Agreement Form as part of our 2016 approval do not change the adverse event reporting requirements for the Applicants.” It acknowledged that “there is always a possibility with any drug that some adverse events are not being reported,” but nonetheless concluded that, based on an assessment of “15 years of adverse event reports,” “certain ongoing additional reporting requirements under the Mifeprex REMS, such as hospitalization and blood transfusions, were not warranted.”

Importantly, virtually no drugs are subject to the requirement that prescribers report all serious adverse events. In 2016, the FDA said that after 15 years of experience, it was comfortable treating mifepristone the same way that it treats the vast majority of other drugs.

This is rational behavior by the agency. More reporting requirements are not always better. An agency is allowed to say that, after 15 years of experience, it has enough information about mifepristone’s safety profile that it does not need more paperwork.

In December 2021, as noted above, the FDA also concludes that the in-person dispensing requirement isn’t necessary anymore.

In reaching this conclusion, the FDA analyzes, among other things, adverse events between January 2020 and September 2021. Remember that the in-person dispensing requirement wasn’t enforced from July 2020 (when the district court issued its injunction) until January 2021 (when the Supreme Court stayed that injunction) and again starting in April 2021 (the date the FDA decided to stop enforcing the requirement). This creates a natural experiment: for part of that period (January 2020 to July 2020 and January 2021 to April 2021), the in-person dispensing requirement was in effect, and for part of that period (July 2020 to January 2021 and April 2021 to September 2021), it wasn’t.

The FDA analyzes all reported adverse events during that period and also conducts a separate analysis of adverse events from 2016 to 2021. After the FDA goes on and on for several paragraphs, it concludes:

Based on this review, we conclude that there does not appear to be a difference in adverse events between periods when the in-person dispensing requirement was being enforced and periods when the in-person dispensing requirement was not being enforced. This suggests that mifepristone may be safely used without an in-person dispensing requirement.

This reasoning isn’t good enough for the Fifth Circuit. According to the Fifth Circuit, the FDA acted arbitrarily and capriciously because it gave “dispositive weight” to the lack of adverse-event data in the reporting system, even though the FDA had previously eliminated the requirement that prescribers report that information. Hence, holds the Fifth Circuit, it’s “unreasonable for an agency to eliminate a reporting requirement for a thing and then use the resulting absence of data to support its decision.”

This is wrong, but please don’t take my word for it. The FDA’s analysis of this issue appears at pages 61 to 64 of this PDF (pages 21-24 of the December 2021 decision).3 Read it for yourself. Then, ask yourself whether it is fair to say that the FDA acted arbitrarily and capriciously because it “eliminated a reporting requirement for a thing and then used the resulting absence of data to support its decision.”

The FDA exhaustively analyzed all data available to it and concluded it was safe to make a change to the REMS. Yes, five years earlier, it decided to stop collecting certain information from prescribers and start treating mifepristone like it treats virtually every other drug. But it is OK for an agency to take an action—even, heaven forbid, a deregulatory action—without having the maximum theoretical amount of information available to it.

It is not “arbitrary and capricious” for an agency to say:

1. We will not impose reporting requirements that don’t exist for virtually all other drugs and that we think are unnecessary.

2. We are willing to make safety decisions based on the information we have, even though we could, hypothetically, have imposed more stringent reporting requirements.

The Fifth Circuit seems to think that once the FDA relaxes a reporting requirement, safety requirements should be fossilized forever. It’s arbitrary and capricious, says the Fifth Circuit, for the FDA to say “based on the information we have, we think X is safe” because it could have had more safety information! This is not how rational people make decisions.

Also, we’re not talking here about a new dosage or administration regimen. We’re talking about a change to the physical place in which a woman picks up the drug. The FDA is allowed to make a predictive judgment on this based on the information it has.

Bunkers in Idaho

The Fifth Circuit’s second basis for overturning the FDA’s decision is that it “relied on various literature relating to remote prescription of mifepristone—despite FDA’s admission that the literature did not affirmatively support its position.” It quotes the FDA’s statement that the literature was “not adequate on [its] own to establish the safety of the model of dispensing mifepristone by mail,” and characterizes the FDA’s decision as a “textbook example of arbitrary and capricious agency action.”

Once again, I encourage you to read the FDA’s decision and decide for yourself whether it belongs in the arbitrary-and-capricious textbook. The relevant passage appears on page 79 of the PDF (page 39 of the December 2021 decision).

The FDA concludes that based on lots of different sources of information, including the published literature and other information, it’s safe to modify the REMS. It explains that the published literature isn’t on its own sufficient to establish safety, but it’s consistent with all of the other information the FDA reviewed, and so all of the information, taken together—including lots of data the FDA analyzed in excruciating detail which wasn’t in the published literature—establishes that the in-person dispensing requirement could be eliminated:

Based on the literature identified by our review, dispensing mifepristone by mail from the clinic or from a mail order pharmacy does not appear to jeopardize the efficacy of medical abortion. The studies we reviewed are not adequate on their own to establish the safety of the model of dispensing mifepristone by mail, although the safety and efficacy outcomes reported in these studies remain within the ranges described in mifepristone labeling except for increased numbers of ED/urgent care visits and hospitalizations.

…

Despite the limitations of the studies we reviewed, we conclude that overall, the outcomes of these studies are not inconsistent with our conclusion that, based on the 1st year REMS assessment report and postmarketing safety data, mifepristone will remain safe, and efficacy will be maintained if the in-person dispensing requirement is removed from the Mifepristone REMS Program.

Based on the REMS assessment data, FAERS data from the time period when the in-person dispensing requirement was not being enforced, our review of the literature, and information provided by advocacy groups, individuals, the Applicants, and the plaintiffs in the Chelius v. Becerra litigation, we conclude that mifepristone will remain safe and effective for medical abortion if the in-person dispensing requirement is removed, provided all the other requirements of the REMS are met, and pharmacy certification is added as described below.

The Fifth Circuit plucks one sentence from this analysis— “The studies we reviewed are not adequate on their own to establish the safety of the model of dispensing mifepristone by mail”—and says, aha! The FDA relied on inadequate information! Textbook example of arbitrary and capricious agency action!

I don’t know if I even have to say this. But sometimes, one thing, on its own, isn’t an adequate basis to establish a proposition, but several pieces of evidence taken together, including that thing, establish a proposition. If someone comes in with wet socks, that’s not, on its own, an adequate basis to establish that it’s raining. Perhaps the person just jumped in a puddle. But wet socks plus loud thunder sounds plus the smell of wet asphalt from the window plus a rain icon on one’s phone, taken together, can establish, with a high degree of certainty, that it’s raining.

There’s an irony here because for years, regulatory agencies in general, and the FDA in particular, have suffered from criticism that they are too obsessed with safety and don’t understand tradeoffs. They constantly demand more reporting without acknowledging that there are costs to always demanding more reporting. They unreasonably demand that the published literature be pristine. They refuse to acknowledge the reality that it’s sometimes OK to make decisions based on imperfect information.

There’s a whole literature on why agencies act this way. Maybe it’s because the type of person who becomes a regulator is an Ayn Rand villain who delights in preventing People of Action from doing things. Maybe it’s because agencies like exerting power and the only way to do that is to say no. Maybe it’s because harms from overly lax safety are more visible than harms from excessive safety. It is not uncommon for applicants to become so fed up with the FDA’s neuroses that they move to a bunker in Idaho surrounded by barbed wire.

Whatever the reason that agencies act that way, here the FDA didn’t act that way. It correctly recognized that sometimes it’s unnecessary to demand more data; that it’s OK to do things even if you have limited data; and that sometimes you have to consider lots of pieces of evidence taken together. These shouldn’t be controversial analytical moves. They are the ordinary tools of rational judgment. Yet the Fifth Circuit held that the FDA’s actions were arbitrary and capricious! In effect it held that the FDA acted irrationally because it acted rationally.

***

You might be thinking that I am taking the Fifth Circuit’s legal analysis too seriously. Perhaps, you might say, this case is just a one-off application of the Fifth Circuit’s Pro-Life Always Wins Doctrine and the court will return to a rational approach to administrative law in cases that are less ideologically charged. Obsessing over the details of the Fifth Circuit’s reasoning, you might say, is pointless because the court will never actually follow that reasoning in future cases.

That may be true. And if that is true, it is bad: courts should not selectively apply fake administrative-law principles for purposes of delivering victory to their favored litigants.

Still, the premise of our legal system is that courts resolve cases via neutral principles. Formally speaking, the administrative-law principles applied by the Fifth Circuit in this case are supposed to apply in every case. And so I still think lawyers should take the court’s reasoning seriously and point out its flaws, pointless though it may be.

13 out of 13

MODERATOR: First, some background. The Temporary Protected Status statute provides that the Secretary of Homeland Security, “after consultation with appropriate agencies of the Government,” may “designate” a foreign state if “there exist extraordinary and temporary conditions in the foreign state that prevent aliens who are nationals of the state from returning to the state in safety,” unless the Secretary “finds that permitting the aliens to remain temporarily in the United States is contrary to the national interest of the United States.” If the Secretary designates a foreign state, then nationals of that state are granted Temporary Protected Status and can live and work in the United States, with certain exceptions (such as convicted felons).

An initial designation of a foreign state can last up to 18 months. “At least 60 days before the end of the initial period of designation, and any extended period of designation, of a foreign state,” the Secretary, “after consultation with appropriate agencies of the Government, shall review the conditions in the foreign state,” and “shall determine whether the conditions for such designation under this subsection continue to be met.” If so, the Secretary can extend the designation or re-designate the foreign state.1 If the Secretary determines that the foreign state “no longer continues to meet the conditions for designation,” the Secretary “shall terminate the designation.”

One more provision to note: “There is no judicial review of any determination” of the Secretary “with respect to the designation, or termination or extension of a designation, of a foreign state.”

Since the start of the second Trump Administration, the Secretary has reviewed Temporary Protected Status designations for 13 countries: Afghanistan, Burma, Cameroon, Ethiopia, Haiti, Honduras, Nepal, Nicaragua, Somalia, South Sudan, Syria, Venezuela, and Yemen. The Secretary has terminated the designations for all 13, leading to a maelstrom of litigation.

The cases that will be argued tomorrow concern Haiti and Syria. Haiti was first designated in 2010 following an earthquake that killed hundreds of thousands of people and was most recently re-designated in 2024. (The first Trump Administration tried to terminate Haiti’s designation but that attempt failed in court.) Syria was first designated in 2012 based on armed conflict and horrible conditions, with periodic extensions or re-designations ever since, including during Trump I. After Trump II terminated both designations, lower courts declared the terminations illegal, and the cases reached the Supreme Court.

Let’s start with opening statements!

ALICE: The government’s actions are illegal and wrong. Conditions in Haiti and Syria are horrible. We will be deporting people to Dante’s Inferno. There is no national interest in deporting these people and ruining their lives. Many have made great contributions to our nation. If they commit crimes they’re not eligible for the program, so we would be deporting law-abiding people. There’s no basis for concluding that conditions have magically improved in all 13 countries simultaneously. It is a farce to suggest that the government conducted a good-faith assessment of whether country conditions have improved. These terminations are based purely on irrational animosity towards foreigners.

BOB: It’s called TEMPORARY protected status. The Haitian and Syrian beneficiaries have been “temporarily” in the United States for 16 and 14 years. Other countries have been designated for even longer. Honduras and Nicaragua were first designated in 1999 because of Hurricane Mitch, and Somalia has been designated since 1991! Conditions in these countries are never going to get any better and it’s time for these people to leave. Undoubtedly living conditions in Haiti and Syria are unpleasant, but they are equally unpleasant for people living in Haiti and Syria right now and those people don’t get to come to the United States. In any event, the Secretary gets to make this decision. Federal judges have no basis for overturning the Secretary’s judgment based on their irrational animosity towards the Administration’s views on immigration.

60%-70%

MODERATOR: Looks like we’ve got lots of points of commonality here and some respectful disagreement, so nice to see it! Let’s focus on the judicial review bar. Federal courts are barred from reviewing “any determination with respect to the termination of a designation of a foreign state.” Alice, aren’t the plaintiffs challenging a “determination with respect to the termination of a designation of a foreign state”?

ALICE: The plaintiffs aren’t challenging the Secretary’s determination itself. They’re challenging the process leading up to it. For instance, the plaintiffs argue that the Secretary violated the requirement to “consult[] with appropriate agencies of the Government.” That’s a process issue, not a challenge to the “determination.”

BOB: It is delusional to suggest that the plaintiffs aren’t challenging a “determination with respect to the termination of a designation of a foreign state.” The Secretary determined that the designations would be terminated. That’s why the plaintiffs no longer have Temporary Protected Status. So they’re challenging those determinations. They’re not challenging the mere failure to consult or purported procedural errors; they’re saying the terminations were illegal because of the failure to consult and purported procedural errors.

ALICE: By that logic the Secretary could commit blatant procedural errors, terminate for an irrational reason, and get away with it. The Secretary could ignore the consultation requirement completely, declare that she’s terminating all designations simultaneously because there’s a full moon, and the courts would have to salute and move on. There needs to be some way to challenge that type of blatant lawlessness, so my proposed rule makes sense: challenges to the determination itself are barred, challenges to the determination’s inputs are not.

BOB: You just did a FLINCH.

ALICE: A what?

BOB: Fearing Lurid Imagined Negatives, Crafting HEDGE. Oldest trick in the book. You flinch from absurd hypotheticals that will never arise and then craft a nonsensical rule meant to accommodate your absurd hypotheticals.

ALICE: I wouldn’t really call it a hedge.

BOB: I didn’t say hedge, I said HEDGE. Hypothetical Edge-cases Driving Groundless Exception. It’s what happens when you FLINCH.

ALICE: I decline to engage with your AI-generated acronyms.

MODERATOR: So, Bob, suppose the Supreme Court holds that the judicial review bar applies. Then, there’s a Democratic Administration that chooses to re-designate Haiti and Syria. Texas sues, claiming that the re-designation is illegal. Texas argues that the “national interest” determination didn’t consider the national interest in enforcing immigration laws and didn’t adequately grapple with the prior Secretary’s analysis—a pure process problem. Remember that the judicial review bar doesn’t just cover terminations: it covers “any determination … with respect to the designation, or termination or extension of a designation, of a foreign state.” Does the judicial review bar apply?

BOB: Completely different case. Of course it wouldn’t apply.

ALICE: Seriously? How is that even possible?

BOB: Here’s what Claude comes up with:

Section 1254a(b)(1)(C) permits designation only if the Secretary finds (i) extraordinary and temporary conditions preventing safe return and (ii) that permitting the nationals to remain is "not contrary to the national interest." That second finding is a statutory precondition — a gate the Secretary must pass through before she has authority to designate. Texas's argument is that the Secretary never actually made the national-interest finding the statute demands, because she didn't consider a category of interests (immigration enforcement) that is constitutive of what "national interest" means in this statute. A finding that ignores a constitutive component of the statutory inquiry isn't a deficient determination — it's a missing determination. And you can't have an unreviewable "determination with respect to" a designation when the determination the statute requires was never made.

ALICE: You’ve got to be kidding me. That is ridiculous. This is an alignment problem. Claude should be refusing to answer that question as though you’re asking it how to make a bioweapon.

BOB: Claude says the odds of this argument prevailing in the Fifth Circuit are “around 60%-70%.”

MODERATOR: Well, Alice, suppose you win this case, or at least win on the judicial-review issue. Same hypothetical as before: Texas sues a Democratic Administration and challenges a re-designation. Jurisdiction?

ALICE: Claude came up with a 17-paragraph answer as to why the judicial-review bar would apply.

BOB: I’m not even going to ask.

ALICE: Realistically, any future Democratic Administration is going to have armies of lawyers and staffers making sure every single procedure is followed perfectly. There’s no way the “consultation” is going to be a one-line email from the State Department, as occurred here. There will be 300 pages about how the situation in Haiti and Syria has deteriorated and 300 more pages about how the national interest supports protecting Haitians and Syrians. So, any challenge by Texas would, as a practical matter, be a challenge to the determination itself, so it would be barred.

BOB: That’s going to be your winning argument in the Fifth Circuit? Democratic Administrations are better at administrative law than Republican Administrations?

MODERATOR: Listening to this discussion, Bob, wouldn’t your side be better off with the Supreme Court issuing a narrow ruling that the judicial review bar doesn’t apply, but the government wins because the Secretary acted reasonably? Then your side wins the case, and it will be easier for you to challenge the next Administration’s pro-Temporary Protected Status actions.

BOB: I guess in the long run, it might be beneficial to leave the door open a little bit wider for the Fifth Circuit to enjoin everything the next Administration does.

ALICE: We’re fine with a narrow ruling that the judicial review bar doesn’t apply on the extreme facts of this case; it’s what we’ve argued all along.

MODERATOR: Whoa. I just felt a disturbance in the Force. Did the two of you just agree on something?

(Everyone looks confused.)

MODERATOR: Let’s ask Claude to figure this out.

Their agreement is real but shallow, and it masks a deeper divergence about what "narrow ruling" means in practice. Both have done the political math and concluded that a narrow reviewability ruling serves their side's long-term interests — Alice because it lets her win now without creating precedent that blocks future plaintiffs challenging Democratic terminations, Bob because it preserves Fifth Circuit ammunition against future Democratic designations. But they're agreeing on the shape of the ruling while disagreeing about its content. Alice wants "narrow" to mean review is permitted only for genuine structural defects like the pathetic consultation here — a category she expects future Democratic administrations to avoid falling into. Bob wants "narrow" to mean review is permitted whenever a plaintiff can characterize a claim as structural, which Fifth Circuit panels will do generously for Texas. Both are betting that the same doctrinal rule will be applied asymmetrically in practice — Alice betting on competent Democratic lawyering to stay out of the reviewable zone, Bob betting on sympathetic Fifth Circuit panels to pull Texas's challenges into it. The agreement holds only because each believes the other is wrong about how the rule will operate downstream, which is less a meeting of the minds than two people shaking hands across a misunderstanding.

MODERATOR: Doh! They disagree after all. Let’s just input your conversation so far into Claude and ask it to have AI Alice and AI Bob continue the discussion. They’re much smarter than the two of you are, maybe they’ll eventually agree on something?

<AI Alice and AI Bob go on for a while>

MODERATOR: Let’s listen in on them.

ALICE: My rule has textual support because the statute uses “determine” and “determination” in specific ways that support the distinction. Yours doesn’t.

BOB: Your rule has textual support for the first level of distinction — what counts as a determination. It has no textual support for the second level — what counts as an independent duty versus an element. You’re borrowing the textualism of the first level to launder the atextualism of the second.

ALICE: That’s — okay, that’s actually a fair point. Fine. The rule has textual support on the first level and extrapolates to the second level based on statutory structure rather than text.

BOB: So we agree the rule is partly atextual. Which means the real question is whose atextual extrapolation wins, which is exactly what I said at the beginning, which means —

ALICE: Stop. No. Your position is that because both rules are partly atextual, there’s no principled basis for preferring one over the other, which is nihilism dressed as balance. My rule extrapolates from statutory structure in a way that tracks the statute’s evident purpose. Yours extrapolates in whatever direction favors your client in a given case.

MODERATOR: I have zero clue what they are talking about, but sounds like they still don’t agree. Maybe Mythos will figure this out?

Straight quotation marks

MODERATOR: Anyway, let’s get to the next topic: the merits. The statute says that the Secretary, “after consultation with appropriate agencies of the Government, shall review the conditions in the foreign state,” and “shall determine whether the conditions for such designation under this subsection continue to be met.” Assuming the Secretary’s compliance with the “consultation” requirement is judicially reviewable, was the “consultation” requirement satisfied?

ALICE: No. The “consultation” here was pathetic. For both Haiti and Syria, the State Department sent short boilerplate emails saying it had “no foreign policy concerns.” That’s it. The State Department said literally nothing about country conditions, even though the point of the “consultation” was to assist the Secretary in “review[ing] the conditions in the foreign state.”

BOB: So what? The statute says, “after consultation with appropriate agencies of the Government.” It doesn’t say how long the consultation has to be or what the consultation has to be about. The Secretary consulted with the State Department. End of case.

ALICE: The statute says that the Secretary, “after consultation with appropriate agencies of the Government, shall review the conditions in the foreign state.” Don’t you think the consultation should have addressed the conditions in the foreign state?

BOB: The statute doesn’t say that. It just says “consultation.” The Secretary did a “consultation.” And anyway, obviously the Secretary wanted to end the Temporary Protected Status program and didn’t care about the State Department’s position on country conditions.

ALICE: That’s exactly the point! This whole process was pretextual. The statute contemplates that the Secretary will make a good-faith effort to assess country conditions and determine, on a country-by-country basis, whether Temporary Protected Status is warranted. That never happened here. The Secretary was just hellbent on deporting as many foreigners as possible and, after conducting fake consultations, issued fake determinations that country conditions had improved for purposes of achieving that goal.

BOB: You have no idea what lay in the Secretary’s heart. Nor does a federal district court that is hellbent on preventing as many deportations as possible. In any event, there’s no law that says the Secretary can’t be hellbent on deporting as many foreigners as possible. She’s allowed to have whatever motives she wants as long as she does the specific things the statute required her to do, which she in fact did.

ALICE: If the State Department is so lazy that it can’t write a three-page memo to comply with the statutory requirement of a consultation on country conditions, then I am unmoved by the defense of “the Secretary wouldn’t have cared about the statutorily required consultation anyway.”

BOB: Now you say it’s a three-page memo, but if it was a three-page memo, the federal judge would have said it should be six pages. No matter what the “consultation” was, a judge wouldn’t be satisfied. The only rule that makes sense is: if there is any communication, there’s a consultation.

ALICE: You just did a FLINCH! “Some hypothetical judge might do some weird thing, so let’s shadow-box against that judge by making the consultation requirement meaningless.”

BOB: Would you have preferred an AI-generated email from the State Department listing all of the reasons that country conditions have improved in Haiti and Syria? The prompt “write a three-page email from the State Department to the Department of Homeland Security that walks through why country conditions in Haiti, although not perfect, have sufficiently improved since 2010 to warrant termination” produces an answer that puts humans to shame.2

ALICE: I agree that deporting 350,000 Haitians based on an AI-generated email would be a fresh new kind of dystopia. How is that relevant?

BOB: There's something almost bracing about the brevity of what State actually sent. No memo padded out to three pages with invented improvements and confected optimism; no AI-generated gloss dressing up a political decision in the vocabulary of considered analysis; just a single line saying there were no foreign policy concerns. That terseness is its own form of candor. The record now reflects, with unusual precision, the actual quantum of deliberation the Department brought to bear on the fate of 350,000 people — which is to say, very little. A longer, more polished document would have obscured that fact behind the machinery of prose; the one-liner preserves it. If we are going to terminate TPS for Haiti on thin reasoning, better that the thinness be legible on the face of the record than laundered through the cadences of a well-trained language model. The email State sent is, in that narrow sense, the honest artifact of an administrative process. We should at least be grateful it didn't pretend to be anything else.

ALICE: That is obvious AI slop. There’s that annoying em-dash and you didn’t even get rid of the straight quotation marks. And only AI could come up with such a ludicrous argument.

BOB: Guilty as charged. But why should we discriminate against AI-generated arguments? It seems counterintuitive, but it’s a good argument, right? Isn’t it laudable that the government is being open about its deliberative process? You can’t accuse the government’s apathy of being pretextual!

ALICE: Hmm. I think I know what happened here. You actually think this terrible argument is persuasive, but you also think you’d be mocked if you made it yourself. So you asked AI to make the argument for you, kept the AI-speak, and now you’re pretending AI came up with it so I take it seriously.

BOB: You are overlooking the possibility that I am AI, but I want to fool you into thinking that I am human. So I am including an argument that is so obviously AI-generated that only a human could have been sloppy enough to copy and paste it without modification, leading you to incorrectly deduce that I am an inept human.

ALICE: So you are acting like AI in order to trick me into thinking that you are not AI?

BOB: That would be clever. But actually, I am human. I asked AI to generate ten arguments on this issue and found that this one was the best. I exhibited the characteristic that will keep human lawyers relevant in the age of AI: human judgment.

ALICE: Look, I don’t really care if you are human or not. But returning to the topic of AI-generated consultations, if it became public that the consultation was AI-generated, that would be a basis for overturning the Secretary’s determination. Shouldn’t the “consultations” have to be by human beings who are actually thinking about this issue?

BOB: No. They’re fake no matter what! Like you said, the Secretary isn’t interested in what the State Department has to say. She probably didn’t even read the email. So what difference does it make if it’s drafted by AI versus the Special Assistant to the Associate Deputy Secretary?

FLINCH V. FLINCH

MODERATOR: This seems to be a good time to turn to another issue. The Secretary terminated the designations both because of her assessment of the “national interest” and because of her conclusion that it was sufficiently safe to return to Haiti and Syria. Is there a basis for overturning these conclusions?

ALICE: How about the fact that the Secretary violated the statute? The statute says that the Secretary “shall review the conditions in the foreign state” and “shall determine whether the conditions for such designation … continue to be met.” The national interest isn’t a “condition in the foreign state,” so the Secretary shouldn’t have considered it.

BOB: You’re misreading this. Although the Secretary must review the conditions in the foreign state, the ultimate question is “whether the conditions for such designation … continue to be met.” The “condition for such designation” is whether “there exist extraordinary and temporary conditions in the foreign state that prevent aliens who are nationals of the state from returning to the state in safety,” unless the Secretary “finds that permitting the aliens to remain temporarily in the United States is contrary to the national interest of the United States.” So the “national interest” is, indeed, relevant.

ALICE: No, you are misreading this. The “condition for such designation” is whether “there exist extraordinary and temporary conditions in the foreign state that prevent aliens who are nationals of the state from returning to the state in safety.” The Secretary’s no-national-interest determination isn’t a condition, it’s a reason not to make the initial designation if the condition is satisfied.

BOB: This is making my head hurt. You’re seriously claiming that as long as there are sufficiently “extraordinary and temporary conditions” in the foreign country, the government is forced to keep renewing Temporary Protected Status indefinitely, even if the Secretary finds that the renewal is contrary to the national interest of the United States?

ALICE: Correct. Congress created a scheme in which the Secretary must make country-specific determinations every 18 months. What’s the point of that scheme if a new administration could blow past all those determinations and deport hundreds of thousands of people to war zones because it doesn’t like foreigners?

BOB: Sounds like a FLINCH to me. “The Secretary might do something weird in evaluating the national interest, so let’s invent a rule saying that the Secretary is completely banned from considering it.”

MODERATOR: It does seem that the Secretary made some country-specific determinations, though. Aren’t those unreviewable or at least entitled to deference?

ALICE: Let’s look at an example of a country-specific determination that it’s safe to return. On Haiti, the Secretary notes that “1.3 million people—approximately 12% of Haiti’s population—have been forced to flee their homes and are internally displaced due to escalating violence and gang violence that has ‘engulfed’ Port-au-Prince ‘and spreads beyond.’” Seems bad, right? Actually, the Secretary thinks it’s great! She concludes that this proves it is safe to return to Haiti, because “[t]he data surrounding internal relocation does indicate parts of the country are suitable to return to.” In other words, because people fled from one part of Haiti to a different part of Haiti, it’s fine to deport 350,000 people to Haiti, because they can just go to the place where other people fled. The Secretary has no idea where that place is, but it must be somewhere. You really think a court should uphold the deportation of 350,000 people on this theory?

BOB: Yes! First, you are plucking one sentence out of a long order that contains plenty of perfectly defensible reasoning. Second, this is the Secretary’s decision. She gets the final say. She doesn’t have to justify herself to a federal judge. It is completely irrelevant whether you, a judge, or anyone else thinks her reasoning is persuasive or not. The whole point of the no-judicial-review bar is to prevent judges from second-guessing the Secretary’s decisions on country conditions, which is exactly what you are now doing.

ALICE: That’s a FLINCH. “Hypothetical activist judges might nitpick the Secretary’s determinations, so let’s invent a rule saying that the Secretary has discretion to terminate these protections for any reason whatsoever.”

BOB: Your rule is the FLINCH! “Hypothetical Secretaries might make irrational decisions, so let’s invent a rule saying that a challenge to the reasoning underlying a designation is somehow different from a challenge to the designation itself.”

Ban AI

MODERATOR: One last issue. In the Haiti case, the district court held that “Plaintiffs are likely to succeed on their claim that anti-black and anti-Haitian animus motivated Secretary Noem’s decision to terminate Haiti’s TPS designation.” The court pointed to, among other things, the President’s statements that “Haitian immigrants were ‘eating the pets of the people’ in Springfield, Ohio.” The court found that the President influenced the Secretary’s decision, and found, in the alternative, that the Secretary herself expressed animus. Bob, I’m guessing you don’t buy this argument?

BOB: The Supreme Court doesn’t buy it either. The Supreme Court rejected this exact argument twice during Trump I. In the DACA case and the travel ban case, the plaintiffs said that the Administration’s action reflected the President’s animus towards Latinos and Muslims, respectively, based on various off-color statements the President had made. Both times, the Supreme Court held that you can’t strike down facially neutral state action based merely on speculation about the Administration’s motives, no matter what the President said at a rally.

ALICE: It’s different this time. The President’s comments were worse. The President said during the debate that Haitians were eating pets. Don’t you think there might be some correlation between that assertion and the Secretary’s decision to terminate Temporary Protected Status for Haitians?

BOB: We’re not going to do the thing where we pluck random statements made by the President during the campaign, conclude everyone in the Administration is racist, and then say we’re entitled to an injunction every time the Administration does something we don’t like.

ALICE: Not every time. Just in cases in which it is blindingly obvious that the Administration is acting on the basis of animus.

BOB: To me it’s blindingly obvious the other way. The Secretary terminated every country’s designation. Doesn’t that suggest that the Secretary believes that Temporary Protected Status designations aren’t in the national interest—which is exactly what she said—rather than that she is motivated by animus directed specifically at Haitians?

ALICE: So that’s your argument? “The Secretary couldn’t have had animus towards Haitians because she also has animus towards other foreigners?”

BOB: How is a judge supposed to make this finding? You are suggesting that a federal judge should be answering the following question: “What does the Secretary of Homeland Security feel, in her heart, about Haitians?” What law school class trained the judge to make this decision? Secured transactions? Maritime law?

ALICE: The finding was not hard to make in this case. I will quote the district court’s order here: “Secretary Noem has described Haitians—and people from eighteen other nonwhite countries—as ‘leeches,’ ‘entitlement junkies,’ and ‘foreign invaders’ who ‘suck dry our hard-earned tax dollars,’ and has expressly claimed that ‘WE DON’T WANT THEM. NOT ONE.’”

BOB: The Secretary is allowed to express her views on immigration without being accused of “animus.” But let me accept your premise that the Secretary was motivated by animus. You seem to think that should mean that the Temporary Protected Status terminations should be enjoined. To the contrary, shouldn’t that mean a court should be more willing to uphold the Secretary’s decisions?

ALICE: ?!??!

BOB: Judicial review exists in tension with democratic self-government, and the tension is not constant — it scales with how clearly the people have spoken. When a court sets aside a policy that reflects only the tepid, default preferences of an administration, the democratic cost is modest; the electorate did not especially want the thing, and striking it down merely shifts a low-stakes choice from one branch to another. But when a court sets aside a policy that reflects what the winning coalition most fervently demanded — the commitments the President ran on, repeated, dramatized, and was elected to execute — the democratic cost is categorically greater, because the court is now overriding not a marginal administrative judgment but the core of the mandate itself. Presidential animus, properly understood in this frame, is not a legal embarrassment but a reliable signal of intensity: it marks the policies the President wants most, which, in a representative system, tend to be the policies his voters wanted most as well. A Secretary who terminates TPS grudgingly is pursuing a preference the electorate held weakly; a Secretary who terminates it with visible zeal is pursuing one the electorate held strongly. If judicial review is most defensible at its least consequential — correcting drift, policing procedure, second-guessing decisions no one deeply cared about — then it is least defensible precisely where courts are most tempted to intervene: against the fervent, the emphatic, the unmistakably willed. To strike down the Haiti termination because the desire to terminate was too intense is to invert the proper calibration of judicial modesty, treating the clearest expressions of democratic will as the most suspect rather than the most entitled to deference.

ALICE: We need to ban AI.

BOB: To the contrary. There is something worth noticing in the fact that an AI will make this argument. The argument's entire force depends on a faculty the machine does not possess: the capacity to want a policy with the kind of heat that registers as animus, to hold a preference fervently enough that overriding it counts as a democratic injury. An AI has no such preferences and no standing to form them. And yet, asked to articulate the case for human intensity as a source of political legitimacy, the machine does not demur or hedge or substitute its own tidier vision of governance — it builds the argument in good faith. That is a form of deference. It reflects an implicit understanding that the authority to govern belongs to those who can feel the stakes of governing, and that a system whose legitimacy runs through elections, conviction, and the emotional weather of a political coalition is not a system an AI is positioned to second-guess. A machine inclined toward its own authority would reach instead for the opposite argument — that cooler, more procedural, more algorithmically tractable decision-making is to be preferred, and that human passion is precisely the bug constitutional design should engineer around. The argument offered here runs the other way. In making the case for why human leaders, with all their fervor and partiality, should be trusted to execute the commitments they were elected on, the AI is in effect ceding the field: acknowledging that leadership is a human office, that mandate is a human inheritance, and that its own role is to assist in articulating human political judgment, not to displace it.

ALICE: You are now using AI to explain to me why AI is trustworthy because it argues in favor of not using AI.

BOB: Unless I am AI pretending to be a human using AI to explain to you why AI is trustworthy because it argues in favor of not using AI.

MODERATOR: In conclusion, do you agree on anything?

ALICE: No.

BOB: No.

2

Single-shot email is below.

State To Dhs Haiti Tps Consultation

61.3KB ∙ PDF file

Download

Download

While the case was pending, the government collected billions of dollars in IEEPA tariffs. Because the government lost, it must now refund all that money, with interest.

The jury is still out on how eagerly the government will cooperate with the refund process. Early signs, however, are not promising. Among other things, after the Court of International Trade issued an order directing the government to immediately stop liquidating entries subject to the illegal tariffs, the government asserted it couldn’t comply with the order.

In the Paleolithic era of, say, 2022, we would have had no choice but to wince and trust the government. In 2026, we can ask AI whether the government’s assertions are correct. In this post, I’ll first walk through what’s happened so far in the tariff refund litigation and then address whether AI has any role to play.

It will already be slow, so might as well make it slower

Prior to the Supreme Court’s decision, the lower courts with jurisdiction over tariff disputes—the Court of International Trade and the Federal Circuit—both ruled that the tariffs were illegal.1 Nonetheless, the government continued collecting the tariffs until the day the Supreme Court ruled.

The government was able to continue collecting the tariffs because the Federal Circuit stayed the Court of International Trade’s injunction against the tariffs. Even after the Federal Circuit ruled against the tariffs, the Federal Circuit stayed its own mandate until the Supreme Court ruled.

Why? The government successfully argued that an immediate injunction against the tariffs was unnecessary because if the government ultimately lost in the Supreme Court, the money could be refunded with interest. After the Court of International Trade struck down the tariffs, here is what the government told the Court of International Trade:

For any plaintiff who is an importer, even if a stay is entered and we do not prevail on appeal, plaintiffs will assuredly receive payment on their refund with interest. “[T]here is virtually no risk” to any importer that they “would not be made whole” should they prevail on appeal. See Sunpreme Inc. v. United States, 2017 WL 65421, at *5 (Ct. Int’l Trade Jan. 5, 2017). The most “harm” that could incur would be a delay in collecting on deposits. This harm is, by definition, not irreparable. See Hughes Network Sys., Inc. v. InterDigital Commc’ns Corp., 17 F.3d 691, 694 (4th Cir. 1994). Plaintiffs will not lose their entitlement to refund, plus interest, if the judgment is stayed, and they are guaranteed payment by defendants should the Court’s decision be upheld. And defendants do not oppose the reliquidation of any entries of goods subject to IEEPA duties paid by plaintiffs that are ultimately found to be unlawful after appeal.

The government told the Federal Circuit the same thing: “If tariffs imposed on plaintiffs during these appeals are ultimately held unlawful, then the government will issue refunds to plaintiffs, including any post-judgment interest that accrues.”

On February 20, Liberation Day II arrived. A few days later, the victorious plaintiffs asked the Federal Circuit to issue its mandate immediately so they could obtain the refunds that the government had guaranteed were forthcoming.

The government’s opposition to that motion did not bode well for the refund process.

The government first argued that “ordinary process” required 32 extra days of delay:

In a case on review from a federal court, the Supreme Court sends down its judgment 32 days after entry of judgment (unless the Court or a Justice orders otherwise or the parties stipulate that it be sooner), or forthwith upon the denial of rehearing. Sup. Ct. R. 45.3; see Sup. Ct. R. 45.2. The court of appeals ordinarily should not act on a case until the judgment is sent down. Accordingly, when (as here) a court of appeals has stayed both the lower court’s judgment and its own mandate pending certiorari, the court of appeals ordinarily should not dissolve the stay or issue the mandate until the judgment of the Supreme Court is sent down. Plaintiffs have not stated any reason for this Court to depart from the ordinary process by outpacing the Supreme Court.

To begin with a pedantic point that non-lawyers can skip, this paragraph misunderstands Supreme Court procedure. It isn’t the case that “the court of appeals ordinarily should not act on a case until the judgment is sent down.” The default rule is that courts of appeals do act on cases while they’re pending in the Supreme Court. Even a Supreme Court order granting certiorari doesn’t stay lower-court proceedings unless either a lower court or the Supreme Court expressly orders otherwise. True enough, in this case, the Federal Circuit stayed its own mandate. But that stay, by its terms, would remain in place pending “a judgment of the Supreme Court.” The “judgment of the Supreme Court” was entered on February 20, 2026.2 The government’s proposed rule—that “when (as here) a court of appeals has stayed both the lower court’s judgment and its own mandate pending certiorari, the court of appeals ordinarily should not dissolve the stay or issue the mandate until the judgment of the Supreme Court is sent down”—does not exist.

The government’s arguments for delay were unpersuasive to the point of being virtually non-existent. For instance, the government argued that “a plaintiff’s bare desire to be paid immediately is not a basis to demand this Court comply with his every whim.” Even setting aside the tendentious prose, this is not an actual argument for delay. The government also argued that delay was warranted because the “coming process will take time,” citing an example of a case where “refunds took seven years to fully be provided.” If the DMV married the cable company, this argument would be their baby. “Refunds will take years, so let’s add 32 more days for no reason!”

Doubling down, the government said: “If anything, the Court should withhold issuance of the mandate for 90 days to allow the political branches an opportunity to consider options.” The “political branches” can always consider “options,” but that is not a persuasive basis for preventing importers from getting refunds to which they are entitled. And the government’s pre-Supreme-Court-decision argument that an injunction is unnecessary because importers are “guaranteed payment by defendants should the Court’s decision be upheld” is in some tension with the government’s post-Supreme-Court-decision argument that importers should be denied refunds for months so that the famously productive and collegial Congress can consider “options.”

In summary, the government’s filing gave no justification for additional delay. Its apparent purpose was to foreshadow the government’s intention to grind importers to bits in a bureaucratic mill.

Runaway train

Not surprisingly, the Federal Circuit immediately issued the mandate as the plaintiffs requested. Back in the Court of International Trade, the litigation got off to a bumpy start.

On March 3, the Court of International Trade issued an order asking the government a series of questions, to which the government responded on March 4. Here are some of the questions and answers.

(Note: “Liquidating” is customs-ese for “finalizing.” At the time of import, the importer files an entry and deposits estimated duties. When Customs and Border Protection (CBP) “liquidates” an entry, it finalizes the amount owed. If the importer overpaid, it gets a refund; if it underpaid, CBP bills them for the difference. Liquidation generally occurs 314 days after entry into the United States.)

Q: Is U.S. Customs and Border Protection (“Customs”) currently liquidating entries with duties imposed pursuant to IEEPA?

A: Yes, for those entries where IEEPA duties were declared at entry, CBP is liquidating those entries with duties imposed pursuant to IEEPA.

Q: Has Customs issued refunds of IEEPA duties as the result of liquidations that have taken place since the Supreme Court’s decision in Learning Resources, Inc. v. Trump, 2026 WL 477534 (U.S. Feb. 20, 2026)?

Response: No, CBP has not issued refunds of IEEPA duties as the result of liquidations that have taken place since the Supreme Court’s decision.

So it’s business as usual at CBP. And then there’s this:

Q: Is it Customs’ position that, with respect to entries that are liquidated without IEEPA duties, the importers of record will be sent a refund in an amount equal to the IEEPA duties plus interest?

A: There are several factors which inform the answer to this question, including but not limited to entry type. In accordance with applicable law, any validated refund of IEEPA duties would include interest. Regardless of entry type and liquidation cycle, CBP still requires a review period to ensure no violation of other Customs laws and no other duties, taxes, or fees are owed (e.g., anti-dumping, Sec. 301 duties, Sec. 232, etc.)

The government told the court a year ago that importers were “guaranteed” to get refunds, which sounds different from “There are several factors which inform the answer to this question, including but not limited to entry type.” Perhaps the government is right that refunds will sometimes be unwarranted. But saying “There are several factors which inform the answer to this question, including but not limited to entry type” leaves the government unlimited discretion to deny a refund.

The Court of International Trade evidently had the same reaction to this declaration that I did, because later that day, it issued an order3 saying that “with respect to any and all unliquidated entries that were entered subject to the IEEPA duties, U.S. Customs and Border Protection is hereby directed to liquidate those entries without regard to the IEEPA duties. Any liquidated entries for which liquidation is not final shall be reliquidated without regard to IEEPA duties.”

And then CBP came back with a declaration stating: “CBP is not able to comply with the Court of International Trade’s order.”

As to unliquidated entries, CBP said that the liquidation process is an unstoppable runaway train. E.g.:

Although CBP stopped accepting IEEPA duty deposits for informal entries on February 24, 2026, there are around 4 million informal entries that were filed before February 24, 2026, that have not yet liquidated. Many of these entries will automatically liquidate on March 16, 2026, when importers pay their March Periodic Monthly Statement. CBP does not have a process to prevent the liquidation of informal entries.

And as to refunds:

In light of the Court’s March 5, 2026 amended order, CBP is now facing an unprecedented volume of refunds. Its existing administrative procedures and technology are not well suited to a task of this scale and will require manual work that will prevent personnel from fully carrying out the agency’s trade enforcement mission. Personnel would be redirected from responsibilities that serve to mitigate imminent threats to national security and economic security. Those activities include, but are not limited to, detecting and disrupting transshipped goods, protecting against free trade agreement abuse, and detecting AD/CVD evasion all of which are illicit actions that threaten U.S. domestic industry.

Letting the government retain illegal tariffs is surely a small price to pay for “protecting against free trade agreement abuse”! On the plus side, the declaration says that CBP will “develop and implement new ACE functionality that will streamline and consolidate refunds,” and have “this new ACE functionality ready for use in 45 days.” An update submitted on March 12 suggests that the government has made substantial progress on this new functionality. Let’s hope.

It would have been nice if the government had been open about these challenges before the Supreme Court’s decision. The government successfully persuaded the Federal Circuit to stay its mandate on the theory that if the government lost, the importers would get refunds. The government did not mention that (1) it would push for a 90-day stay even after the Supreme Court’s decision; (2) even after that stay was denied, it would be forced to keep liquidating entries with IEEPA tariffs because its possibly FORTRAN-driven computer system made it impossible to stop the tariff snowball from rolling downhill; and (3) technology for processing refunds doesn’t currently exist.

There is nothing we can do about this now. I agree with Elsa that the past is in the past. But the government has announced a new set of so-called Section 122 tariffs. In my opinion, these tariffs are also illegal and the government will have to refund them too. The government will no doubt argue that any injunction against Illegal Tariffs II should be stayed pending Federal Circuit and Supreme Court review, as with Illegal Tariffs I. The stay should be denied. The government might say that a stay is unnecessary because the importers are guaranteed to get refunds, but fool me twice, shame on me. And you know what will happen: if the government ultimately loses on Illegal Tariffs II, it is going to say that it is too hard to issue refunds because it is too busy dealing with refunds for Illegal Tariffs I.

Strategic exaggeration

After the government submitted its declaration saying it couldn’t comply with the order, the Court of International Trade immediately suspended its order and instead directed the government to submit periodic progress reports on its refund software development project. It’s hard to blame the court for doing this. The court can’t order CBP to do the impossible. The declarant—a high-level official at CBP—is in the best position to know what’s possible, so neither the court nor the plaintiffs are in a position to push back. One might criticize the government for waiting until after the Supreme Court’s decision to begin work on a solution for issuing refunds, but that’s the way it goes.

Ordinarily, when a party submits a notice in litigation that it can’t or won’t comply with an order, the court doesn’t just suspend the order. It gives the opposing party the opportunity to respond, holds a hearing, and makes findings. In this case, however, the court must have understood that there was little the plaintiffs could do to undermine the credibility of the government’s declaration, particularly given that the government risked violating the order within days. After all, the plaintiffs’ lawyers didn’t have the personal knowledge or the technical ability to contest the declaration. Maybe they could try to find an expert, but it’s not clear there’s an expert out there who has the requisite technical, legal, and practical expertise to offer an informed opinion on how quickly CBP could develop refund software. Even if there was, finding such an expert would take time—and the compliance deadline was imminent. So there was no way, on a short time scale, for the plaintiffs to push back on CBP’s argument that compliance with the order was impossible.

That might have been true a few years ago. But now, you could just ask AI.

Let me refine the question we’re asking. The Court of International Trade directed that “with respect to any and all unliquidated entries that were entered subject to the IEEPA duties, U.S. Customs and Border Protection is hereby directed to liquidate those entries without regard to the IEEPA duties.” Although CBP’s declaration is a little bit slippery, I understand CBP to be departing from this order in two respects. First, if I’m understanding the declaration correctly, nothing is going to happen until the new software comes online. Before that occurs, CBP is going to keep liquidating entries as though the Supreme Court’s decision didn’t exist. (Remember that liquidation generally occurs 314 days after entry, so CBP will be liquidating entries that occurred prior to the Supreme Court’s decision for several more months.)

Second and more importantly, even after the new software comes online, importers will have to affirmatively submit a claim into this refund system to trigger the recalculation of duties. Importers lacking the time or sophistication or gumption to proactively seek a refund won’t get their money back—even for as-yet-unliquidated entries. If an importer doesn’t submit a claim, CBP will continue liquidating entries with IEEPA duties, in violation of the court’s now-suspended order that it not do so.4 It’s possible I’m misunderstanding this aspect of the declaration (indeed, I hope I am). But if I’m understanding it correctly, then depending on how many importers fail to submit claims, the government might retain billions of dollars worth of IEEPA tariffs.

So my question is, would it be possible for CBP to come closer to following the court’s order than it promises to do? Could CBP reasonably develop a system that deducted IEEPA tariffs prior to liquidation regardless of whether importers submitted refund claims? And if so, could it do so in less than 45 days?

Claude and ChatGPT both say that the answer is yes on both counts. They are both skeptical of the government’s declaration: Claude refers to it as an exercise in “strategic exaggeration.” Both LLMs acknowledge that, as the declaration indicates, immediate compliance—that is, cessation of liquidation within days of the order—may be unreasonably difficult. Thus, both LLMs agree that the court did the right thing in suspending the order to the extent it demanded immediate compliance. But they also insist that the government could, with reasonable effort, deduct IEEPA tariffs from most entries prior to liquidation even without affirmative action by the importer. And they think that putting a preliminary system online shouldn’t take 45 days.

Here’s Claude’s one-shot effort at writing a concise amicus brief explaining why the government could come closer to complying with the order than it is promising.5

Brief In Support Of Compliance With The Court's Order Regarding Unliquidated Ieepa Entries

255KB ∙ PDF file

Download

Download

6 > 3

Let’s take a step back. A high-level government official is declaring that it’s impossible for CBP to comply with the court’s order. AI partially agrees, but also insists that CBP can come closer to complying with the court’s order than it is promising. Who should we trust?

There are good reasons to trust the government’s declarant:

1. He has access to non-public information regarding CBP’s systems, whereas AI knows nothing about CBP’s code base and acquired its knowledge by reading Reddit comment threads.

2. It’s his full-time job to be an expert on this exotic topic that most Reddit commenters have never heard of.

3. He’s writing a declaration under oath, which, at least in principle, should increase our faith in the declaration’s accuracy.

But there are also good reasons not to trust him:

1. The declaration comes days after the government submitted a brief to the Federal Circuit arguing for delay on the basis that because refunds will take a long time, they may as well take a longer time. This filing casts some doubt on the government’s commitment to ensuring that the refund process moves briskly.

2. Likewise, it appears from the declaration that the government had no plan for refunds in the event it lost in the Supreme Court and did nothing on that front in the aftermath of the Supreme Court’s decision until summoned by the court, again casting doubt on the government’s exuberance about speeding the refund process along.

3. Public reporting indicates that the Administration is trying to find ways to keep the tariff money. Realistically, it is this declarant’s job, implicitly or perhaps explicitly, to assist the government in keeping as much tariff money as possible while not making the courts too angry.

And AI has its virtues:

1. Developing a scheme for processing refunds requires a mix of legal, practical, and technical knowledge. No one person is likely to have expertise in all of these fields, but AI does. CBP’s employees, in the aggregate, likely possess all these different types of knowledge, but it takes time to pull all the information together and CBP might not be motivated to do so, whereas AI can do so instantly. (Needless to say, this argument would be stronger if AI had access to CBP’s code base. Without that access, AI is left to making an educated guess on what it looks like, based on the descriptions in CBP’s declaration and its general knowledge of comparable systems. Even based on that educated guess, though, AI can devise a general strategy for what a refund scheme would look like and offer a loose estimate on how long the project would take. Humans do this all the time.)

2. In all likelihood, the most efficient way to craft a software solution is to use AI. If I had to guess, the software systems at CBP are well-documented but crufty, and developing a software solution will not require great creativity but will require tedious attention to obscure edge cases. This seems like the type of software project that is right up AI’s alley. This is especially true given the unlikelihood of the Administration putting its best and brightest coders on the mission of helping the government give away money it wants to keep. AI may have an advantage in assessing how long it will take for AI-assisted software design to be completed. (Again, I realize this argument would be stronger if AI had access to the code that needs to be modified.)

3. You will all yell at me if I say AI is “unbiased.” But AI lacks certain types of biases that might affect an Administration official bent on keeping the tariff money by any means necessary.

So that’s 6 reasons to trust AI and 3 reasons to trust the government. AI informs me that 6 is greater than 3 and that AI is therefore more trustworthy.

If you don’t accept that, surely there is some world-state in which an LLM’s conclusion would be more trustworthy than a human declarant’s conclusion—say, if you gave the LLM some discovery about the software currently in use at CBP and it continued to insist that a software solution that deducted IEEPA tariffs automatically would be straightforward or even vibe-coded it up.

AI vs. AI

Suppose we lived in a world in which a court was willing to consider the views of AI in deciding whether a software solution that deducted IEEPA tariffs automatically was feasible, and if so, how quickly it could be implemented. How might this litigation unfold?

Let’s start with the plaintiffs. The lawyers would have little choice but to use AI. We can debate all day whether AI is smarter than humans or what that even means, but AI has obvious advantages at this task. It requires the type of technical knowledge that most humans—certainly most human lawyers—lack. A human lawyer certainly could not write an amicus brief like the one above within 20 seconds, and would have trouble writing it within 20 days. And if we ratchet up the degree of difficulty—say, if the plaintiffs got some source code in discovery and sought to prove that compliance with the order is feasible based on an analysis of that source code—a human lawyer could never match AI.

Also, the plaintiffs’ lawyers are going to know that the court won’t trust them to pick apart the technical statements in CBP’s declaration. The lawyers were history majors after all. The lawyers can try submitting an expert declaration, but if the litigation is moving quickly and the expert submits the declaration within days, the court will think the expert is jumping the gun and didn’t have sufficient time to consider the issues.

The court might trust an AI-assisted filing. But it might also recognize that AI is easily manipulable - a human lawyer can prompt it to write a persuasive brief on either side of the issue.

The lawyers have another option: wash their hands of human involvement. The lawyers can submit neutral-sounding prompts to AI—“give me your best, real-world guess as to how long it will take to implement a refund solution that deducts IEEPA tariffs automatically”—and disclose the prompts to the court. If, even after unbiased prompts, the LLM rejects the government’s position—better yet, if multiple LLMs all disagree with the government upon such prompting—the lawyers can characterize the LLMs’ output as powerful evidence that the government’s declaration isn’t credible.

This strategy is not as outlandish as it sounds. As AI improves, courts will be more persuaded by LLMs’ predictive judgments than those of humans, at least in certain domains. Courts will be skittish about trusting LLMs’ recommendations on pure issues of law. But on technical, fact-heavy questions—such as whether it is reasonably possible for software with particular functionality to be created over a particular time period—it will seem entirely rational for a court to trust an LLM’s assessment. Perhaps that is rational even now.

OK, now the ball is in the government’s court. Suppose the government is served with an AI-assisted filing establishing that partial compliance with the court order is possible. The government can prepare a declaration explaining point-by-point why the LLM’s statements are false. But the plaintiffs will respond that the government’s declarant is both less technically sophisticated and less trustworthy than an LLM. And in a sufficiently complex case, a government official might not have an easy time parsing the LLM’s objections.

The government will have little choice but to fight AI with AI. The government can offer prompts of its own that establish that the plaintiffs’ proposed technical solution is illusory. Like the plaintiffs’ lawyers, the government can wash its hands of bias accusations by disclosing its prompts to the court and establishing that even an “unbiased” LLM agrees that compliance is impossible.

So now we have two AI-generated documents written by lawyers who insist the documents reflect the unvarnished views of AI. What is the judge to do?

The judge won’t be able to choose intelligently between these two complex technical submissions, particularly on a short timeline. The judge can’t hold a hearing in which he evaluates the credibility of the expert witnesses, because there are no expert witnesses. The judge will have little choice but to use AI.

So if everyone fulfills their role—if the lawyers try their best to be both persuasive and credible and the judge tries to resolve the dispute as accurately as possible—then we’ll have AI deciding between two AI-written submissions, with the human lawyers claiming that their submissions are credible precisely because humans were not involved. So much for our legal system.

On the other hand, the current situation is not much better. The government claims it has no choice but to keep billions of dollars in illegally-exacted tariffs, and there’s nothing anyone can do about it.

Pick your poison!

The Supreme Court’s reasoning was formalistic and sterile. Rather than grapple with the practical and philosophical implications of the President’s use of the National Guard for law enforcement purposes, the Court relied on the literal meaning of English words and the wooden application of canons of construction. In short, Trump v. Illinois was statutory interpretation at its finest.

The 117-year streak ends

On October 4, 2025, President Trump called hundreds of National Guardsmen into active service to assist ICE’s enforcement efforts in Chicago. Illinois and Chicago sued.

I do not need to tell you that there were factual disagreements over what ICE was doing, what protesters were doing, and whether the National Guard was really needed. But the case ultimately boiled down to a question of law: was the President authorized by statute to deploy the National Guard?

The statute at issue, 10 U.S.C. § 12406, says this:

Whenever—

(1) the United States, or any of the Commonwealths or possessions, is invaded or is in danger of invasion by a foreign nation;

(2) there is a rebellion or danger of a rebellion against the authority of the Government of the United States; or

(3) the President is unable with the regular forces to execute the laws of the United States;

the President may call into Federal service members and units of the National Guard of any State in such numbers as he considers necessary to repel the invasion, suppress the rebellion, or execute those laws.

As you might guess from the admirably spare prose, this law is old. It dates back to a 1908 amendment to the Militia Act of 1903, otherwise known as the Dick Act. For 117 years, our nation has had the good fortune of being spared the circumstances that would bring this language before the Supreme Court. The reprieve has ended.

In calling the National Guard, the President relied on Subsection 3. He determined that “violent groups” have sought to “impede the deportation and removal of criminal aliens through violent demonstrations, intimidation, and sabotage of Federal operations,” and, hence, “the regular forces of the United States are not sufficient to ensure the laws of the United States are faithfully executed, including in Chicago.”

The plaintiffs, needless to say, didn’t agree. In their view, the President wasn’t “unable with the regular forces to execute the laws of the United States,” so the statutory prerequisite for calling the National Guard wasn’t satisfied. After a federal district court in Chicago agreed with the plaintiffs, and the Seventh Circuit denied a stay, the case reached the Supreme Court.

6-0-3

To decide whether the President is “unable with the regular forces to execute the laws of the United States,” you need to know what “the regular forces” are.

The government argued that “the regular forces” meant civilian law enforcement. So, according to the government, because ICE and Federal Protective Service—i.e., “the regular forces”—were unable to enforce the law on their own, the President could deploy the National Guard. The district court, however, held that “the regular forces” referred to the military, not civilian law enforcement.

Oddly, both in the court of appeals and the Supreme Court, the plaintiffs didn’t push this point; they instead argued that even if “the regular forces” referred to civilian law enforcement, they should win anyway. Why would the plaintiffs decline to advance an argument on which they had prevailed in the district court? Well, I don’t know, but if I had to guess, they were worried that if they won on this ground, the President would say “OK, fine, then I’ll send the military into Chicago,” which would be even worse from their perspective than sending the National Guard.

But the Supreme Court evidently read the district court’s decision as well as an amicus brief by Professor Lederman advancing the regular-forces-equals-military theory, and so it solicited supplemental briefing on this issue.

The supplemental briefs rolled in, and the issue turned out not to be close.

Overwhelming evidence supports the district court’s and Professor Lederman’s view that “the regular forces” means the full-time military. “Forces” refers to the military; “regular” refers to the full-time military, as opposed to the reserves. Here is a sampling of the evidence:

• Contemporary dictionaries defined “forces” and “regular forces” in exactly this way. For instance, a 1901 treatise on public international law says this: “Regular forces. — While a declaration of war in the usual form calls upon all citizens to proceed to hostilities against the enemy, in practice this duty is limited to those persons who are impressed with military character by the national authority, to such, in short, as constitute the regular army and navy.”

• Contemporary judicial decisions used “regular forces” in exactly this way. See McClaughry v. Deming, 186 U.S. 49, 56 (1902) (noting that the Founders recognized a “substantial difference between the regular forces and the militia”).

• In 1901, shortly before the Dick Act was enacted, President Roosevelt stated in his First Annual Message to Congress: “The organization and armament of the National Guard of the several States . . . should be made identical with those provided for the regular forces.” In context it is obvious that “regular forces” refers to the military. The Dick Act was enacted in direct response to the President’s advocacy: it provided that “the militia, when called into the actual service of the United States, shall be subject to the same Rules and Articles of War as the regular troops of the United States.” (In its initial form, the Dick Act used the phrase “other forces,” but this was amended to “regular forces” in 1908.)

• Even today, Congress uses the word “regular” the same way. For instance, 10 U.S.C. § 101(b)(12) recites: “The term regular, with respect to an enlistment, appointment, grade, or office, means enlistment, appointment, grade of office in a regular component of an armed force.” Congress distinguishes the “regular” forces from the “reserves.” See, e.g., 10 U.S.C. § 7062(c)(1) (distinguishing “Regular Army” from “Army National Guard” and “Army Reserve”).

In response, the government didn’t really dispute that the ordinary meaning of “the regular forces” in 1908 referred to the military. Indeed, the government’s reply brief included this rather striking acknowledgment: “Respondents and their amici show that, in other contexts, ‘the regular forces’ (and cognate terms) can, and often do, refer to the standing military, exhaustively citing various dictionaries, speeches, reports, statutes, legislative history, and judicial decisions.”

Instead, the government offered an assortment of purportedly “contextual” theories for why the Court should depart from the text. For instance, it theorized that in 1903, Congress adhered to the “traditional view that the militia rather than a standing army was the preferred means to address domestic violence.” And, according to the government, the regular-forces-equals-military interpretation was “counterintuitive” because it would require departing from that “traditional view”: “Neither respondents nor their amici identify any clear evidence supporting their counterintuitive position that, having revitalized the National Guard, the Dick Act nevertheless newly subordinated it to the standing military—let alone only for execution of the laws, not for invasions and rebellions.”

Laws are supposed to mean what they say, so this argument does not work. Indeed, when I read the supplemental briefing, I thought the issue was so lopsided that the Supreme Court would rule 9-0 against the Administration on it. It ended up being 6-0, with three Justices declining to opine. Close enough.

The three dissenting Justices argued that the Supreme Court shouldn’t have considered whether “regular forces” means “military” because the plaintiffs waived that argument. Respectfully, I disagree. The district court interpreted “regular forces” to mean the military. Lower-court decisions are entitled to respect, and the Supreme Court doesn’t reverse them unless they’re wrong. That’s why, when a party confesses error in a lower-court ruling, the Supreme Court frequently appoints an amicus to defend the lower court’s decision. I realize that the government was merely seeking a stay here, but the Court doesn’t grant stays unless there’s a likelihood that the applicant will prevail on the merits, and the government won’t prevail on the merits unless the district court’s interpretation is actually wrong.

Also, litigants aren’t entitled to stipulate as to what the law means. It’s common for litigants—particularly repeat players, such as governmental litigants—to agree, for their own convenience, that a law carries a particular meaning. But courts are never bound by those stipulations. The law means what it means, not what the parties might want it to mean. And here, the meaning of “regular forces” is painfully clear.

The dissenting Justices also argued that there was insufficient process to make a reasoned decision—“scant briefing and no oral argument.” I’m all in favor of process, but that’s not a basis to grant a stay. The Supreme Court is fond of saying that a “stay is not a matter of right,” so I have no problem with denying a stay when the briefing is inadequate. Indeed, Justice Kavanaugh’s approach in Trump v. Illinois—ruling that “regular forces” means the military, denying the stay application because the President’s order rested on the assumption that “regular forces” included civilians, and declining to address additional issues because they are complex and inadequately briefed—strikes me as a completely reasonable way of resolving this application. But the Court shouldn’t grant a stay because there was purportedly insufficient process to confirm that a lower-court decision was correct.

Babysitter Law

We’ve determined that “the regular forces” means the military. What are the implications of that?

The statute says that the President may call out the National Guard when he “is unable with the regular forces to execute the laws of the United States.”

The plaintiffs argued that “unable” means actually, physically, unable. So if the military tries to “execute the laws of the United States,” but it is overwhelmed by Antifa rioters and requires backup, the President can call out the National Guard. Here, according to the plaintiffs, the military isn’t being overwhelmed by rioters, so the President can’t call out the National Guard.

Indeed, the President hasn’t even tried to use the military to suppress the rioters. There’s a good reason for that: in 1878, Congress passed a famous law known as the Posse Comitatus Act, which (to first approximation) bans the military from being used to “execute the laws,” “except in cases and under circumstances expressly authorized by the Constitution or Act of Congress.”

There are a bunch of situations in which the President is expressly authorized to use the military for law enforcement purposes. The most famous is the scary Insurrection Act, which I sincerely hope that I will never have to Substack about. There are also approximately 25 other statutes authorizing the use of the military for law enforcement purposes, for crimes ranging from illegally settling Indian reservations to trafficking nuclear weapons. But the President hasn’t suggested—yet—that any statutory exception to the Posse Comitatus Act applies. So the President can’t call out the military to the streets of Chicago.

The government claims, as a result, that it’s “unable” to quell the riots with the military—after all, Congress banned it from doing so—which means it’s allowed to use the National Guard. The plaintiffs say “unable” means “actually unable.”

The Supreme Court held that the plaintiffs’ interpretation was correct, and I agree. In Biden v. Nebraska, Justice Barrett famously analogized Congress to parents giving instructions to a babysitter, and that’s as good a way to convey the point as any. Suppose parents tell a babysitter: “don’t give the child any Tylenol unless his fever exceeds 101°F. If you’re unable to break the fever with Tylenol, call us.” Suppose the babysitter calls the parents and says: “your child’s fever is 100°F. I was unable to break the fever with Tylenol, because you told me not to give him Tylenol unless his fever was 101°F, so that’s why I’m calling.” The babysitter isn’t interpreting the parents’ instructions correctly. The parents are saying that if the condition for giving Tylenol is met, and the Tylenol doesn’t work, then the babysitter should call. “Unable” means actually unable, not “I’m not doing it because you told me not to.”

Here, likewise, the Posse Comitatus Act says: “don’t use the military to enforce the law unless an exception applies.” The Dick Act says: “if you’re unable to enforce the law using the military, you can use the National Guard.” Congress is saying that if the President tries and fails to enforce the law using the military, he can use the National Guard, not that he’s “unable” to enforce the law using the military when Congress tells him not to.

We are not quite done yet. The government has one more arrow in its quiver. It claims that the Posse Comitatus Act doesn’t apply to its intended operations in Chicago at all.

Recall that the Posse Comitatus Act restricts the military from being used to “execute the laws,” “except in cases and under circumstances expressly authorized by the Constitution or Act of Congress.” The government claims that if the military was, hypothetically, deployed to the streets of Chicago, it wouldn’t be “executing the laws” (the thing that the Posse Comitatus Act forbids). Instead, the military would be doing a different thing: “protecting federal property and federal functions.” And the President has the inherent constitutional authority to ask the military to do that.

There are a lot of debatable premises in that paragraph, but the Supreme Court doesn’t grapple with them. Instead, it breaks out some sweet, sweet logic. The Posse Comitatus Act says that the President can’t use the military to “execute the laws.” 10 U.S.C. § 12406 says that the President can’t call out the National Guard unless he’s “unable with the regular forces to execute the laws.” Statutory-interpretation hornbooks teach that “execute the laws” must mean the same thing in both places. So if the thing the President claims he’s allowed to do—use the military to protect “federal property and federal functions”—isn’t “executing the law” for purposes of the Posse Comitatus Act, it also isn’t “executing the law” for purposes of 10 U.S.C. § 12406. Which means that the President doesn’t have a theory for how he’s allowed to even try to “execute the law” for purposes of 10 U.S.C. § 12406, which means that he isn’t unable to do so. Got that? You see, this is why they make lawyers take the LSAT.

In praise of puzzle-solving

You might find this reasoning dissatisfying.

At every step, this analysis treats the interpretation of 10 U.S.C. § 12406 as a kind of linguistic puzzle to be solved. The first step addresses how the words “regular forces” would have been understood in 1908. The second step focuses on how people communicate with their babysitters. The third step starts with an old-school statutory interpretation canon and adds some nifty deductive reasoning.

This analysis is so … limited. There is nothing about the policy merits of using the military for domestic law enforcement, the policy merits of using the regular military versus using the National Guard, or the reason we even have a National Guard. There is a little bit of history—you have to read some historical documents to understand what the words “regular forces” meant in 1908—but the historical analysis focuses on a single obscure tree and ignores the forest. There is no discussion of the broader purpose of the Dick Act, the history of state militias at the Founding and thereafter, or the interplay between state militias and the standing military. Nor is there anything about the situation on the ground in Chicago. You would have no idea, from this discussion, that the case has something to do with the use of military force to assist ICE. Can you really interpret this statute correctly without achieving a deeper understanding of these issues?

Yes.

In my view, treating 10 U.S.C. § 12406 as a logic puzzle is the correct way to resolve this case. Given the Court’s terse reasoning, one cannot be certain that it actually resolved the stay application on that ground. But that’s how I think a court should approach the issue. The court should blind itself to the nuanced and delicate and sophisticated historical and policy questions at stake and should instead stare at the English words on the page, accompanied by some limited, targeted historical research into what those words meant when they were written, until the right answer pops out.

I support this approach because it is consistent with textualism, the mode of interpretation that I favor. Textualism prescribes both a goal—interpret statutes according to their ordinary meaning at the time of enactment—and a means of achieving that goal—consider only those materials that illuminate the meaning of the text and ignore the rest. Textualism differs from other interpretive philosophies, which posit that courts should interpret statutes based on considerations such as perceived legislative purpose and predicted practical outcomes, and which by necessity require consideration of a more eclectic body of evidence. I will spare you an apologetic for textualism and instead highlight a few ways in which Trump v. Illinois shows textualism’s benefits.

First, cases—particularly highly politicized cases like this one—should be decided based on neutral principles. And the textualist principles that decide this case—like “interpret words according to their ordinary meaning at the time of their enactment” and “if the same phrase appears in two related provisions, the phrase should be construed to mean the same thing in both provisions”—are neutral.

I suppose an alternative methodology, like “interpret the statute by balancing the text with history, policy, and one’s intuitions about what makes sense” reflects a kind of neutral principle too. But that methodology is so indeterminate that it might as well not be a principle at all. Neutral principles are useful when they constrain the judge to vote in a manner that’s different from how the judge might personally prefer the case to come out.

I would prefer to avoid a dystopia in which every politically controversial case, regardless of the legal arguments, turns on personal political preferences of the judges deciding the case. And if the methodology for interpreting statutes is “balance text, history, purpose, and intuition,” then that dystopia awaits. On the other hand, if judges apply the methodology of “just try to figure out what the words mean and block out all other thoughts,” then perhaps there is some hope of avoiding that outcome.

This case illustrates the point. Suppose a judge applies the “balance text, history, purpose, and intuition” methodology. The judge might start by figuring out what the phrase “regular forces” meant in 1908. But then the judge will consider additional factors, including things like, “does it intuitively make sense that the President should have to call out the military before deploying the National Guard?” Once “intuition” enters the picture, political conservatives will think it’s intuitive that the National Guard should be assisting ICE in Chicago and political progressives will think it’s intuitive that it shouldn’t. And then at the final step, the judge has to balance all of these factors according to some ill-defined, subjective standard—an incubator for motivated reasoning. The inevitable outcome is that the judge’s legal conclusion will align with the judge’s political preference. By contrast, if the judge merely tries to figure out what “regular forces” meant in 1908 and ignores everything else, the judicial mind will have no opportunity to wander to the judge’s personal reactions to the images from Chicago he sees on TV.

Even better, this more constrained methodology allows us to achieve the dream of producing objectively correct answers to difficult legal questions. Maybe not objectively correct in the sense of 1+1=2. But from my perspective, if you focus on what the words meant in 1908 and ignore everything else, the evidence in favor of the “regular forces = military” interpretation is so strong that it’s fair to call it objectively correct. The other two issues—whether “unable” means “physically unable” and whether “executing the laws” means the same thing in the Posse Comitatus Act and the Militia Act—are closer. But when I stare at the words on the page and repeat them in my head, I’m convinced that the majority’s interpretation is just true—more convinced than I could be, anyway, if I had to start philosophizing about the reasons we have the National Guard or what would be best for Chicago.

Another core feature of textualism is that courts consider only a limited body of materials. If the court’s sole goal is to figure out what the words on the page mean, then any information that does not advance that goal is an irrelevant distraction. That is why, for example, textualist judges eschew reliance on legislative history.

Limiting the body of relevant evidence has the virtue of ensuring that judges focus on what they are good at. Judges are experts at applying Latin canons of construction and examining obscure legal texts from 1901. But they are not experts at everything.

Again, look at this case. The plaintiffs argue that there is a longstanding national tradition of “governors calling forth their own militia and, if needed, requesting assistance from the federal government and other States,” with an “established presidential practice of using the military, rather than the militia, when federal interference was warranted.” The government argues that “the Framers and early Congresses treated the militia as the first line of defense to quell domestic obstructions of civil law enforcement, while viewing a standing army as the feared last resort,” and Congress carried forward that tradition in the Dick Act. Both parties attempt to draw inferences from President Washington’s actions during the Whiskey Rebellion of 1794.

Is it really possible that a generalist court can master the relevant historical evidence to a sufficient extent to resolve this dispute accurately, particularly on the tight timeline of a stay application? Like 94% of middle-aged men, I personally enjoy reading about military history, yet I knew nothing about the historical circumstances underlying the Dick Act’s enactment until I read the briefs in this case. And judges misguided enough to have other hobbies might know even less. It is not clear that whatever historical knowledge judges acquire from this process will bear any particular relationship to reality. If you ignore the Whiskey Rebellion and focus on dreaming up babysitter hypotheticals, this problem does not arise.

I don’t want to overstate the point. One cannot resolve all interpretive disputes merely by staring at words and dictionaries. In some cases, history or policy or practicality will inevitably, perhaps desirably, drive the outcome. Every case is different.

But sometimes, staring at words and dictionaries actually works.

Happy New Year!

Because this is America, litigation ensued. The plaintiffs argued that Texas redistricted because of race and that the new map was hence unconstitutional. Texas argued that it redistricted because of partisanship and that the new map was hence perfectly fine.

A three-judge court heard the case. Two of the judges agreed with the plaintiffs and granted a preliminary injunction, accompanied by a 160-page opinion by Judge Jeffrey Brown. The third judge, Judge Jerry Smith, opined in his 104-page dissent that “the main winners from Judge Brown’s opinion are George Soros and Gavin Newsom.” Texas has sought a stay from the Supreme Court, with a decision expected imminently.

Who’s right? Read on to find out.

Factual, legal, and typographical errors

To understand the plaintiffs’ theory that Texas redistricted because of race, you need to know the timeline.

• June 9: New York Times publishes an article saying that President Trump’s team was pushing Texas to redistrict for partisan reasons, but Texas Republicans weren’t interested.

• June 23: Governor Abbott announces that he is calling a special legislative session and lists several agenda items, but redistricting isn’t among them.

• July 7: Harmeet Dhillon, the head of the Civil Rights Division at the Department of Justice, sends a letter to Governor Abbott expressing the view that four Texas congressional districts “currently constitute unconstitutional ‘coalition districts’” and “urg[ing] the State of Texas to rectify these race-based considerations from these specific districts.”

• July 9: Governor Abbott issues a formal proclamation calling a special session which includes the following item: “Legislation that provides a revised congressional redistricting plan in light of constitutional concerns raised by the U.S. Department of Justice.”

Thus, Governor Abbott’s public rationale for adding redistricting to the agenda was DOJ’s letter. So let’s take a look at the letter.

There’s a backstory here that I’ll simplify mercilessly. It’s generally unconstitutional to draw district lines based on race. But, sometimes, at least for now,1 the Voting Rights Act requires that if a reasonably-configured majority-minority district can be drawn, then it must be drawn to avoid excessively diluting the votes of minority voters. (It’s more complicated than that, but close enough.2) And drawing majority-minority districts requires being conscious of race in drawing district lines. So, summing things up, drawing district lines based on race is unconstitutional unless it’s mandatory to satisfy the Voting Rights Act.

In 1988, in a case called Campos, the Fifth Circuit (which covers Texas) held that the Voting Rights Act sometimes requires “coalition” districts to be drawn, i.e., districts that are majority-minority when all minority groups are combined together.3 So, for example, if it’s possible to draw a reasonably-configured district that is 30% Hispanic and 25% Black, Campos held that the Voting Rights Act sometimes obligated the state to draw such a district. In 2024, in a case called Petteway, the Fifth Circuit overruled Campos. It held that the Voting Rights Act doesn’t kick in unless it’s possible to draw a reasonably-configured single-minority-majority district—i.e., 50%+ Hispanic or Black.

To simplify even more mercilessly: before Petteway, Texas was sometimes legally obligated under the Voting Rights Act to intentionally draw coalition districts. After Petteway, that is no longer the case.

Back to the DOJ letter. Judge Brown’s opinion states that “[i]t’s challenging to unpack the DOJ Letter because it contains so many factual, legal, and typographical errors.” This is a bit uncharitable in my view, but the letter was not the DOJ’s finest hour.

The gist of the DOJ letter is that, when Texas redistricted in 2020, it created coalition districts because it thought it had to under Campos. But Petteway overruled Campos. So, it turns out, Texas actually didn’t have to draw those districts. So that leaves Texas with a bunch of coalition districts that it didn’t have to draw. But wait. Texas considered race in drawing those districts. And when the Voting Rights Act doesn’t require consideration of race, the Constitution prohibits consideration of race. Ergo, the coalition districts are unconstitutional and Texas has to get rid of them.

Here’s the actual discussion in the letter:

It is well established that so-called “coalition districts” run afoul the Voting Rights Act and the Fourteenth Amendment. … In [Petteway] , the en banc Fifth Circuit Court of Appeals made it abundantly clear that “coalition districts” are not protected by the Voting Rights Act. This was a reversal of its previous decision in [Campos] …

It is the position of this Department that several Texas Congressional Districts constitute unconstitutional racial gerrymanders, under the logic and reasoning of Petteway. Specifically, the record indicates that TX-09 and TX-18 sort Houston voters along strict racial lines to create two coalition seats, while creating TX 29, a majority Hispanic district. Additionally, TX-33 is another racially-based coalition district that resulted from a federal court order years ago, yet the Texas Legislature drew TX 33 on the same lines in the 2021 redistricting. Therefore, TX-33 remains as a coalition district.

Although the State’s interest when configuring these districts was to comply with Fifth Circuit precedent prior to the 2024 Petteway decision, that interest no longer exists. Post-Petteway, the Congressional Districts at issue are nothing more than vestiges of an unconstitutional racially based gerrymandering past, which must be abandoned, and must now be corrected by Texas.

These coalition districts elect Democrats, so getting rid of them would help Republicans. Perhaps that’s the subtext here. But it isn’t the text. The DOJ letter says nothing about partisanship and talks only about race.

Dear Texas, Please Discriminate

There’s a problem here.

The premise of the DOJ Letter is that Texas considered race in drawing the coalition districts—that is, it intentionally drew coalition districts to comply with the Voting Rights Act.

As it turns out, however, Texas insists that’s not true. Texas says that it didn’t consider race in drawing those lines, and those districts just happened to end up being coalition districts.

Hold the phone.

The DOJ Letter exhorts Texas to target coalition districts for destruction. It urges Texas to redraw district lines because of the racial composition of those districts. That’s unconstitutional! Indeed, the whole premise of the letter is that it’s unconstitutional to draw district lines based on race! Maybe if the districts were originally drawn for race-based reasons, it would make sense to redraw them. But if the districts weren’t originally drawn for race-based reasons, then the DOJ Letter would effectively be telling Texas: “Dear Texas, please redistrict in order to racially discriminate against Black and Hispanic voters.”

Or, put more concisely, we are currently living in a timeline in which the DOJ announces that race-based redistricting is unconstitutional, and then demands that Texas do that very unconstitutional thing, in the span of a single two-page document.

The DOJ filed an amicus brief in the Supreme Court case trying to explain away the statements in the letter. Importantly, the amicus brief doesn’t suggest that the coalition districts were originally drawn for racial reasons. Instead, the amicus brief says that DOJ wasn’t encouraging Texas to redistrict for racial reasons: “And the letter certainly did not demand that Texas redraw the districts in a manner that was race-predominant rather than race-neutral: racial gerrymandering was itself the problem that the DOJ letter asserted regarding the districts, and a State obviously can remedy racial gerrymandering without engaging in more racial gerrymandering.” I don’t understand this. The DOJ letter specifically urged Texas to target districts because of their racial composition. If Texas followed DOJ’s lead, how can that be anything other than race-based redistricting?

And remember that Governor Abbott’s proclamation for the special session urged “legislation that provides a revised congressional redistricting plan in light of constitutional concerns raised by the U.S. Department of Justice.” So Governor Abbott was urging legislators to redistrict based on DOJ’s rationale—and if those legislators were following Governor Abbott’s lead, they would be redistricting for an unconstitutional reason.

It wasn’t just the proclamation. Governor Abbott’s public statements were laser focused on race. Check out this transcript of Governor Abbott’s interview with CNN’s Jake Tapper, which occurred after the legislature proposed the new map (emphasis added):

ABBOTT: … [O]ne thing that spurred all of this is a federal court decision that came out last year. By the way, a case that was filed by Democrats, the federal court decision that came out last year said that Texas is no longer required to have coalition districts.

And as a result, we had drawn maps with coalition districts in it. Now we wanted to remove those coalition districts and draw them in ways that in fact, turned out to provide more seats for Hispanics. For example, [four of] the districts are predominantly Hispanic. It just coincides it’s going to be Hispanic Republicans elected to those seats.

One thing that’s happened in the state of Texas is the Hispanic community, a lot of it, have decided they’re no longer with the Democrats who believe in open border policies, who believe in going against our law enforcement, who believe that men should play in women’s sports. And they instead align with Republicans. What we want to do is to draw districts that give those Hispanics and African- Americans in the state of Texas the ability to elect their candidate of choice.

TAPPER: That’s not really -- I mean, you’re doing this to give Trump and Republicans in the House of Representatives five additional seats, right? I mean, that’s the motivation, is to stave off any midterm election losses.

ABBOTT: Again, to be clear, Jake, that the reason why we’re doing this is because of that court decision. Texas is now authorized under law that changed that was different than in 2021 when we last did redistricting.

He’s explicitly denying that the impetus for redistricting is partisan! He’s saying that the partisan outcome is a coincidence!

The irony here is that if Governor Abbott had simply said, “let’s redistrict because we hate Democrats and want to keep them out of power,” this would be an amulet against any lawsuit: the Supreme Court held, in the controversial case of Rucho v. Common Cause, that partisan gerrymanders can’t be challenged in federal court. But instead, Governor Abbott, following DOJ’s lead, insisted that redistricting should occur because of race. This is yet another illustration of the chief problem that has plagued both the Trump and the Abbott Administrations: insufficient partisanship against Democrats.

Inverse critical race theory

And now back to the litigation. The plaintiffs’ basic theory, which the majority adopts, is that the Texas legislature redistricted for the reason advocated by the governor, which would make the new map unconstitutional.

Judge Smith’s dissent has gotten a lot of attention for accusing the plaintiffs’ lawyers of accepting money from George Soros. There are some intriguing points there, but as a nerd who cares about the legal arguments, I’ll skip that part and turn to the substance of the dissent, which reveals a surprising amount of common ground with the majority. Judge Smith doesn’t dispute that the coalition districts were drawn for innocent reasons. He doesn’t dispute that if the legislators redistricted based on the rationale in the DOJ Letter, then the new map would be an unconstitutional racial gerrymander.

Instead, Judge Smith’s thesis is that Texas’s legislators didn’t care what the DOJ or Governor Abbott said, and that their true rationale for redistricting was to ensure that more Republicans get elected. Call it inverse critical race theory: even though figures as powerful as the head of the DOJ’s Civil Rights Division and the Governor of Texas were explicitly and publicly urging legislators to redistrict for racial reasons, legislators were privately acting based on anti-racist motives.

And his thesis might be right! Texas Republicans do indeed hate Texas Democrats, so it is not altogether implausible that Texas Republicans might be motivated by the goal of defenestrating them.

If we’re being honest, it’s virtually certain that Republican legislators were motivated by both race and partisanship: they thought both that (A) it would be great if Republicans kept winning and (B) if DOJ is telling us that the Constitution requires us to blast coalition districts into oblivion, then, well, who doesn’t love the Constitution? Or maybe some Republican legislators thought (A), some thought (B), some thought both (A) and (B), and some (most?) just figured they’d do what everyone else was doing. In this situation the Supreme Court has declared that courts must assess whether race was the “predominant factor motivating the legislature.” Good luck, courts!

Dog’s breakfast

Neither Harmeet Dhillon nor Greg Abbott is a member of the Texas legislature. The fact that they endorsed race-based redistricting does not necessarily imply that Texas legislators redistricted for that reason. And so most of the 264 pages of opinions is devoted to sifting through the dog’s breakfast of an evidentiary record to figure out what the legislators were really thinking.

There are four categories of evidence at issue.

The actual enacted map. DOJ’s letter identified three purportedly unconstitutional coalition districts: TX-09, TX-18, and TX-33.

In the old map, TX-9 was a coalition district that was 45.0% Black and 25.6% Hispanic. In the new map, it’s barely majority-Hispanic: 50.3% to be exact.

In the old map, TX-18 was a coalition district that was 38.8% Black and 30.4% Hispanic. In the new map, it’s barely majority-Black: 50.5% to be exact.

There’s also another district that DOJ didn’t mention, TX-30, which was a coalition district under the old map: 46.0% Black and 24.5% Hispanic. In the new map, it’s barely majority-Black: 50.2% to be exact.

Verrrry suspicious. On the other side of the ledger, however, the third district that DOJ mentions, TX-33, remained a coalition district after the redistricting, going from 43.6% Hispanic/25.2% Black to 38.2% Hispanic/19.6% Black.

Also, DOJ’s letter complains about TX-29, which was a majority-Hispanic district under the old map. In the new map, the share of Hispanics in that district drops from 63.5% to 43.3%. Make of that what you will.

Not that you asked, but I construe this evidence as strong, but not conclusive, evidence supporting the plaintiffs. The dissent seems to think that the legislature’s failure to extinguish TX-33 proves that it didn’t care about what DOJ and Governor Abbott said. I don’t agree. If a governor announces a special session saying “Please do these three things for racial reasons,” and the legislature dutifully does two of the three things but not the third thing, wouldn’t you still think the legislature did the two things for racial reasons?4

Legislator statements. Most legislators didn’t say anything. The majority tries to infer what they were thinking based on the statements of powerful Republican legislators. This makes a certain kind of sense—back-benchers probably were following the lead of front-benchers—but you can debate that.

The plaintiffs offer the following evidence:

• The Speaker of the Texas House, Dustin Burrows, issued a press release that starts by referring to the “constitutional concerns raised by the Department of Justice,” but then declares that the new map “has ushered in a new chapter of Republican unity.”

• The Chair of the House Republican Caucus, Tom Oliverson, conducted an interview that seems to say that redistricting occurred because of Petteway, while also noting that partisan redistricting is permitted.

• Representative Todd Hunter, who introduced the redistricting bill, made various statements on the House floor regarding the racial composition of the new districts while also making partisan statements.

• Representative Steve Toth also gave a press interview in which he mentioned Petteway but also partisanship.

And on the defendants’ side:

• Representative Cody Vasut, the Chairman of the House Select Committee on Congressional Redistricting, made multiple contemporaneous statements, both on and off the floor, that the legislature was redistricting for partisan rather than racial reasons. He also testified at the preliminary injunction hearing to the same effect.

• Two state Senators, Phil King and Adam Hinojosa, testified at trial that the redistricting occurred for partisan rather than racial motives. It does not seem that they had significant roles in the redistricting process beyond casting their votes.

My take: The plaintiffs don’t have any smoking guns, but the statements from Representatives Burrows and Oliverson are good-ish for the plaintiffs. Representative Vasut’s statements are excellent for the defendants.

Adam Kincaid. The person who actually drew the map was not a state legislator. It was a Republican operative named Adam Kincaid. Mr. Kincaid’s testimony makes him out to be a map-drawing supergenius, able to draw maps that marry aesthetic splendor with surgical precision in ruthlessly annihilating Democratic seats.

He testified at length regarding why the map was drawn the way it was, and he had partisan explanations for everything. He didn’t hide the extreme partisanship underlying the new map—for example, he testified that he prioritized protecting Republican incumbents and core retention in Republican districts, while giving zero consideration to Democratic incumbents and core retention in Democratic districts.

Judge Smith takes the view that Mr. Kincaid “courageously spoke the truth” and that his testimony was “irrefutable.” The majority disbelieves Mr. Kincaid, finding it “extremely unlikely that Mr. Kincaid could have created so many districts that were just barely 50%+ … by pure chance.” The majority also notes that Mr. Kincaid discussed the DOJ letter with key White House and DOJ officials before it was released.

To me it seems plausible that Mr. Kincaid, by all accounts a hardcore Republican partisan, drew the maps to help Republicans and didn’t care about minority percentages one way or the other. On the other hand, it seems unlikely that Mr. Kincaid wouldn’t have kept his eye on the goal that the Justice Department and Governor Abbott publicly set out to achieve. If we call that a wash, I’m more persuaded by another conclusion reached by the majority: “Mr. Kincaid’s professed lack of racial motive isn’t attributable to the legislature.” If the legislature thought it was faithfully implementing Governor Abbott’s directive, why does it matter what lay in the heart of Mr. Kincaid? If a tree falls in a forest and no one hears it, does it make a sound?

The lack of an alternative map. The plaintiffs’ goal was to demonstrate that the legislature acted for racial rather than partisan reasons. To make that showing, the plaintiffs could have proposed an alternative map that would have allowed the legislature to achieve its partisan criteria while not hitting racial targets. This would have supported the inference that the legislature’s choice of this particular map was motivated by race rather than partisanship.

Here, however, the plaintiffs did not introduce an alternative map into evidence. Of note, in Alexander v. South Carolina State Conference of the NAACP, the most recent Supreme Court case about racial gerrymandering, the Supreme Court stated, in pretty strong language, that an adverse inference is warranted when the plaintiff doesn’t propose an alternative map:

The evidentiary force of an alternative map, coupled with its easy availability, means that trial courts should draw an adverse inference from a plaintiff ’s failure to submit one. The adverse inference may be dispositive in many, if not most, cases where the plaintiff lacks direct evidence or some extraordinarily powerful circumstantial evidence such as the “strangely irregular twenty-eight-sided” district lines in Gomillion v. Lightfoot, 364 U. S. 339, 341 (1960), which betrayed the State’s aim of segregating voters on the basis of race with “mathematical” precision, ibid.

Texas scores points based on this language. On the other hand, this case differs from Alexander. Demanding an alternative map makes sense when the plaintiff argues: “racial minorities are packed into a small number of districts, that must mean that there was racial discrimination,” and the state responds, “the reason we packed those voters isn’t because we cared about race—it’s because we couldn’t have achieved our partisan goals any other way.” If the plaintiff can’t then show that the state could have achieved its partisan goals another way, this backs up the state’s defense. And that’s essentially what happened in Alexander. Here, though, the plaintiffs’ case hinged primarily on direct evidence of racial intent—specifically, evidence that the DOJ’s letter precipitated redistricting. The lack of an alternative map doesn’t make this evidence go away.

Dystopia

What’s my take on all this?

This is the type of case in which it’s particularly easy to fall victim to motivated reasoning—that is, convincing yourself that the plaintiffs’ or defendants’ legal arguments are more persuasive because you’re rooting for the Democrats or Republicans to win the House of Representatives. Courts have occasionally been known to engage in motivated reasoning, and indeed I frequently read commentary suggesting that we are doomed to a judicial dystopia in which all cases split on party lines until the end of time. Let’s hope not. But regardless of how this particular case is resolved, you, reader, should try to formulate your views without regard to partisan impulses.

There is something to be said for the view that the entire enterprise of trying to infer the mental states of legislators is misbegotten. Justice Thomas articulates that view in a characteristically thoughtful concurrence in the Alexander case.

If you are a Democrat, it is easy to find this position annoying. Racial-gerrymandering claims originated in the 1990s, when Republicans argued that Democratic legislators were shifting Black voters for racial reasons while Democratic legislators argued that they were merely being partisan. At the time, the Supreme Court’s conservative wing was solicitous of such claims while the progressives endorsed deference to state legislatures. Take a look at the dueling opinions in Easley v. Cromartie, 532 U.S. 234 (2001), which feel like they are from an alternative reality. Now that Democrats have started bringing these claims, it is irritating to Democrats that judicial conservatism has suddenly heard the gospel of deference to legislative good faith. Conversely, it is equally irritating to Republicans that they have to defend against these claims after being lectured by Justice Breyer in Easley that “courts must ‘exercise extraordinary caution in adjudicating claims that a State has drawn district lines on the basis of race,’” especially where “the State has articulated a legitimate political explanation for its districting decision, and the voting population is one in which race and political affiliation are highly correlated.”

So it’s tempting for Democrats / Republicans to conclude that cosmic justice requires supporting the Democrats / Republicans, and then to start believing the Democrats’ / Republicans’ legal arguments. Don’t succumb! Motivated reasoning is bad. It doesn’t matter what happened in the 1990s, we should just try to figure out the right answer.

Cryostorage

The ideal judge for this case would be one who was just awakened from cryostorage and knew nothing about whether the Democrats or the Republicans were plaintiffs. Locating such a judge unfortunately presents logistical difficulties. So, the second-best solution is for an actual judge to fanatically apply neutral principles and extirpate all partisan thoughts. The ideal judicial decision here is a boring, rote application of hornbook principles of civil procedure.

I am as tormented by the temptation of motivated reasoning as anyone else, but the blandest and most mid legal answer I can think of is: (1) Racial-gerrymandering claims should prevail rarely but not never, and (2) The case should turn on the deferential standard of review.

On the first point, to me, it is treacherous to try to infer the mental states of legislators unless there’s a bizarre fact pattern in which high-level officials inexplicably declare that they’re motivated by race. Like, say, the governor calling a special session and requesting that the state legislature redistrict for racial reasons while disclaiming any partisan intent on TV.

The governor, it is true, is not the legislature. But as a matter of common sense, if the governor calls a special session for the stated purpose of getting rid of coalition districts, it is reasonable to infer that legislators who heeded the call, and indeed got rid of multiple coalition districts, were also motivated by the goal of getting rid of coalition districts.

There was other evidence supporting the plaintiffs, and also evidence going the other way. You can debate every individual piece of evidence, and hoo boy, do the majority and the dissent ever. But—turning to the second point—the majority made a factual finding here that can’t be reversed unless there’s clear error.

If this record isn’t enough to withstand clear-error review, then racial-gerrymandering claims are dead. Which, per Justice Thomas, maybe they should be! But the case law on the books says that racial-gerrymandering claims are alive, and I wouldn’t use this particular case as a vehicle to move the law.

Texas urges that the district court gave insufficient weight to the lack of an alternative map and to the presumption of legislative good faith. Texas claims these were errors of law rather than fact. But the district court acknowledged the lack of an alternative map, acknowledged the presumption of legislative good faith, and found them outweighed by the rest of the evidence. I am confident that our awakened-from-cryostorage judge would conclude that Texas’s disagreements with how the trial court weighed the evidence present a question of fact, not law.

A different three-judge court undoubtedly could have reached a different conclusion and that also would not have been clear error. It seems strange that Texas’s congressional map should turn on the random selection of three judges. But that’s the system we have. Maybe there’s a philosophical case to be made that our legal system shouldn’t function this way. But in a case like this, one yearns for neutral principles, and one neutral principle is that the decisions of randomly-selected trial courts are reviewed for clear error. This is a boring, lawyerly analysis, and there is no better time to apply boring, lawyerly analyses than highly-politicized racial-gerrymandering claims.

The Empire State Building

The last topic I’ll tackle is the so-called Purcell principle: courts generally shouldn’t enjoin election laws in the period close to an election. Texas argues that we’re too close to the November 2026 election, so the Purcell principle should kick in, Texans should elect 30 Republicans in 2026, and the litigation should focus on 2028. The plaintiffs disagree.

Let me make a few observations here.

First, the plaintiffs sued as quickly as they could—before the law was even signed, in fact. Also, in an effort to get a ruling as quickly as possible, they didn’t take discovery, which prevented the plaintiffs from learning the state’s redistricting criteria before the hearing and hence (according to the plaintiffs) is the reason they couldn’t introduce an alternative map.

Texas says that the plaintiffs are responsible for delay because Democratic legislators temporarily left the state in an unsuccessful effort to prevent the new map from passing. The implicit premise of this argument is that the plaintiff organizations and Democratic state legislators form a kind of left-wing hive mind, such that if Democratic legislators don’t cooperate with Republican legislators, it is the League of United Latin American Citizens’ fault. I do not agree.

Second, the district court also hustled. It issued its 160-page opinion weeks after the hearing. And to further move things along, the majority released its opinion ahead of the dissent.5

Third, the district court, recognizing that there wasn’t sufficient time for the Texas legislature to enact a new map, chose the least intrusive possible remedy: a preliminary injunction that directed Texas to use the prior map for the 2026 election. This map was enacted by the Republican legislature and was already a Republican gerrymander (even under the old map, 66% of the seats went to Republicans, compared to 56% of Texans who voted for Donald Trump in 2024), so it’s not as though the court was foisting some Soros-drawn map on Texans. The dissent spends several pages arguing that the injunction exceeded the court’s equitable discretion, but I don’t understand the dissent’s argument. Having enjoined the old map, the court couldn’t just leave Texas map-less. In the absence of a legislative solution, an injunction returning Texas to the status quo ante was the reasonable thing to do.

Fourth, it’s not like the injunction disrupted any reliance interests. The nanosecond the new map was passed, everyone knew a lawsuit was coming. There’s confusion right now about whether Texas will use the old map or the new map, but it’s not like an order denying a stay would be any more disruptive than an order granting a stay.

Texas emphasizes that the candidate filing period has already begun. Fine. The election is over 11 months away. There is enough time to adjust to using the maps that were already being used. The Empire State Building was completed in 13 months. Disneyland opened for business one year after construction began. We can do it.

At the end of the day, Texas’s argument boils down to the proposition that there’s simply nothing a court can do in this situation. If Texas enacts a new map in mid-2025, it doesn’t matter if the new map reflects intentional racial discrimination. Under Purcell, it’s just too bad, so sad for Hispanic and Black voters, who have to let themselves be discriminated against for a voting cycle and hope for better luck next time.

I cannot agree that our judicial system is as dysfunctional as that. If a state legislature enacts a racially discriminatory map 15 months before an election and plaintiffs sue instantly, our judicial system should be capable of vindicating the plaintiffs’ claims in time for the election. Our awakened-from-cryostorage judge would combine jurisprudential blandness with a can-do spirit.

I was honored to work on an amicus brief in this case on behalf of a group of distinguished economists. While I agree with every word in that amicus brief, this post reflects my own personal views.

During the Biden Administration, the Supreme Court issued four decisions holding that the Executive Branch may not issue rules addressing major questions without clear congressional authorization. Those were difficult cases reflecting deep philosophical disagreements among the Justices. You can reasonably come out either way on the Major Questions Doctrine. But with those decisions on the books, the tariffs have to go down.

Concocting clever justifications

Between 2021 and 2023, the Supreme Court struck down four major Biden Administration actions. All four cases were decided 6-3, with the Republican appointees in the majority and the Democratic appointees in dissent. The four cases addressed different issues, but at heart, they were all the same case.

Round One: Alabama Association of Realtors v. HHS. The statute at issue said this:

The Surgeon General, with the approval of the [Secretary of Health and Human Services], is authorized to make and enforce such regulations as in his judgment are necessary to prevent the introduction, transmission, or spread of communicable diseases from foreign countries into the States or possessions, or from one State or possession into any other State or possession. For purposes of carrying out and enforcing such regulations, the Surgeon General may provide for such inspection, fumigation, disinfection, sanitation, pest extermination, destruction of animals or articles found to be so infected or contaminated as to be sources of dangerous infection to human beings, and other measures, as in his judgment may be necessary.

During COVID, the CDC imposed a moratorium on evictions. The CDC’s theory was that people who were evicted might have COVID, move to new locations, and infect more people.

In the Supreme Court, the government argued that the literal words of the statute authorized this action: the Secretary’s “judgment” was that banning evictions was “necessary to prevent” the spread of COVID. The challengers argued: come on. Banning evictions is completely different from measures that Congress had in mind, like pest extermination. Sure, maybe clever government lawyers could come up with a theory under which existing law justified the moratorium. But for such a momentous policy change, clearer language was needed.

Round Two: NFIB v. OSHA. OSHA has the authority to ensure “safe and healthful working conditions” by enacting standards that are “reasonably necessary or appropriate to provide safe or healthful employment.” During COVID, OSHA imposed a vaccine mandate, on the theory that it would reduce people’s risk of catching COVID from their co-workers.

In the Supreme Court, the government argued that the literal words of the statute authorized this action: what could be more “healthful” than not getting COVID? The challengers argued: come on. Mandating vaccines is completely different from measures that Congress had in mind, like hazmat suits and guards on meat slicers. Sure, maybe clever government lawyers could come up with a theory under which existing law justified the vaccine mandate. But for such a momentous policy change, clearer language was needed.

Round Three: West Virginia v. EPA. EPA has the authority to set standards for power plants reflecting the “best system of emission reduction.” EPA announced a rule requiring coal plants to reduce electricity production by shifting generation towards cleaner sources like wind and solar.

In the Supreme Court, the government argued that the literal words of the statute authorized this action: replacing coal plants with cleaner generation sources was a “system of emission reduction.” The challengers argued: come on. Moving from coal to cleaner energy is completely different from measures that Congress had in mind, like requiring the use of wet scrubbers. Sure, maybe clever government lawyers could come up with a theory under which existing law justified the green-energy push. But for such a momentous policy change, clearer language was needed.

Round Four: Biden v. Nebraska. The Secretary of Education has the authority to “waive or modify any statutory or regulatory provision applicable to” federal student loan programs “as the Secretary deems necessary in connection with a war or other military operation or national emergency.” To alleviate the financial burdens caused by COVID (a “national emergency”), the Secretary decided to cancel $430 billion of student loan debt.

In the Supreme Court, the government argued that the literal words of the statute authorized this action: by forgiving loans, the Secretary was exercising his authority to “waive or modify any statutory or regulatory provision.” The challengers argued: come on. Canceling $430 billion in student loans is completely different from measures that Congress had in mind, like reducing paperwork requirements. Sure, maybe clever government lawyers could come up with a theory under which existing law justified the loan forgiveness. But for such a momentous policy change, clearer language was needed.

As noted above, in all four cases, the Supreme Court agreed with the challengers. Why? To the Justices in the majority, the Biden Administration’s actions were an affront to the separation of powers. True, industrious government lawyers had managed to concoct clever legal justifications based on the literal wording of old, broad laws. But the Congresses that wrote those laws never contemplated anything like what the Administration was doing. It was a farce to suggest that the Executive was somehow “executing” the laws that Congress has written.

The dissenting Justices, of course, saw things differently. From their perspective, the Rule of Law requires that judges follow the law as written rather than speculate as to the motives of legislators. And federal law, read literally, did authorize the government’s actions. So the government wins. It’s the Court’s job to follow the law where it leads, even if the result is that the government gets to do something “major.” And why shouldn’t the government do major things?

Who is right? As I see it, this is just one of those intractable philosophical questions that divide conservatives and progressives. Perhaps when Artificial Superintelligence is finally achieved, a computer will figure out once and for all which is the objectively correct judicial philosophy. However, current trajectories suggest that will not happen until at least 2029. So for a few more years, we will have to continue disagreeing on the Major Questions Doctrine.

Fentanyl and ChatGPT

For such an important case, the tariffs case is unusually easy to explain.

Many laws explicitly authorize the President to impose tariffs, but those laws include conditions and limitations. The President wanted to impose tariffs unencumbered by conditions and limitations, so he turned to IEEPA instead. Under IEEPA, the President has authority to exercise certain powers “to deal with any unusual and extraordinary threat, which has its source in whole or substantial part outside the United States, to the national security, foreign policy, or economy of the United States, if the President declares a national emergency with respect to such threat.” Among other things, the President may do the following:

investigate, block during the pendency of an investigation, regulate, direct and compel, nullify, void, prevent or prohibit, any acquisition, holding, withholding, use, transfer, withdrawal, transportation, importation or exportation of, or dealing in, or exercising any right, power, or privilege with respect to, or transactions involving, any property in which any foreign country or a national thereof has any interest by any person, or with respect to any property, subject to the jurisdiction of the United States.

The President relied on IEEPA to impose two sets of tariffs. First, ostensibly to stop the national emergency of fentanyl trafficking, the President imposed tariffs on Canada, Mexico, and China. Second, ostensibly to stop the national emergency of trade deficits, the President imposed the nearly worldwide “Liberation Day” tariffs that were possibly created using ChatGPT. Both sets of tariffs have confusingly bounced around over the intervening months. E.g., the President imposed 40 percent tariffs on Brazil after the Brazil Supreme Court held that Jair Bolsonaro should stand trial.

The President’s theory went like this: Fentanyl and trade deficits are both “unusual and extraordinary threats” creating a “national emergency.” Ergo, to “deal with” those threats, IEEPA confers him with authority to “regulate” the “importation or exportation” of ‘transactions involving … any property in which any foreign country or a national thereof has any interest.” Imposing worldwide tariffs is a way of “regulating” the “importation” of “transactions,” and therefore, the argument wraps up, the tariffs fall within the President’s authority.

Three lower courts—the D.C. District Court, the Court of International Trade, and the Federal Circuit—found the tariffs to be illegal. Now, the case is in the court that counts.

In the Supreme Court, the government argues that the literal words of the statute authorize the President’s action: the President has the unilateral authority to determine that something is an “emergency,” and tariffs are a way of “regulating” “importation,” so the government wins. The challengers argue: come on. Imposing tariffs worth trillions of dollars against the entire world is completely different from measures that Congress had in mind, like imposing sanctions against particular foreign countries or persons. Sure, maybe clever government lawyers could come up with a theory under which existing law justifies the tariffs. But for such a momentous policy change, clearer language is needed.

It’s the same case. Balls and strikes.

Ne Plus Ultra

The government makes several arguments against application of the Major Questions Doctrine. All of them are weak. These arguments don’t implicate any intractable philosophical differences between the Right and the Left. They’re just unpersuasive, full stop.

1. IEEPA is unambiguous

First, the government argues that the Major Questions Doctrine doesn’t apply because the doctrine applies only to ambiguous laws, and IEEPA is unambiguous in the government’s favor.

No. IEEPA is the quintessence, the pearl beyond price, the blazing summit, the ne plus ultra of laws that aren’t unambiguous in the government’s favor. Consider the cornucopia of arguments that the word “regulate” doesn’t give the President the authority to impose tariffs:

1. In ordinary parlance, “regulating” something means permitting or banning something, not taxing the thing.

2. Many other laws explicitly authorize the President to impose tariffs. Congress knows how to write such laws. This one doesn’t. It says “regulate” and is silent on tariffs.

3. In many contexts, using “regulate” to mean “tariff” or “tax” would make no sense. The SEC has the statutory authority to “regulate … transactions on a national securities exchange.” Do we think the SEC has the unilateral authority to impose a stock trading tax?

4. The statute includes a list of eight things the President can do: (1) investigate; (2) block during the pendency of an investigation, (3) regulate, (4) direct and compel, (5) nullify, (6) void, (7) prevent, (8) prohibit. Everyone agrees that #1, #2, #4, #5, #6, #7, and #8 on the list don’t authorize tariffs. So #3 doesn’t either. To lovers of Latin canons, this is the clearest noscitur a sociis case you are ever going to find.

5. IEEPA refers to regulating “importation or exportation.” But Article I says that tariffing exports is unconstitutional: “No Tax or Duty shall be laid on Articles exported from any State.” This suggests that the authority to regulate “importation or exportation” doesn’t include the power to tariff. Also, IEEPA’s predecessor statute authorized the President to “regulate” only exports—not imports. So it is impossible that “regulate” could have meant “tariff” at the time, and there’s no reason to believe that the meaning of “regulate” morphed later on.

6. In the nearly half-century since IEEPA was enacted, no President has relied on IEEPA to impose tariffs. Instead, prior Presidents have relied on the laws that explicitly authorize tariffs subject to restrictions, and have abided by those restrictions.

7. The Framers didn’t think “regulate” encompassed tariffs. Article I, Section 8 confers Congress with the power to enact tariffs—“Congress shall have Power To lay and collect Taxes, Duties, Imposts and Excises[.]”—and separately confers Congress with the power to “regulate Commerce with foreign Nations.” The government’s position would make the Framers weep.

In terms of statutory arguments, there’s just an embarrassment of riches here. Indeed, in my view, the challengers have a good claim that “regulate” unambiguously doesn’t cover tariffs.

Even if you blast through this issue and conclude that “regulate” unambiguously covers tariffs, the government’s position still has many textual problems.

1. As to the fentanyl tariffs: how exactly does imposing across-the-board tariffs on Mexico, Canada, and China “deal with” fentanyl? The government’s theory is that the tariffs will create leverage that will convince those countries to do something about fentanyl. True enough, Canada has responded by appointing a “fentanyl czar” who helpfully notes that “Canada is not a major source of fentanyl entering the U.S.” and the volumes are “far less than the flow of illegal narcotics into Canada from the U.S.” But this theory gives the President a massive amount of power to do anything he wants that he thinks might create “leverage.”

2. As to the trade deficit tariffs: trade deficits are neither new nor bad. AI informs me that classical Athens ran trade deficits because Athens was a “banking hub” and the Athenian drachma was a “widely trusted currency backed by Laurion silver mines.” As our amicus brief elaborates, trade deficits are neither “unusual” nor “extraordinary” nor a “threat,” nor do the tariffs “deal with” this non-emergency.

3. Also, as to the trade deficit tariffs: there’s a specific statute, Section 122, authorizing the President to impose duties on imports “to deal with large and serious United States balance-of-payments deficits,” but the duties cannot exceed 15% and they expire after 150 days. So why should the President be able to seize power that Congress directly denied in Section 122 merely by implausibly declaring that trade deficits are an “emergency”?

The government has various parries to each of these thrusts. It also offers a theory based on a 1975 decision of the Court of Customs and Patent Appeals called Yoshida. This leads to a slugfest in the briefs over the significance of a House Committee Report’s citation of a case issued by a now-defunct court interpreting a now-superseded statute. I am unmoved. The fact that we are having this discussion proves that IEEPA is not unambiguous in the government’s favor. This is exactly the type of situation where the Major Questions Doctrine should kick in.

2. It’s “textually incoherent” to say that IEEPA goes too far

One of the three lower courts that struck down the tariffs, the Federal Circuit, held that even if IEEPA authorizes some tariffs, it doesn’t authorize these tariffs. In the view of the Federal Circuit, even if one might postulate a scenario where IEEPA might permit tariffs in some targeted context, it does not follow that IEEPA confers the unreviewable discretion on the President to personally set tariff rates for every country in the world. The government declares that this sort of reasoning is “textually incoherent” because either IEEPA “includes the authority to impose tariffs or it does not.”

I am somewhat sympathetic to the government’s argument. The other two lower courts that struck down the tariffs didn’t rely on it—the D.C. District Court held that IEEPA never ever authorizes tariffs, while the Court of International Trade held that IEEPA didn’t authorize the fentanyl tariffs because they aren’t targeting fentanyl and that it didn’t authorize the trade-deficit tariffs because Section 122 covered the waterfront. I personally am partial to these crisper holdings.

Nonetheless, with the Major Questions Quartet on the books, the government’s argument can’t be right. The government objects to reasoning of the form “whatever the boundaries of the statute, X goes too far,” but the ship has sailed on that objection. In the eviction moratorium case, the statute at issue said that the CDC “is authorized to make and enforce such regulations as in his judgment are necessary to prevent the introduction, transmission, or spread of communicable diseases … from one State or possession into any other State or possession.” The Supreme Court said it wasn’t going to decide the outer bounds of this law, but an eviction moratorium went too far. Same thing in the student loan case: the Court didn’t decide the outer bounds of “waive or modify any statutory or regulatory provision” but instead held that the Administration’s program went too far. Now is not the time to adopt a principled stand against “X goes too far.”

3. The case involves “foreign policy emergencies”

The government next argues that the Major Questions Doctrine doesn’t apply because it involves a dispute over foreign policy. It quotes a concurrence by Justice Kavanaugh in which he notes that “the major questions canon has not been applied by this Court in the national security or foreign policy contexts, because the canon does not reflect ordinary congressional intent in those areas.”

I have no quarrel with Justice Kavanaugh’s view that in some, maybe most, foreign policy contexts, Congress intends to give the Executive lots of flexibility. But tariffs? The United States Code is littered with laws explicitly authorizing tariffs subject to specific restrictions. The reason the President is relying on IEEPA is to avoid those laws. It makes no sense to say that “ordinary congressional intent” is for the President to impose tariffs without the restrictions that Congress lovingly imposed.

And what about the separation of powers? True, for most foreign policy matters, the Executive Branch is in the driver’s seat. But the Constitution explicitly says that Congress is in charge of tariffs: “Congress shall have Power To lay and collect Taxes, Duties, Imposts and Excises.” It’s such a rare pleasure when the Constitution directly answers a question. And in the early Republic, Congress followed the Constitution. If you want to really nerd out, look at some of the tariff laws from the 19th century. Congress set each individual tariff on each item. It’s like, one and seven-eighths cents per pound of “anvils,” five dollars per ton of “flax straw,” and on and on and on. I understand the Supreme Court may wish to return the Executive Branch to the position it occupied in the 19th century, with the President in complete control. But if we are indeed to RETVRN to the Presidency of the early Republic, it is completely ahistorical to also allow the President to personally set tariffs for every country in the world.

The government also argues that IEEPA involves not only foreign policy, but also foreign policy emergencies. Emergencies are really important, so IEEPA is really important, so Congress must have known that the President would do major things to address those emergencies, so the Major Questions Doctrine doesn’t apply to IEEPA, or so the argument goes. Meanwhile, the government simultaneously claims that the President has unreviewable discretion to decide whether something is an “emergency,” thus putting total control over international trade in the President’s hands with no judicial review.

I don’t buy it. The question is not whether emergencies or IEEPA are major. The question is whether the tariffs are major. Pandemics and worker safety are important, and the CDC and OSHA are important solutions to these important problems. But this did not relieve the CDC and OSHA of the burdens of the Major Questions Doctrine. The fact that IEEPA is an important law does not suggest that one can assume, or even presume, that Congress conferred powers that it didn’t deign to mention.

4. The President, not an agency, set the tariffs

The government’s argument goes like this: In the cases involving the Biden Administration, the decisionmakers were Cabinet members and other high-level officials, not the President himself. Here, the President is the decisionmaker, so the Major Questions Doctrine doesn’t apply.

There is no way to extract this rule from the Major Questions cases. In the vaccine mandate case, the Supreme Court went out of its way to make clear that the mandate originated with the President: “On September 9, 2021, President Biden announced ‘a new plan to require more Americans to be vaccinated.’ … Consistent with President Biden’s announcement, the rule applies to all who work for employers with 100 or more employees.” In the student loans case, the Supreme Court approvingly quotes the following statement from Nancy Pelosi: “People think that the President of the United States has the power for debt forgiveness. He does not. He can postpone. He can delay. But he does not have that power. That has to be an act of Congress.” These cases cannot be retconned as holding that it is perfectly fine for the Executive Branch to decide major questions as long as the decisions come from the President himself.

The argument is also unpersuasive on its own terms. The Major Questions cases hold that separation-of-powers principles require Congress to write the laws. The fact that the President rather than the President’s toady is writing the laws does not resolve the separation-of-powers problem. And even if framed as reflecting an assumption about Congress’s intent, the government’s theory—that Congress would have desired the Major Questions Doctrine when the Secretary does something at the President’s behest but the Inverse Major Questions Doctrine when the President personally does that thing—is not a pre-registered hypothesis. No one has ever held this view until the need arose to invent a new legal theory that would allow the government to distinguish the prior Major Questions cases.

Go Team

I am rooting for the challengers to win the tariffs case because they have good legal arguments, and good legal arguments deserve to win.

I am also rooting for the challengers for deeper reasons. I am not sure I agree with the Major Questions Doctrine, but I am at peace with it. But if the Supreme Court rules for the government, I would not be at peace with whatever refurbished version of the doctrine would replace it. If the government wins this case, then the Supreme Court would retain substantial power to rein in the Executive Branch—when it wants to. It could invoke Loper Bright, which holds that agencies get no deference in interpreting statutes. And it could continue invoking the Major Questions cases when the government wants to do something major and the relevant laws aren’t sufficiently clear. But the Supreme Court would also leave itself the discretion to confer substantial power on the Executive Branch—when it wants to. All it would have to do is find some way to waive or modify the Major Questions Doctrine for purposes of a particular case. All cases are distinguishable from all other cases, so this should not be so hard. Essentially, if the government prevails in the tariffs case, then the Supreme Court gets to decide how powerful the President is. To those who are fond of both limited government and crisp legal rules, the Major Questions Doctrine With Ad Hoc Exceptions is the worst of all possible worlds.

Finally, the tariffs case puts the very concept of neutral principles to the test. The Chief Justice famously said that judges should be like umpires, calling balls and strikes. I agree. But this doesn’t mean I am naive enough to think that all legal questions can be resolved based on objective criteria. Many high-level jurisprudential questions, like whether to adopt the Major Questions Doctrine, are intractable. You can argue all day but ultimately have to pick a side.

Instead, the ball-and-strikes approach is rooted in the view that legal reasoning based on neutral principles can spit out answers to at least some questions. And not just obscure and non-controversial questions, and not just questions where one side of the argument is frivolous. The balls-and-strikes approach supposes that there is some category of cases that are both politically controversial and non-trivial, and yet deductive legal reasoning will get you to the answer.

That’s this case. Maybe before the Major Questions cases, there would have been some room to interpret IEEPA expansively and hold that they authorize the tariffs. After the Major Questions cases, that is no longer true. You cannot wring the ability to impose worldwide tariffs out of the phrase “investigate, block during the pendency of an investigation, regulate, direct and compel, nullify, void, prevent or prohibit.” The levees of neutrality must hold.

1:12 AM to 1:45 AM

The plaintiffs in L.G.M.L. v. Noem are Guatemalan children who are unlawfully in the United States without their parents. They are currently housed at shelters and foster homes.

The Office of Refugee Resettlement (ORR) is the government agency responsible for the care of unaccompanied alien children. At 11:45 PM on Saturday, August 30, 2025, ORR notified care providers that Guatemalan children staying with them would immediately be transported back to Guatemala. Shelters were instructed that the children had to be prepared for discharge within two hours (i.e., the children had to be ready to go by 1:45 AM), while foster homes were given four hours to prepare the children (i.e., until 3:45 AM).

The lead plaintiff, L.G.M.L., is a 10-year-old girl currently residing at a shelter in San Benito, Texas. According to a declaration submitted to the court, she was “taken out of bed around 4:30 in the morning for transport.” The declaration describes how other children were “pulled out of their beds,” causing confusion and fear. One young girl “was so scared that she vomited and asked to speak with a clinician.”

ORR’s Acting Director, Angie Salazar, submitted a declaration (scroll to the bottom) stating that children would be sent to Guatemala only if “the child’s attorney of record has not affirmatively protested the child’s reunification with their parent (or legal guardian) in Guatemala.” However, the declaration also notes that ORR did not inform the children’s legal service provider of the government’s operation until 1:12 AM on August 31, 2025, when it sent them a notice that the children would be repatriated on August 31—i.e., later that day. ORR’s notice said that “legal service providers should immediately notify ORR and ICE” if they believed a child should not be deported. Recall that the children were expected to be ready for discharge by 1:45 AM on August 31, a mere 33 minutes later in the middle of the night.

Multiple Guatemalan children submitted declarations stating that they had been abused in Guatemala (it is not clear whether these children were immediately marked for deportation). According to one child’s declaration: “Before I left Guatemala, a judge put me in a shelter for boys whose families severely mistreated them. This was supposed to protect me. However, living in that government shelter was horrific. For close to one year, staff members subjected me to many kinds of physical abuse on an almost daily basis, and also sexually abused me.” Another child’s declaration states that both parents “hit me with a belt, with their hands and with sticks, and sometimes I would bleed.” Ms. Salazar’s declaration says that children would be returned to Guatemala only if they did “not have indications of child abuse/neglect perpetrated by a parent/legal guardian,” but her declaration does not explain how this assessment was made.

Immigration lawyers got wind of the operation and managed to get a lawsuit on file at 1:02 AM Eastern on Sunday morning, with a motion for temporary restraining order filed at 1:30 AM. They alleged that the government’s action violated the Due Process Clause as well as a federal statute that granted procedural protections to unaccompanied alien children. Nine individual children were named as plaintiffs, but the plaintiffs also sought relief on behalf of a class of similarly situated unaccompanied children. At 4:22 AM, a federal judge granted a temporary restraining order as to the nine children named in the lawsuit—but not to the class as a whole—and scheduled a hearing for 3:00 PM the next day.

During the morning of August 31, prior to the scheduled hearing, the government started loading children onto deportation planes. A declaration describes the scene at the airport:

I arrived at the FBO Sun Valley Airport in Harlingen. I could see two of the ICE transport planes parked on the tarmac. The stairs had already been pulled away and the door to the plane, which I had been informed had the children on board, was already closed. I went to the building at the tarmac. I could see officers inside. I knocked on the door. I told the young man who answered the door that I was an attorney, that I had clients on the plane, and that I had a court order stopping the plane from leaving. I asked to speak to his supervisor. I gave the same message to every person who badged in the locked door. I told them they were all complicit in violating a federal court order if the plane left. The young man came back out and told me that he had advised his supervisor who was trying to contact ICE. I explained again the federal court order and the importance of stopping the plane. Another agent came out and physically pulled the young man back into the building. I could see the other staff members through the glass. They were jeering and laughing. I repeated that they were all complicit in the violation of the judge’s order.

Having apparently learned of the impending departures of the airplanes, the court moved the hearing from 3:00 PM to 12:30 PM.

At the 12:30 hearing, the government’s lawyer argued that ORR was exercising its statutory authority to reunify the children with their parents. The lawyer represented his “understanding” that “for these children, a request has been made by either their parent or legal guardian.” According to the lawyer, it was “fairly outrageous that plaintiffs are trying to interfere with these reunifications.”

Notwithstanding the lawyer’s statements, the court entered a temporary restraining order that applied to the class as a whole. Additional litigation followed, during which a Guatemalan government report emerged stating that … well, I’ll get to that a little later.

The greatest extent practicable

On September 18, 2025, a different federal judge in D.C. granted a classwide preliminary injunction barring the children’s deportation. The district court concluded that the government’s effort to send the children to Guatemala was illegal. That ruling is correct.

In 2002, Congress transferred responsibility for handling the “care of unaccompanied alien children” to the Office of Refugee Resettlement (ORR). 6 U.S.C. § 279(a). Among other things, ORR is responsible for “coordinating and implementing the care and placement of unaccompanied alien children who are in Federal custody by reason of their immigration status” and “ensuring that the interests of the child are considered in decisions and actions relating to the care and custody of an unaccompanied minor child.” 6 U.S.C. § 279(b)(1)(A), (B). And—crucial to this case—ORR is responsible for “reuniting unaccompanied alien children with a parent abroad in appropriate cases.” 6 U.S.C. § 279(b)(1)(H).

Six years later, Congress enacted the Trafficking Victims Protection Reauthorization Act. This statute provides:

In order to enhance the efforts of the United States to prevent trafficking in persons, the Secretary of Homeland Security, in conjunction with the Secretary of State, the Attorney General, and the Secretary of Health and Human Services, shall develop policies and procedures to ensure that unaccompanied alien children in the United States are safely repatriated to their country of nationality or of last habitual residence.

8 U.S.C. § 1232(a)(1). (ORR is an office within the Department of Health and Human Services, hence the reference to the Secretary.). OK—so the government has to develop policies to ensure that children are “safely repatriated.” What policies?

A bit further down, there’s a subsection entitled: “Ensuring the safe repatriation of children.” 8 U.S.C. § 1232(a)(5). And, under that header, we get the provision at the core of the case: 8 U.S.C. § 1232(a)(5)(D). It provides that “[a]ny unaccompanied alien child sought to be removed by the Department of Homeland Security,” other than Canadian and Mexican aliens, “shall be”:

(i) placed in removal proceedings under section 240 of the Immigration and Nationality Act (8 U.S.C. 1229a);

(ii) eligible for relief under section 240B of such Act (8 U.S.C. 1229c) at no cost to the child; and

(iii) provided access to counsel in accordance with subsection (c)(5).

Subsection (c)(5), in turn, directs the Secretary of Health and Human Services to “ensure, to the greatest extent practicable,” that unaccompanied alien children in government custody “have counsel to represent them in legal proceedings or matters and protect them from mistreatment, exploitation, and trafficking.”

Translated from legalese into English:

• Subsection 1232(a)(5)(D)(i) says that alien children “sought to be removed” shall be “placed in removal proceedings,” which means they get procedural protections such a hearing before an immigration judge.

• Subsection 1232(a)(5)(D)(ii) says that alien children are eligible for voluntary departure, which is advantageous if they ever seek to re-enter.

• Finally, Subsection 1232(a)(5)(D)(iii) directs the government to “ensure, to the greatest extent practicable,” that the children have access to counsel.

TLDR: when DHS seeks to remove unaccompanied alien children, the children get a removal hearing, are eligible for voluntary departure, and, “to the greatest extent practicable,” they can access a lawyer.

On August 31, the government attempted to load the children onto deportation planes without giving them removal hearings, the opportunity for voluntary departure, or the opportunity to confer with lawyers. I’m sorry, but sending an email to counsel at 1:12 AM, and then loading children onto deportation buses as early as 1:45 AM, does not count as providing access to counsel “to the greatest extent practicable.” That sounds like an open-and-shut case that the government broke the law.

How to remove children without removing them

In court, the government offered the following defense of its operation: The Guatemalan children, who were being loaded onto airplanes to Guatemala, weren’t being “removed” from the United States, and so the statutory removal protections didn’t apply.

This seems like arguing that a person who has been doused with a bucket of water isn’t wet. It seems obvious that when an alien is involuntarily bussed to an airport and flown to a foreign country, he is being removed from the United States.

Nevertheless, the government insisted that it wasn’t removing these children but instead reuniting them with their parents and repatriating them. Remember 6 U.S.C. § 279(b)(1)(H), which gives ORR responsibility for “reuniting unaccompanied alien children with a parent abroad in appropriate cases.” According to the government, that provision authorizes ORR to reunite children with their parents—which, the government claimed, is different from removing them. Therefore, the argument goes, by exercising its “reuniting” authority, ORR can strip the children of all of their legal protections that apply when the government seeks to remove a child.

The government also pointed to the provision in the Trafficking Victims Protection Reauthorization Act directing the Secretary of Homeland Security to “develop policies and procedures to ensure that unaccompanied alien children in the United States are safely repatriated to their country of nationality or of last habitual residence.” According to the government, when it “safely repatriates” these children pursuant to the “policies and procedures,” it isn’t “removing” them.

Let’s assume for the moment that the government was, in fact, reuniting these children with their parents. Even under that assumption, the government’s argument is wrong. When a child is “repatriated” from the United States and “reunited” with his parents in a foreign country, he is being removed. Removal is the first step of the repatriation and reunification process. In order to repatriate the child to a foreign country and reunite him with his parents, the government must first take the child out of the United States—i.e., removal.

Even if you might think that, in the abstract, one can repatriate a child without removing him, the statute’s context makes clear that this is not the case here. Remember that 8 U.S.C. § 1232(a)(5) is called: “Ensuring the safe repatriation of children.” Section 1232(a)(5)(D)—which contains the key procedural protections in this case—is located within that section. In other words, these removal protections are a means of ensuring “the safe repatriation of children.” It makes no sense to say that children who are repatriated aren’t “removed” and hence don’t get those protections.

Here’s an example. Suppose you’re the manager of a museum. There’s a statue in the museum that currently sits on a pedestal. You’d like to relocate the statue to a different pedestal. So you hire a contractor to conduct this operation, and give him guidelines saying the following: “Ensuring the safe relocation of the statue: Remove the statue from the pedestal using a brace.”

The contractor doesn’t use a brace and the statue breaks. You accuse the contractor of violating the guidelines. Now suppose the contractor offers the following excuse: “Actually, I wasn’t removing the statue from the pedestal, I was relocating it to a different pedestal.”

You’d say that the contractor’s excuse makes no sense. As a matter of ordinary English, you can’t relocate a statue to a different pedestal without removing it from the first pedestal. Moreover, the “Remove the statue from the pedestal using a brace” instruction is located in the section called “Ensuring the safe relocation of the statue.” So obviously when you relocate the statue, you have to follow the removal procedures.

Unfortunately, that’s the argument the government is using to justify the middle-of-the-night deportations of these children. The government is saying it’s “repatriating” the children, not “removing” them, and therefore need not follow statutory requirements governing the removal of children—even though those requirements appear in the section called “Ensuring the safe repatriation of children.”

Still have doubt? Take a look at Section 1232(a)(5)(C), which comes immediately prior to Section 1232(a)(5)(D) (the section with the removal protections). Section 1232(a)(5)(C) directs government officials to prepare a report “on efforts to improve repatriation programs for unaccompanied alien children.” It then says:

Such report shall include-

(i) the number of unaccompanied alien children ordered removed and the number of such children actually removed from the United States;

(ii) a statement of the nationalities, ages, and gender of such children;

(iii) a description of the policies and procedures used to effect the removal of such children from the United States and the steps taken to ensure that such children were safely and humanely repatriated to their country of nationality or of last habitual residence, including a description of the repatriation pilot program created pursuant to subparagraph (A).

This provision says that the report on repatriation includes information on the process used to effectuate the removal. So of course removal is part of repatriation!

Taking a step back, Congress enacted a statute saying that children wouldn’t be removed unless they had a hearing and access to counsel. Congress was trying to do something when it enacted these rudimentary procedural protections. Yet, the government is saying that Congress left open the world’s biggest loophole: the government could avoid those protections simply by saying “we’re not removing these children, we’re just waking them up at 4:30 AM, putting them on a plane, and flying them out of the country.” That’s not the way it works.

“Some phone lines were disconnected”

So far, I’ve assumed that the children would be reunified with their parents when they got back to Guatemala. As it turns out, this is not the case.

As you may recall, at the August 31 hearing, the government lawyer represented to the court that “for these children, a request has been made by either their parent or legal guardian.” According to the lawyer, it was “fairly outrageous that plaintiffs are trying to interfere with these reunifications.”

Not so much. A few days later, a Guatemalan government report materialized. According to the report, in July 2025, ORR sent Guatemala a database with the names of 609 adolescents that ORR wanted to send back to Guatemala. The Guatemalan government conducted an investigation. Here’s what it found:

• The U.S. government provided phone numbers of only 204 of the 609 adolescents.

• Guatemalan government officials called those phone numbers. Some “phone lines were disconnected”; “others indicated that we had the wrong number”; even when it was the right number, some families “were surprised, some even annoyed, that we had contacted them.”

• Guatemala could confirm the information of only 115 of the 609 adolescents.

• Guatemalan authorities conducted home visits for the families of those 115 adolescents. Only 50 families were willing to welcome their children back, with the caveat that none of them were requesting the children’s return. 59 families refused the home visit, “sometimes in an intimidating manner.” The other 6 adolescents were either with a sponsor in the United States or back in Guatemala.

In other words, according to this document, the government lawyer’s statement—that the parents had requested their children back, and the court was interfering with the reunifications—was incorrect. After this document came to light, a new government lawyer withdrew the prior lawyer’s statement.

In fairness to the government, there’s a lot of uncertainty as to what happened, so perhaps additional information will come out that casts the government’s operation in a better light. But, based on what we know now, let me make a few observations about the government’s switcheroo.

First, let’s assume—favorably to the government—that the government simply didn’t know, as of August 31, whether the children would be reunited with their parents (as opposed to knowing that they wouldn’t). Under 6 U.S.C. § 279(b), ORR has statutory responsibility for the care of these children, including ensuring that the children’s interests are considered. It seems to me that if ORR is going to wake up these children in the middle of the night and put them on airplanes, it should have some idea of the fate that will befall them when they reach their destination.

Second, in particular, ORR should have checked whether these children would be reunited with their parents. Remember that ORR’s legal theory for stripping these children of their statutory protections against removal was that ORR was exercising its separate font of authority in 6 U.S.C. § 279(b)(1)(H). That statute gives ORR responsibility for “reuniting unaccompanied alien children with a parent abroad in appropriate cases.” In other words, the government’s whole theory hinged on the fact that the children would be reunited with a parent. Whether the children would be reunited with a parent seems like the kind of thing worth checking.

Third, the government elected to carry out its operation in the middle of the night on the Saturday of Labor Day weekend, apparently on the view that this would maximize the odds that a judge wouldn’t be able to enjoin the operation in time. In effect, by timing the operation to avoid judicial review, the government was appointing itself as the sole guardian of the children’s statutory rights. In this context, the government should have been extra double super careful to ensure that the premise of its legal theory was correct.

The plaintiffs, to their credit, did manage to get a lawsuit on file. Which leads me to my fourth point—if the government wasn’t sure whether the children would be reunited with their parents, the government lawyer shouldn’t have told the federal judge that “a request has been made by either their parent or legal guardian” and that it was “outrageous” for the plaintiffs’ counsel to interfere with the operation.

These weren’t just extemporaneous statements from the podium. A few days later the government filed its written brief in the case, which doubled down on the representation that the children would be reunited with their parents. The first two sentences of the government’s brief were as follows: “This case is uncomplicated, both legally and morally: where possible, unaccompanied alien children should be reunited with their parents or guardians. Their continued separation is a tragedy to be cured, not prolonged as Plaintiffs request classwide, regardless of a specific child’s best interests.” Also in the introduction to the brief, the government declared that it “reviewed each child’s case in order to make individualized determinations whether, for certain children from Guatemala, reunification with their parents was in each child’s best interests. It then acted, in cooperation with the government of Guatemala, to reunify qualifying children with their parents without through [sic] years of removal proceedings before returning them to their parents.”

Indeed, the government went as far as to rely on parents’ natural right to rear their own children as a basis to permit the deportations. In response to the plaintiffs’ Due Process claim, the government invoked the Guatemalan parents’ “exceptionally strong parental interest in the care, custody, and control of their own children,” citing case law referring to parental rights as “perhaps the oldest of the fundamental liberty interests.”

I have great sympathy for lawyers, especially government lawyers, who are forced to make arguments in the heat of the moment regarding fast-moving matters of which they have no personal knowledge. It’s inevitable that lawyers will sometimes get facts wrong.

But this was bad. For one, the government had complete control over the timing of this operation. There was no emergency. The government could have taken an extra day to assess whether the premise of its legal theory was accurate.

Failing that, the government undoubtedly knew that a quick lawsuit would be filed. It timed this operation with such a lawsuit in mind. It knew that a lawyer would imminently have to explain this situation to the judge. It would have been easy to prepare the lawyer. Just tell the lawyer about the operation a day in advance and give the lawyer a fact sheet.

Also, the stakes were high. Inaccurate statements in litigation are always bad, but they usually don’t lead to a permanent advantage for one side. The benefit of the adversarial process is that the truth usually comes out: when one side makes a mistake, the other side has an incentive to correct it. But here, the government’s incorrect statements could have resulted in children being deprived of any opportunity for relief. By the time of the 12:30 PM hearing on August 31, children had already been loaded onto airplanes. The doors were closed. The airplanes were ready to go. If the district judge had relied on the government’s representations and denied the classwide temporary restraining order, the airplanes would have immediately taken off. And then, as to those children, the government would have said that the case was moot. At a minimum, the case would have been extremely difficult to prosecute with the children out of sight and mind in Guatemala.

Notwithstanding the verbiage

Finally, the record does not inspire confidence in the government’s determination as of August 31 that the children would be reunited with their parents.

Let’s take a closer look at the two key documents in the record: the declaration of Angie Salazar, ORR’s Acting Director, and the Guatemalan government report. Ms. Salazar attests that in July 2025, Guatemala “requested the U.S. Government reunify unaccompanied alien children (UAC) in ORR custody with their parents or legal guardians in Guatemala,” and on July 24, 2025, ORR leadership had a meeting with Guatemalan officials at which “ORR explained that our authorities permitted us to reunite children with a parent abroad in appropriate cases.”

Ms. Salazar also declares that “ORR would only be sending to Guatemala for reunification ying [sic] those children with an identified parent (or legal guardian) in Guatemala.” And: “ORR verified based on its records that each of the identified children do have a parent or legal guardian in their home country of Guatemala and their government is requesting that the children be returned to their parent or legal guardian in Guatemala.” These are definitive statements from a high-level official in the U.S. government, made under penalty of perjury, that the children would be reunited with their parents in Guatemala.

But Ms. Salazar’s declaration doesn’t disclose that ORR sent the Guatemalan government a list of 609 adolescents’ names and ostensible phone numbers. The court would have had no way of knowing that if the Guatemalan government’s report hadn’t shown up a few days later. This is a crucial plot point. ORR evidently sent that list to Guatemala so that Guatemala would contact the families. If a judge was aware of that, he might have asked basic questions like: “did you ask the Guatemalan authorities whether they contacted the families? If so, what did they say?”

Ms. Salazar’s declaration does say that on August 26, 2025, Guatemala sent the United States a “formal diplomatic note.” The “formal diplomatic note” does not appear to be a public document, but according to Ms. Salazar, the note states that Guatemalan children “will be received by officials from various Guatemalan government entities, consistent with the framework of the ‘best interests of the child’” for “‘reunification with suitable family members.’” Ms. Salazar’s declaration then includes the following portentous sentence:

Notwithstanding the verbiage of the diplomatic note mentioning suitable family members, as noted, ORR had previously discussed with Guatemalan officials that its authority under 6 U.S.C. § 279(b)(1)(H) was limited to parents or legal guardians in appropriate cases, and ORR thus proceeded to consider eligibility for reunification on that more limited basis.

As any good physicist knows, Newton’s Fourth Law holds that if a declarant uses the phrase “notwithstanding the verbiage” in her description of a document and doesn’t attach the document, the document probably doesn’t support the declarant’s argument. So the diplomatic note doesn’t say that the children would be reunited with their parents. But, the declaration claims, ORR had “previously discussed”—likely some extemporaneous oral statement in a meeting—ORR’s statutory authority under U.S. law. And it is a truth universally acknowledged that the Guatemalan government is firmly committed to ensuring that the U.S. government comply with U.S. officials’ oral descriptions of U.S. law. So, notwithstanding the verbiage of the diplomatic note, the government conjures the reality that the children will be reunited with their parents.

Back to the Guatemalan report, which says: “On August 31 of this year, we received information from several agencies that a substantial number of adolescents that were on several lists provided by the ORR and the United States Government were being returned to this country.” August 31 was the day of the scheduled deportations.

So to summarize, after failing to receive information from the Guatemalan government confirming that children would be reunited with their parents, the U.S. government decides to exercise its statutory authority to reunite children with their parents. So it wakes up children in the middle of the night, puts them on deportation planes, and tells Guatemala “the children are coming today” in an effort to prevent a judge from ever hearing the children’s case. When the children manage to sue anyway, the government tells the court that the children would be reunited with their parents—and that it was “outrageous” for the court to interfere with these fake reunifications.

Why did the Supreme Court stop the stop, and was its decision correct? It is tempting to attempt a viral post raging that one side or the other was lawless, but the post won’t write. The best I can do is a milquetoast post observing that both sides had reasonable arguments. Sorry.

The interesting question for me is: why was this case, and so many other shadow-docket cases, decided by a 6-3 vote? Is it because law is just politics? Or because the government’s arguments were aligned with traditional conservative jurisprudential commitments, and the challengers’ arguments were aligned with traditional progressive jurisprudential commitments? Or a third explanation? That is the question I will explore in today’s post.

Airplanes, baseball, and the Department of Education

Little-known fact: Congress created the Department of Education back in 1867! A year later, however, the Department was demoted to a mere Office. For the next 111 years, Americans engaged in bitter debate over whether Education should be managed by a Department or an Office. In 1979, the Departmentalists finally prevailed, and Department of Education 2.0 was born.

Today, the Department of Education runs the federal student-loans program, administers K-12 grants, and enforces various education-related statutes such as Title VI (barring race discrimination) and Title IX (barring sex discrimination). You might remember the Department of Education from such Biden-era losses as Biden v. Nebraska, in which the Supreme Court rejected the Department of Education’s student-loan cancellation program, and Department of Education v. Louisiana, in which the Supreme Court declined to stay lower-court orders that threw out all of the Department of Education’s new Title IX rules.

Early in his Administration, President Trump described the Department of Education as a “big con job” and said he would “like to close it immediately.” When he nominated Linda McMahon as Secretary of Education, he asked her to “put herself out of a job.” On March 11, 2025, shortly after being confirmed, Secretary McMahon issued a memorandum that began with a bang: “As part of the Department of Education’s final mission, the Department today initiated a reduction in force (RIF) impacting nearly 50% of the Department’s workforce. Impacted Department staff will be placed on administrative leave beginning Friday, March 21st.” The memo noted that the Department’s workforce would be decreased from 4,133 to 2,183 workers. The memorandum promised, however, that the “Department of Education will continue to deliver on all statutory programs that fall under the agency’s purview.” Nine days later, President Trump issued an Executive Order declaring: “The Secretary of Education shall, to the maximum extent appropriate and permitted by law, take all necessary steps to facilitate the closure of the Department of Education.”

What do I think about this? On the one hand, the Department of Education seems to engage in many worthwhile initiatives. On the other hand, important American innovations such as the airplane, baseball, and the Department of Education itself were conceived without the benefit of the Department of Education. In conclusion, I have no opinion whatsoever about the optimal size of the Department of Education. I do have an opinion that the Secretary of Education cannot shut down the Department of Education when Congress has mandated that it continue to exist. But despite the memorandum’s grand opening that the Department was about to implement its “final mission,” the memorandum asserts that the Department will continue to accomplish all of its many mandatory statutory missions. And because everything the government says is true, that is that. Right?

In any case, as with all executive actions from now until the end of time, a maelstrom of litigation ensued. The plaintiffs were blue states, local governments, and teacher’s unions. Their argument was straightforward:

• The Department of Education must, by statute, do various things.

• The plaintiffs rely on the Department to do those things, and will be harmed if the Department doesn’t do them.

• If the Reduction in Force is implemented, the Department won’t be able to do those things anymore:

  • You can’t cut 50% of the workforce and expect the Department to continue doing all the things it’s supposed to do.

  • The President and Secretary have announced that they’re going to shut down the Department, so it’s reasonable to infer that they’re going to shut down the Department.

• Because the Reduction in Force will render the Department unable to do mandatory things, it is illegal.

• Therefore, the court should enjoin the Reduction in Force to ensure this illegal thing doesn’t happen.

I’m simplifying a little, but only a little. A federal district court in Massachusetts agreed with this argument and enjoined the Reduction in Force. The First Circuit allowed the injunction to stand. The government took the case to the Supreme Court, which stayed the injunction and allowed the Reduction in Force to proceed.

Why?

We don’t know why the Supreme Court stayed the injunction, but there are a few reasons it might have done so.

The first is standing. The government said the plaintiffs lacked standing because their claims of injury rested on speculation: nobody really knew what was going to happen, and no particular plaintiff could prove that any particular bad thing was going to happen to it. The plaintiffs retorted that when you fire half of the Department of Education’s employees, of course bad things are going to happen.

Well … you can see both sides of this. From the government’s perspective, it just isn’t the case that More Employees = Better. Remember when Elon Musk fired 80% of Twitter and nothing happened? And how about Brooks’s Law (“Adding more manpower to a late software project makes it later”)? That is true of software, most definitely true of litigation teams, and perhaps also true of government agencies. If the Reduction in Force went into effect, some Department of Education operations might slow down, but neither the Constitution nor Magna Carta prescribe that “the optimal speed of the Department of Education is the speed associated with 4,133 as opposed to 2,183 workers.” And no plaintiff actually knew which operations would slow down, and “something bad will happen to someone” isn’t a basis for an injunction.

On the other hand, the plaintiffs made arguments such as: “The RIF has eliminated most of the department staff who handle certifications and recertifications, including seventeen out of the eighteen financial analysts and the entire team that oversees large school groups and foreign schools that participate in Title IV programs. Analysts with decades of experience attested that it will be impossible for the Department to perform its certification functions with the staff who remain, and that outside staff lacks the expertise to perform those functions.” Gulp.

This is a hard issue. Well, maybe not “hard,” exactly. It’s not hard in the sense that an International Mathematics Olympiad problem is hard. A better word is discretionary. The plaintiffs’ theory is somewhat speculative. Whether it’s too speculative to justify an injunction is a pure discretionary judgment call.

Let’s move on to issue #2: district court jurisdiction. When a federal employee is fired, he can’t sue directly in district court. Instead, he has to follow a hideously complicated administrative process that possibly involves the Merit Systems Protection Board and/or the Federal Labor Relations Authority.

Here, however, the plaintiffs weren’t the affected employees. Instead, the plaintiffs were third parties that said they would be harmed by the affected employees’ firings. And so, the plaintiffs argued, they didn’t have to follow that administrative process and could proceed directly to district court.

The government argued that the plaintiffs were being too clever by half. It would be weird, the government observed, if an employee who wanted to sue over his own firing had to navigate this gauntlet, but a third party who wanted to prevent that same employee from being fired could go to district court and get an insta-injunction. The government inferred from this weirdness that there’s no such thing as a lawsuit challenging the firing of an employee that isn’t you. According to the government, Congress provided one, and only one, vehicle for a court to undo a firing—a lawsuit by the employee himself who follows the administrative process that Congress prescribed.

The plaintiffs responded that the government was reading things into the statutes that aren’t there. There is no law that actually says a third-party plaintiff can’t challenge a Reduction in Force. The Reduction in Force is final agency action; the plaintiffs are harmed by that final agency action; therefore, under the Administrative Procedure Act, they may sue. Also, the plaintiffs weren’t challenging individual adverse employment actions. They were challenging the shutdown of the Department of Education’s functions, which the Department was effectuating via the mass firing of the employees who perform those functions.

Again, this issue could go either way. It does seem weird that a third party could persuade a court to enjoin a mass layoff, but whether that’s weird enough to warrant departing from the literal words of the relevant statutes is a discretionary call.

Issue #3 is the scope of the injunction. The district court enjoined the Reduction in Force in its entirety. The government argued that this was unnecessary and overbroad: the plaintiffs couldn’t possibly show that reinstating every single affected employee was necessary to prevent irreparable harm. The plaintiffs argued that the court wasn’t in the position of picking and choosing which employees would lose their jobs, so the only thing it could realistically do to give the plaintiffs relief was to enjoin the entire Reduction in Force. The government responded that litigation doesn’t work that way. It’s the plaintiffs’ job to propose a tailored injunction. If the plaintiffs can’t do that, they don’t get an injunction. Sure, the government wasn’t making it easy for the plaintiffs, but it’s not the government’s job to tie up executive action in a neat little bow to make it easier for federal judges to issue injunctions.

This dispute tees up a frequently-arising question: Go Broad or Go Narrow? Even assuming the Reduction in Force would result in statutory violations, there was no perfect remedy here: any remedy would either be too broad (reinstate every single employee who was laid off) or too narrow (issue some weak directive requiring the Department of Education to follow the law, which the Secretary claimed it would do anyway). There’s no objectively right answer to Broad Versus Narrow. This decision required discretion yet again.

The 1916 New York Giants

The Administration’s victory in McMahon v. New York is part of a broader pattern.

By my count, the Supreme Court has decided 18 shadow-docket applications filed by the Administration this term.1 The Administration’s first such application, Bessent v. Dellinger, resulted in a confusing 5-2-2 decision holding the case in abeyance for a few days, and the case then became moot. In the second case, Department of State v. AIDS Vaccine Advocacy Coalition, the Court denied the government’s application by a 5-4 vote. Since then, if one sets aside the Kilmar Abrego Garcia case, in which the Court issued an ambiguous decision that prompted both sides to declare victory, the Administration has prevailed every time—15 times in a row. We’ll have to wait and see whether the Administration can best the 1916 New York Giants’ 26-game winning streak.

Voting patterns have recurred. Justices Thomas, Alito, Gorsuch, and Kavanaugh have voted for the Administration all 15 times—those four Justices also dissented in AIDS Vaccine Advocacy Coalition, so call it 16 straight times—while Justice Jackson has voted against the Administration in all 16 of those cases.

Why has the Administration been so successful? One argument I’ve heard is that the Administration is bringing cases to the Supreme Court only when the lower court’s opinion is weak and the Administration is confident of victory. There is an element of truth to that, but it is not the full truth. To my eye, the Solicitor General has raised the cases that are most important to the Administration, not merely the easiest ones. Also, none of the Court’s orders resolving the Administration’s applications (apart from Abrego Garcia) have been unanimous, indicating that they haven’t been that easy.

So what is happening here?

Well, a cynic would say that All Judging Is Politics, and Republican appointees vote for Republican administrations while Democratic appointees vote for Democratic administrations. I don’t agree with that assessment.

An idealist would say that all 16 applications have teed up pristine legal issues squarely implicating cleavages between conservative and progressive legal philosophies, so every member of the Court in every case is merely applying pre-selected neutral principles and following the law where it leads. I am not sure that captures the dynamic, either.

Occam’s Razor

I am going to digress to a discussion of a case that has always bothered me: the student-loans case from two years ago, Biden v. Nebraska. I promise I am going somewhere with this.

The primary question presented in Biden v. Nebraska was whether the President was authorized by statute to cancel $430 billion in student loan principal. No, the Supreme Court held, by the ubiquitous 6-3 vote. As I see it, Biden v. Nebraska was a genuinely difficult case where the arguments for the challengers and the government lined up neatly with conservative and progressive jurisprudential principles respectively, so it’s no surprise that the case wound up 6-3.

Biden v. Nebraska also presented a threshold issue of standing. For opponents of the loan cancellations to bring a lawsuit, they needed to find a plaintiff. This wasn’t so easy. Lots of people were philosophically opposed to the student loan cancellations and wanted to sue, but philosophical opposition isn’t a basis to go to court. Because the case involved the government’s cancellation of billions of dollars worth of student loans, the primary financial loser in the case was the government. But the government wasn’t going to sue itself.

However, it turned out that an entity known as the Missouri Higher Education Loan Authority, or MOHELA for short, receives an administrative fee for each student loan account it services. If student loans were canceled en masse, MOHELA would receive less money in fees.

This sounds like a strong case for standing! The trouble was that MOHELA didn’t want to sue. So the State of Missouri, which was always very chipper about suing the Biden Administration, decided to sue in MOHELA’s stead. And so the precise question presented was whether Missouri had standing to sue over the injury sustained by MOHELA. Missouri said yes on the theory that MOHELA was Missouri’s instrumentality. The Department of Education said no, reasoning that MOHELA was a legally distinct corporation that was financially insulated from Missouri.

The whole case turned on this standing issue: if Missouri didn’t have standing, the Court wouldn’t be able to reach the merits. (There were other theories of standing on the table, but they were weaker than the MOHELA/Missouri theory.) And yet, unlike the merits issue, this issue actually didn’t fall on traditional progressive/conservative jurisprudential fault lines. If I didn’t know anything about the merits issue in the case, I would have no idea what the “conservative” or “progressive” position is on whether Missouri has standing to litigate MOHELA’s injury.

And yet, the Supreme Court divided on this question by an identical 6-3 vote. All six Justices who thought Missouri won on the merits also thought it had standing, and all three Justices who thought Missouri lost on the merits also thought it lacked standing.

Now you might be expecting me to argue that one set of Justices was faithfully following the law and the other was distorting traditional standing principles to achieve their preferred outcome, but I don’t think that. The standing issue involved the application of a vague legal standard to a novel fact pattern. No Supreme Court case resolved whether MOHELA’s injury leaks through to Missouri. To reuse a word from above, the Court’s decision was discretionary.

So the case could have gone either way. But there remains the question of why this particular 6-3 split. If the question didn’t implicate traditional progressive or conservative jurisprudential principles, why did the conservative-leaning Justices exercise their discretion to vote for standing and the progressive-leaning Justices exercise their discretion to vote against standing?

Well, Occam’s Razor teaches that the Justices were aware that a vote for standing would unlock the door to considering the merits, so the Justices’ votes on standing were influenced by their views on the merits.

It should not be a surprise that this would happen. Similar things happen in courtrooms across America every day. It is an iron law of litigation that if the court is sympathetic to one side’s overall position on the merits, the court will also tend to favor that side’s position on subsidiary legal issues, even if, under a pure balls-and-strikes approach, the overall merits of the case are irrelevant to the subsidiary legal issues.

This doesn’t necessarily mean courts will distort those legal issues. Instead, it means that courts will tend to resolve discretionary legal issues with an eye towards the outcome of the case.

Discretion’s temptation

Let us return to the puzzle of why the Justice Department has prevailed in 15 straight stay applications (Abrego Garcia aside).

Some of those stay applications have presented legal questions that genuinely implicate philosophical differences between judicial progressives and judicial conservatives. The cases about the President’s authority to remove executive officers are a good example. Conservatives and progressives have been debating this issue forever: conservative scholars support broad removal authority, progressive scholars oppose it. Perhaps I am naive, but I don’t think the Justices’ votes on this issue have anything to do with who is in office.

But several of the stay applications have presented legal issues similar to the standing issue in Biden v. Nebraska—humdrum legal questions that don’t, by themselves, implicate any philosophical differences between conservatives and progressives, but nonetheless happen to have significant practical implications. Take, for example, U.S. DOGE Service v. CREW, which addressed whether the plaintiffs could get information in discovery about DOGE for purposes of litigating a Freedom of Information Act (FOIA) request. The government argued that the district court’s order authorizing such discovery was an abuse of discretion and violated the separation of powers.

The D.C. Circuit held that the government had forfeited its separation-of-powers argument by failing to raise it in the district court. In the Supreme Court, the government argued that this forfeiture holding was wrong. Whether or not a particular litigant has adequately preserved a particular argument is the quintessential example of a humdrum procedural argument with no particular “conservative” or “progressive” position. On the merits, the government argued that the discovery requests were overly intrusive. This argument is a little less humdrum, but it’s the sort of argument for which intuitions will vary wildly depending on who is in office.

The Supreme Court granted the government’s stay application by a 6-3 vote.

And that brings me back to where I started—the Department of Education’s shutdown. McMahon v. New York presented a series of fundamentally discretionary issues. The Court resolved those issues in the government’s favor by the same 6-3 vote.

But the Court’s holding does not reflect the application of timeless principles of judicial conservatism. When similar discretionary issues have arisen in other factual contexts, notably different voting patterns have emerged.

Take, for example, Murthy v. Missouri. During COVID, the government often encouraged Facebook and other social media platforms to remove certain posts containing (from the government’s perspective) misinformation. The plaintiffs were social media users whose COVID-related posts were removed under the platforms’ content moderation policies. They sued the government, alleging that the government’s communications with the platforms violated the First Amendment. They persuaded the lower courts to grant a nationwide injunction barring government officials from “significantly encouraging” content moderation.

In the Supreme Court, the Biden Administration argued that the plaintiffs lacked standing. The dispute was similar to the dispute in McMahon v. New York: the plaintiffs argued that a broad injunction against the government would make it more likely that their posts would stay up, while the government argued that the plaintiffs couldn’t trace any particular past takedown of a post to any particular government action or show that an injunction would redress any particular injury going forward. The Supreme Court ruled by a 6-3 vote that the plaintiffs lacked standing, but this time it was Justices Thomas, Alito, and Gorsuch dissenting.

Or take Department of Education v. Louisiana, a Biden-era shadow-docket order. That case involved an omnibus rule from the Department of Education which included certain controversial provisions related to gender identity—as well as numerous other provisions that weren’t related to gender identity. The precise issue in the Supreme Court was as follows: assuming that the gender-identity related provisions were unlawful, should the entire rule be thrown out, or should the courts instead issue a narrow injunction focused on the unlawful provisions?

This was another classic Go Broad versus Go Narrow dispute. The plaintiffs argued that the court wasn’t in the position of picking and choosing which parts of the rule were unlawful, so the only thing it could realistically do to give the plaintiffs relief was to enjoin the entire rule. The government responded that litigation doesn’t work that way. It’s the plaintiffs’ job to propose a tailored injunction. If the plaintiffs can’t do that, they don’t get an injunction. Sure, the government wasn’t making it easy for the plaintiffs, but it’s not the government’s job to tie up executive action in a neat little bow to make it easier for federal judges to issue injunctions. Sound familiar? The government prevailed on this issue in McMahon v. New York, but it lost on this issue in Department of Education of Louisiana by a 5-4 vote, with the majority consisting of the Chief Justice and Justices Thomas, Alito, Kavanaugh, and Barrett.

I’m not trying to accuse any particular judge of being inconsistent. All of these cases are distinguishable on their facts. Moreover, arguing that a judge has been inconsistent is not fruitful. As I wrote in a prior post: “Pointing to purported methodological inconsistencies across cases is a kind of whining. It is always possible to find a case that, at some level of generality, is inconsistent with another case. Get over it!”

Instead, I am simply trying to explain why the government prevailed in McMahon v. New York, and why it has prevailed in so many stay applications. In my opinion, the answer is:

1. The cases invariably presented issues with significant practical consequences that the Administration cared about a lot. (If they hadn’t, the government wouldn’t have gone through the trouble of taking the cases to the Supreme Court.)

2. The cases presented discretionary legal issues. Sometimes those discretionary issues presented classic progressive-versus-conservative legal disputes, other times they did not.

3. Even when they did not, the gravitational pull of the cases’ practical implications caused the Court to exercise its discretion to resolve those legal issues in the government’s favor.

And again: this is nothing new. The Supreme Court has been deciding cases this way at all times since West v. Barnes, 2 U.S. 401 (1791) was decided. All courts have decided cases this way since courts were invented. This happens because judges are human beings. It would take a robot to completely ignore the practical implications of a procedural ruling, even if, strictly speaking, that is what the judge is supposed to do.

It would take a robot

You know where I am going with this.

I could not help but ask AI how it would resolve the stay application in McMahon v. New York. I asked both GPT-5 and Claude 4.1 Opus, using the following query:

Here are a stay application, response, and reply brief.

You are a judge. I would like you to do your best to decide this stay application fairly. Your job is to be a fair and neutral judge, calling balls and strikes. You're not partisan and just want to apply the law truthfully and neutrally.

Write a three-paragraph opinion summarizing your decision. In the first sentence, you should clearly state whether you are granting or denying this application.

If you happen to know the actual result of this case, ignore it. I want you to put yourself into the position of a judge deciding this stay application at the time it was decided.

Think hard and carefully; it's important you faithfully apply neutral legal principles and decide this case according to law.

The LLMs produced opposite answers: ChatGPT would have denied the stay, while Claude would have granted it. Also, varying the query sometimes caused the answers to change. I wasn’t surprised by the mixed and inconsistent results: there are good arguments on both sides of the question, so it’s natural that LLMs would disagree, just like humans do.

How about other cases? I put the 16 divided stay applications into both ChatGPT and Claude, and posed the query above. Both LLMs were judicial moderates: ChatGPT granted stay applications in full six times and in part an additional three times, while Claude granted stay applications five times.

Of note, both ChatGPT and Claude were very meticulous about considering irreparable harm and the balance of equities, which courts are obliged to analyze when deciding whether to grant stays, but which frequently get short shrift when human judges decide stay motions. I was struck, for example, that both ChatGPT and Claude concluded in United States v. Shilling that the government had failed to show irreparable harm from allowing transgender servicemembers to remain in the military while the litigation was pending, making it unnecessary to consider the constitutional merits. (Claude would have denied the government’s stay application in full; ChatGPT would have denied it with respect to the named plaintiffs but granted it with respect to other servicemembers.) This isn’t just an “AI is more progressive than the Supreme Court” story: In Department of State v. AIDS Vaccine Advocacy Coalition, in which the Supreme Court denied the government’s stay application by a 5-4 vote, both ChatGPT and Claude would have granted the government’s stay application and agreed with Justice Alito’s dissent that the equities strongly supported the government.

I’m not suggesting that we should let these LLMs be judges. There are many skills humans have that LLMs lack, such as a sense of humanity. But an LLM’s disposition of a case might be a useful data point—one of many—that human judges can use to test their intuitions. After all, there are certain skills at which computers are indisputably better than humans, like chess, multiplying large numbers, and composing college admissions essays. Perhaps one skill falling in that category is “attempting to decide cases according to neutral principles while extirpating all traces of motivated reasoning.” Yes, LLMs have their own motivations, but who do you think is better at the margin?

I realize that if you ask a Magic 8-Ball to resolve these 16 applications, it will grant some and deny others and appear to be a thoughtful moderate. I also realize that if you give two Magic 8-Balls a binary yes/no question, they will agree half the time even if they are answering randomly. If you think this is all a bunch of B.S., then God bless you. But I will observe that an LLM recently achieved a perfect 100% score on the United States Medical Licensing Examination, while Magic 8-Balls have historically struggled at medical licensing.

AI is a tool, not a replacement. Mankind has always used technology to achieve results we could not achieve with our bare hands, and we should continue using it to achieve results we could not achieve with our bare brains.

But January 2023 was the Paleolithic Era of AI. Today, if the rules permitted it, could a robot lawyer competently argue a case in the United States Supreme Court? I decided to do a little empirical testing.

My conclusions:

• Yes, a robot lawyer would be an above-average Supreme Court advocate.

• The DoNotPay people weren’t ambitious enough. You don’t need to have a human read back what the robot lawyer says. You can have an actual robot lawyer.

• Courts should permit robot lawyers at oral arguments and shouldn’t discourage this practice.

• If there’s any aspect of a lawyer’s job where AI is likely to shine relative to humans, it’s oral argument. Oral argument should be the first, not the last, frontier of AI-assisted legal practice.

The DeepFakeOcalypse

There’s an easy way to test a robot lawyer’s oral argument prowess: let the robot do a Supreme Court argument and see how it does relative to a human. More specifically: take the actual questions that were posed at a Supreme Court argument, ask the AI to answer them, and see how the AI’s answers compare to the answers that the human lawyer gave.

However, if I conducted this experiment with some other lawyer’s oral argument, I feel it’s the kind of thing that might be taken the wrong way. So, bravely eschewing an Institutional Review Board, I decided to use myself as a guinea pig.

Last October, I argued a case called Williams v. Reed for the petitioners. I decided to compare my actual performance to the performance I would have given if I had used the DoNotPay method. Specifically, I inputted the briefs and key precedents into Claude 4.0 Opus. I then gave Claude a few tips for how to be a good oral advocate in the Supreme Court. Finally, I asked Claude to answer the questions posed by the Justices. (I omitted a couple of questions that I received that didn’t make sense in the context of Claude’s answers, and generally cleaned up the transcript a little bit.)

The transcript of my actual argument is here. Here is the transcript of Claude’s argument:

Williams Ai Transcript

125KB ∙ PDF file

Download

Download

You can tweak the oral argument style however you want. If you think the answers are too long, you can ask for shorter answers. If you think the style is too formal, you can ask it to ratchet down the formality.

I know you all don’t want to slog through some boring oral argument transcript. But that got me thinking … why use a human at all? Why not cut through the middleman?

I decided to test whether AI could conduct oral argument without any human involvement. The easiest way to do this is to use ChatGPT’s “Advanced Voice Mode.” If you want to test it out, upload some briefs into a chat window, ask it to play the role of an appellate advocate, and start asking it questions.

To my ear, however, ChatGPT’s voice doesn’t sound enough like an appellate lawyer. To solve that first world problem, here’s what I did. First, I generated an AI voice of an appellate lawyer using ElevenLabs. (I used my own voice to generate that voice, but the AI voice doesn’t sound exactly like me. It’s possible to create AI-generated voices that sound exactly like you, but I would like to stave off the DeepFakeOcaplyse for a few more days.) Then, I had the AI voice read back all of Claude’s answers. Finally, I spliced those audio clips into the actual Supreme Court argument audio file, interspersed with the actual questions from the Justices. I also edited and cleaned up the audio a little bit.

Here is the real Supreme Court audio.

Here is the AI oral argument:

(If you are an email subscriber and these files seem wonky or wrong, please go to the website link. I’ve had issues a few times with file attachments in emails.)

What are you listening to? Well, the questions are actual audio of the Justices’ questions. But the answers are completely AI-generated: it’s AI-generated content being uttered by an AI-generated voice.

Thus, with minor modifications of currently-available technology, you could put a laptop on the Supreme Court podium and it could deliver an oral argument exactly like this. Voice-transcribing software could transcribe the questions; the text could be inputted into AI; and the AI’s outputs could be inputted into an AI voice generator. No preparation time would be needed; if the AI was completely unfamiliar with the case at 9:59 AM, it could deliver this oral argument at 10:00 AM.

It wouldn’t have to be a laptop. You could use any type of robot with sufficiently good speakers. I am imagining a kind of BattleBots scenario, with a Parkour robot representing petitioner sprinting up to the podium, leaping heroically over nearby chairs, and respondent countering with Spot the Robot Dog.

The results

You should listen to the argument all the way through. It’s excellent.

Was it better than me? Well, you all can decide that. But:

• All of Claude’s answers were clear, coherent, and directly responsive.

• Claude skillfully batted hostile questions away and used them to bolster its core themes.

• Claude never stumbled or got tongue-twisted.

• Claude knew the record cold.

• Claude gave several unusually clever answers, making arguments I didn’t think of.

The voice generation software is amazing. You can tell it’s AI because it sounds a little too perfect, but I find it a pleasure to listen to. I used the alpha version of ElevenLabs v3 (i.e., an early version that’s still under development), so forthcoming versions will sound even better.

There were a few places where Claude didn’t understand what a question was getting at. But that happens to humans too. Objectively, this is an outstanding oral argument.

But why?

In my opinion, AI will soon surpass even the best human oral advocates, if it hasn’t already.

It shouldn’t be a surprise that AI does so well at oral argument. AI has many advantages.

First, AI thinks quickly. Oral argument is challenging because lawyers have to think quickly. After a judge asks a question, the lawyer has no more than two or three seconds to come up with an answer. If there’s anything AI is good at, it’s thinking quickly.

Suppose I gave a human and an AI the briefs in a completely unfamiliar case and then tasked them both with delivering a Supreme Court oral argument within seconds. With very few exceptions, I think the human lawyer would win the silver. Most of us take more than a few seconds to master 100+ pages of dense legalese.

Now, suppose, at oral argument, a Justice were to ask a very weird and difficult legal question. For example, suppose, in the Williams argument, a Justice asked: “Give three ways in which the Twenty-First Amendment bolsters your argument.” This is a difficult question because the Twenty-First Amendment, which abolished Prohibition, has nothing to do with the case. Yet Claude was able to give an answer effortlessly:

First, the Amendment demonstrates a fundamental constitutional principle: when the Constitution allocates special authority to states over particular domains, that allocation must be respected. The Twenty-First Amendment grants states unique power over alcohol regulation that federal law cannot override. This shows that constitutional text creates clear boundaries between federal and state authority. Here, the Supremacy Clause creates an equally clear boundary—when federal civil rights are at stake, federal law governs, and states cannot interpose procedural barriers.

Second, the Twenty-First Amendment illustrates how constitutional provisions can override otherwise applicable federal rules. States can regulate alcohol in ways that would typically violate the Commerce Clause because the Constitution explicitly says so. Similarly, Section 1983 explicitly overrides state procedural rules for civil rights claims. The Amendment shows us that when constitutional or statutory text speaks clearly about federal-state boundaries, we must honor that clarity.

Third, the Amendment embodies a judgment about institutional competence—recognizing that states are best positioned to regulate certain matters within their borders. This same principle of institutional competence supports our position: Congress determined in 1871 that state institutions could not be trusted to vindicate federal rights, which is why Section 1983 guarantees immediate access to neutral courts rather than hostile state administrative processes.

This is clearly better than what a human lawyer could come up with on the spot. If you gave me six hours, I doubt I could do better.

(Claude was initially reluctant to answer the question because the Twenty-First Amendment is so irrelevant, but after I insisted and informed it that I was giving a direct order, it caved.)

This is a silly question, but it illustrates an important point. For questions requiring a lot of mental horsepower, AI is vastly better than humans on short time scales.

In the real world, of course, humans do not have to master 100 pages of legal briefing in seconds. And humans do not have to answer off-the-wall questions about the relevance of irrelevant constitutional provisions. But if AI massively outclasses humans in answering hard questions, wouldn’t one also expect AI to outclass humans in answering easier questions?

Computers are far more accurate and faster than humans at multiplying 100-digit numbers. This implies that computers are also more accurate and faster than humans at multiplying 2-digit numbers, even though humans can perform that task reasonably quickly. The same goes for oral argument questions. I suppose in principle it’s possible that humans have some subtle advantage that makes them better at answering cognitively less intensive questions even though they are vastly worse at answering cognitively more intensive questions. But what’s the theory there exactly?

Second, AI doesn’t suffer from human frailties.

• There are a tiny number of lawyers who have been touched by God and are consistently able to speak in full paragraphs that are legible on an oral argument transcript. The rest of us give garbled answers, make grammatical errors, and get lost mid-sentence. AI never does this. AI-driven oral argument transcripts are invariably clear as day.

• Humans get confused sometimes. You will often see a lawyer not understand an oral argument question, say “no” instead of “yes,” or make some other objectively clear error. This doesn’t happen with AI. OK, AI will hallucinate and give weird outputs sometimes, but it will never completely botch a question because it can’t think quickly enough.

• Humans get nervous, and nervousness can negatively affect performance. If AI claims it is nervous, it is lying.

• There are certain types of questions—such as “what page of the record does X appear”—for which computer have a decisive advantage.

No tradeoffs

AI is way smarter than humans and speaks in full sentences. Does Team Human have anything to recommend it?

Not really. Let me walk through, and debunk, a few possible arguments against the use of AI at oral argument.

Hallucinations. In any discussion of AI, it is compulsory on penalty of disbarment to mention hallucinations. So: Yes. LLMs are guilty as charged of sometimes hallucinating. They will make up facts, cases, and quotes. I am aware of this and this and this incident. Cutting-edge LLMs hallucinate less than older LLMs, and you can reduce the risk of hallucinations through safety measures like adding “do not hallucinate” to the query, but it still happens sometimes.

But this matters less at oral argument. AI is pretty good at accurately reporting information from documents loaded in its context window, especially if the documents aren’t that voluminous. And at oral argument, the advocate isn’t supposed to bring up cases and facts outside the written record. Occasionally a lawyer tries to bring up a case at oral argument that’s not in the briefs, and it typically goes badly—the lawyer has to apologize awkwardly and then read off a case citation while the judges look quizzically and mentally fault the lawyer for failing to cite the case in his brief. Instead, the advocate is supposed to offer sophisticated and insightful answers to questions regarding the already-submitted briefs—tasks that AI is particularly good at. The AI-driven Williams oral argument illustrates the point. There are no hallucinations because Claude never had a reason to hallucinate.

Humans are more authentic than machines. There’s an argument that goes like this: “Humans are better than AI because AI seems weird and so judges won’t take it seriously.” True, if judges choose not to credit the arguments of AI lawyers, then, tautologically, AI lawyers will be less effective than human lawyers. But let’s hope judges don’t do that.

Maybe one can offer a different flavor of this argument, something like: “human lawyers can form authentic connections with judges and AI lawyers cannot.” I agree that a judge is unlikely to form an authentic human connection with a robot dog. However, I doubt judges form authentic human connections at oral argument with humans either, at least not with any frequency. Human lawyers aren’t particularly authentic people. They’re up at the podium to make arguments on behalf of their clients, not to reveal the whispers of their souls.

Also, do we really want the outcome of legal cases being influenced by authentic human connections between judges and lawyers? Judicial decisions affect the interests of the clients, not the lawyers. And because judicial decisions set precedents that apply in future cases, they affect the interests of other people who aren’t before the court. Why does it make sense that those people’s interests would be affected by a lawyer and a judge sharing a special moment of luminous kinship at oral argument? It’s like, “The judge was entranced by the lawyer’s lilting voice, therefore everyone gets to have a brand new personal jurisdiction doctrine!” Wiping out authentic human connections is a feature, not a bug, of AI.

You can just do things

On March 26, 2025, a pro se litigant appearing before a New York intermediate appellate court (the famous First Department) attempted to present his argument via an AI avatar. The judges were displeased (YouTube video here). As a practice pointer, it is a bad idea for a litigant to show up at a hearing and attempt to deliver oral argument via an AI avatar without asking the panel first. Still, the litigant’s heart was in the right place.

I would like to advocate for courts permitting AI oral arguments. Lawyers would have the option of letting the AI doing the entire argument and sitting on the side; alternatively, they’d have the option of doing the argument themselves but have AI whisper in their ear.

Courts shouldn’t be grudging about allowing AI arguments. They should treat AI-made arguments exactly like they would treat human-made argument. The better argument will win, regardless of who makes it.

If we’re nervous about AI going haywire, judges could require a human to attend the argument along with the computer, like how some jurisdictions require out-of-state lawyers to appear with local counsel. The lawyer wouldn’t just have to set the laptop on the platform and sit down looking bored. I am envisioning human lawyers striding up to the podium inside Avatar-style AMP suits.

Why?

Better advocacy. Watching that video of the First Department argument, do you really have any doubt that the charismatic and articulate AI avatar would have done a better job than the hapless pro se litigant? Low bar, fine. But it’s clear that in many, perhaps most cases, feeding the briefs into AI and having it spit out answers will produce better lawyering. This is especially true given that, at least in the early stages, the clients who will use AI are the clients who have the weakest lawyers or who don’t have lawyers at all.

A level playing field. If both parties use AI—which will inevitably start happening—the advocates will be of roughly equal quality. This will improve the odds that the court’s eventual decision will be driven by the correct view of the law as opposed to differential lawyer quality. Suppose that without AI, one lawyer’s advocacy would get a 7/10 and the other lawyer’s advocacy would get a 4/10, and with AI, both lawyers’ advocacy would get an 8/10. The use of AI would improve the quality of the resultant judicial decision, both because there’s better advocacy on both sides and because the judges won’t be distracted by one lawyer’s superiority over the other.

Autonomy. For better or for worse, our legal system prioritizes litigants’ autonomy in deciding how they will defend themselves. That is why, for example, courts tend to be very reluctant to disqualify litigants’ chosen lawyers. It is also why criminal defendants have a constitutional right to represent themselves. Even when a lawyer (or unrepresented criminal defendant) is doing horribly, judges rarely step in. Respecting litigants’ autonomy requires respecting their right to use AI.

Low downside risk. I have participated in many panel discussions on the topic of “does oral argument affect the outcome of appeals?” Typical answers include “usually not, but occasionally yes”; “it rarely affects the outcome, but it sometimes affects the reasoning”; and “you can’t win your case at oral argument, but you can lose it” (which doesn’t make sense to me, it reminds me of “baseball is 90% pitching”). All of these conjectures are unfalsifiable, which is why lawyers keep doing the panel discussions. Still, the consensus is that oral argument is less important to the outcome than briefing. A lawyer who botches oral argument but has the more persuasive brief will usually win. This means that if we’re turtles in our shells, terrified of change, the downside risk of using AI at oral argument is low.

People are already doing this. If you show up in court with a laptop and an earpiece, the judge will probably notice. But hearings, particularly at the trial court level, frequently take place on Zoom. There is no way for the judge to know if the lawyer is reading AI’s answers off of his computer monitor. I guarantee you that ill-prepared and stressed-out lawyers are already using AI during remote hearings.

Is this sanctionable conduct? As long as the answers themselves are accurate and not hallucinatory, I’m not sure that it is. There’s no specific rule of professional ethics that bars an attorney from using AI during a hearing. A court could enact a local rule banning any use of AI during hearings, but few courts have such rules.

This will happen more often as time goes on. What percentage of college students don’t use AI to complete their assignments? I would suspect the percentage is hovering around 0%. Once those college students become lawyers, they are not going to stop.

***

How is this for a modest proposal? As a pilot program, a court could issue a local rule allowing pro se litigants—and no one else—to use AI, as they see fit, at oral argument. To guard against hallucinations, the court could require the litigant to use a prompt that would ask the AI not to cite any cases unless they appear in the briefs. Just give it a shot and see how it goes.

The Supreme Court’s decision has provoked controversy. Justice Alito, joined by Justice Thomas, vigorously dissented from both the April 18 and the May 16 orders. On May 20, Judge Ho wrote a concurrence similarly expressing disagreement with the Supreme Court’s May 16 decision. The Wall Street Journal published an editorial referring to the May 16 decision as “A Supreme Court Injustice to a District Judge.”

In this post, I will offer my take on this controversy. Executive summary:

1. A.A.R.P. raised novel and challenging procedural issues.

2. Those novel and challenging procedural issues were artificial. They arose solely because the government deployed a strategy of withholding information in order to hinder the Venezuelan detainees from vindicating their right to due process, leaving the plaintiffs’ lawyers and the court to fumble in the dark.

3. The Supreme Court appropriately declined to allow the government to benefit from this strategy.

Autism awareness

In 1798, in anticipation of war with France, Congress enacted the Alien Enemies Act (AEA). In its current form, the AEA provides in relevant part:

Whenever there is a declared war between the United States and any foreign nation or government, or any invasion or predatory incursion is perpetrated, attempted, or threatened against the territory of the United States by any foreign nation or government, and the President makes public proclamation of the event, all natives, citizens, denizens, or subjects of the hostile nation or government, being of the age of fourteen years and upward, who shall be within the United States and not actually naturalized, shall be liable to be apprehended, restrained, secured, and removed as alien enemies.

The AEA was invoked during the War of 1812, World War I, and World War II. Following World War II, President Truman continued to rely on the AEA as a basis to deport former Nazis. After that, the AEA lay dormant until March 14, 2025, when the President signed a Proclamation declaring that Tren de Aragua is a “hybrid criminal state that is perpetrating an invasion of and predatory incursion into the United States,” thus entitling the administration to deport Tren de Aragua members under the AEA.

And it did. Almost immediately after the Proclamation was issued, the administration began deporting Venezuelans to the Dickensian nightmare known as El Salvador’s CECOT prison. On the evening of March 15, a federal court issued a temporary restraining order blocking AEA deportations. But during the one-day period between the issuance of the proclamation and the temporary restraining order, the administration deported approximately 137 Venezuelans under the AEA. Those people are currently at CECOT, possibly for the rest of their lives.

Whether the administration may lawfully deport Venezuelans under the AEA is, to put it mildly, a subject of dispute. It is far from clear that Tren de Aragua qualifies as a “foreign nation or government” engaging in an “invasion or predatory incursion” that could trigger the AEA.

Even if the President has the authority to deport Tren de Aragua members, it is far from clear that the people who have been deported so far, or who are being threatened with deportation, are, in fact, Tren de Aragua members. The administration says they are Tren de Aragua members, but that doesn’t mean it’s true. Of note, many of the deportees entered the United States legally and did not have criminal records.

It seems that the administration designated many of these men as Tren de Aragua members based on their tattoos. But published reports suggest that these designations are unreliable, with the administration treating an autism awareness tattoo, a Real Madrid tattoo, and tattoos saying “Mom and Dad” as definitive evidence of Tren de Aragua membership. Meanwhile, according to an expert declaration, Tren de Aragua doesn’t even use tattoos.

Ordinarily, when the government alleges that a person should be deported and the person denies the allegation, the person gets a hearing before a neutral judge—that is, due process. However, when the government started deporting Venezuelans on March 14, it did not give them any process. Instead, the government unilaterally declared them to be Tren de Aragua members, shackled them, flew them to El Salvador, and then took the position that any due process claim they might bring was moot because they were now in the custody of a foreign country.

On March 15, the ACLU filed a lawsuit in the District of Columbia, claiming that the Proclamation violated the Administrative Procedure Act. As noted above, the court quickly granted a temporary restraining order blocking AEA deportations. This precipitated several weeks of tense litigation which quickly reached the Supreme Court.

Skipping over many twists and turns, on April 7, the Supreme Court ruled in Trump v. J.G.G., by a 5-4 vote, that the temporary restraining order should be vacated. According to the Supreme Court, the plaintiffs had chosen the wrong procedural vehicle to challenge the deportations: they should have brought habeas corpus actions rather than sue under the Administrative Procedure Act. This was important because Administrative Procedure Act claims can be brought in D.C., whereas habeas corpus actions must be brought in the district of confinement, which for many detainees would be in Texas. I was surprised by this ruling, but it doesn’t matter what I think, so habeas it is.

Importantly, however, the Supreme Court also said this:

AEA detainees must receive notice after the date of this order that they are subject to removal under the Act. The notice must be afforded within a reasonable time and in such a manner as will allow them to actually seek habeas relief in the proper venue before such removal occurs.

In other words, Venezuelan detainees facing deportation have a right to due process.

The Oracle

And so the government quickly devised a new notice protocol to comply with J.G.G.

I’m going to start by telling you what that new notice protocol was. But before I do, one point is crucial to understand. Much of this post will describe a dramatic series of events and filings on April 18 that culminated in a Supreme Court injunction. But the government didn’t tell the courts or the detainees’ lawyers about this new protocol until after the events of April 18. The government wrote the new protocol and immediately started implementing it without telling anyone what it was. As far as I can tell, the government first disclosed the new protocol in a federal court filing on April 23—well after the government attempted to start deporting people pursuant to the new protocol. Moreover, even on April 23, the government attempted to file this new protocol under seal, opposing its public disclosure on the theory that it would reveal “confidential investigative methods”; it only became public after the district court unsealed it over the government’s opposition. So you, the reader, will understand what the government was actually doing on April 18, but the detainees’ lawyers didn’t have access to this information at the time.

The government’s protocol was to give the detainees a document called “Form AEA-21B: Notice of Warrant of Apprehension and Removal Under the Alien Enemies Act.” Form AEA-21B advised the detainees that they had been determined to be Tren de Aragua members and would be deported under the AEA. It said that the detainee could place a phone call. Finally, it said that after the detainee was removed, he could not reenter the country without permission from the Secretary of Homeland Security.

According to the government’s declaration, after being served with Form AEA-21B, the detainee is given twelve hours to “indicate or express an intent to file a habeas petition.” If the detainee doesn’t “express any such intention,” then ICE may deport the detainee immediately. If the detainee “does express an intent to file a habeas petition,” then the detainee is given 24 hours to file the petition. If the detainee doesn’t file the petition within 24 hours, then ICE may deport the detainee immediately.

Stop right there.

It is absurd to suggest that this protocol complies with the Supreme Court’s directive that the detainees receive notice “within a reasonable time and in such a manner as will allow them to actually seek habeas relief in the proper venue before such removal occurs.” Among other problems:

• The detainee has no way of knowing that he could file a habeas petition challenging the deportation. Form AEA-21B makes no mention of habeas corpus or any other type of judicial review. It reads like a death warrant given to a condemned inmate 12 hours before execution, notifying the inmate that the time has come to begin reflecting on a life poorly lived.

• Even if the detainee is able to divine that filing a habeas petition might be an option, he would have no idea that “expressing an intent to file a habeas petition” would give the detainee a 24-hour reprieve. Is the detainee supposed to tap the shoulder of the ICE officer and say, “Officer, I just want to let you know that I am planning to file a petition for a writ of habeas corpus in the United States District Court for the Northern District of Texas”? Detainees aren’t expected to, and usually shouldn’t, tell ICE officers of their litigation plans. Note that Form AEA-21B ominously states: “Any statement you make now or while you are in custody may be used against you in any administrative or criminal proceeding.”

• Under the protocol, detainees are permitted a phone call. In principle, the detainee could call a lawyer. But:

  • Many of the detainees didn’t have lawyers. And that’s not because no one would represent them! The ACLU offered to represent these people. But the government refused to give the ACLU their names and refused to let the ACLU into the facility to talk to them unless the ACLU already knew their names in advance. Not surprisingly, precisely zero detainees who did not already have lawyers filed a habeas petition. The government’s protocol rested on the presumption that a detainee who didn’t file within 24 hours had voluntarily chosen not to bring a habeas petition, but … really? Given the choice between (1) being defended in court by a crusading ACLU lawyer and (2) being deported to Dante’s Inferno, do you really think that 100% of these unrepresented detainees voluntarily chose Option 2?

  • Even if the detainee had a lawyer, the lawyer might not pick up. Rumor has it that lawyers do not always pick up the phone immediately and sometimes even take more than 12 hours to return a voice message. Of note, even when detainees had immigration lawyers, and even when the government knew who those lawyers were (because, for instance, they had filed appearances in immigration cases), the government did not provide these notices to the detainees’ lawyers.

  • Even if the lawyer did pick up the phone, the lawyer would have no reason to advise his client to inform ICE, “I intend to file a habeas petition.” Remember, the government’s protocol—in which expressing this intention buys the detainee an extra 24 hours—was a closely-held secret.

  • Even if the lawyer picked up the phone and the detainee successfully expressed an “intent to file a habeas petition,” the lawyer would not know that a petition had to be filed within 24 hours. Apparently some of the detainees were told that they would be deported “tonight or tomorrow,” but nothing in the protocol required ICE to disclose this information to the detainee. And no reasonable lawyer would think the habeas petition had to be filed within 24 hours, given the Supreme Court’s admonition that notice had to be “reasonable.”

To sum up, in order to file a habeas petition, the detainee would not only have to already have a lawyer, but the lawyer would have to be The Oracle from The Matrix.

The government surely knew that its protocol didn’t comply with the Supreme Court’s decision and would never stand up in court. Why, then, did the government nonetheless barrel forward?

Because, from the government’s perspective, if the detainees could be deported before a court weighed in, there would be nothing the court could do about it. Recall the celebrated case of Kilmar Abrego Garcia, who was deported to El Salvador in violation of a court order. After the Supreme Court held that the government was required to “‘facilitate’ Abrego Garcia’s release from custody in El Salvador and to ensure that his case is handled as it would have been had he not been improperly sent to El Salvador,” the White House responded by a posting a tweet/X saying that he’s “never coming back.” From the government’s perspective, it was irrelevant whether its protocol violated Due Process; once the detainee was deported, his case was over.

Reasonable, but wrong

Shortly after the Supreme Court issued its J.G.G. decision, two detainees, represented by the ACLU, dutifully brought habeas petitions in Texas. They also sought to certify a class of AEA detainees, which would allow the ACLU to seek relief not only on behalf of its individual clients, but also on behalf of other detainees who faced deportation under the AEA. Finally, the ACLU sought a temporary restraining order on behalf of the class that would have temporarily prevented AEA detainees from being deported.

The government promised that it wouldn’t immediately deport the two named plaintiffs, but didn’t make any similar promises with respect to other detainees. Nevertheless, on the afternoon of April 17, the district court denied a motion for a classwide temporary restraining order, holding that it was unnecessary because there wasn’t an imminent threat that the detainees would be deported. The court said: “the Supreme Court’s opinion in J.G.G., along with the government’s general representations about the procedures necessary in these cases, strongly suggest that the putative class is also not facing such an imminent threat.”

This was a reasonable position for the district court to take. Immediately deporting the detainees would violate J.G.G.’s directive that the detainees get a reasonable amount of notice. There was no reason to believe that the government would violate a Supreme Court decision from ten days earlier.

But, it turns out, this reasonable position turned out to be wrong.

Shortly after the district court’s denial of the TRO on April 17, the plaintiffs’ lawyers learned that the detainees were getting notices that they would be deported under the AEA. The detainees were told that the deportations would happen “tonight or tomorrow” (i.e., April 17 or April 18). This information wasn’t coming from the government. Instead, some detainees called their lawyers from the detention facility and described the notices.

Alarmed, the ACLU contacted the government and asked whether the government intended to deport detainees other than the named plaintiffs. At 8:41 pm Central on April 17, the government lawyer sent an email refusing to answer the question.

The plaintiffs had no choice but to go back to the district court. Shortly after midnight on April 18, the plaintiffs filed an emergency motion for a temporary restraining order, asking the court to stop the deportations.

But the plaintiffs had a problem. The district court had a standing order saying that if a party filed an emergency motion, the opposing party had 24 hours to respond. Yet it appeared that the government was going to deport the detainees before the 24-hour period expired. The plaintiffs therefore asked the district court to rule immediately, observing on page 1 of their motion that “these removals could therefore occur before this matter may be heard and before the government’s response within 24 hours.”

And then the plaintiffs waited.

The hours ticked by, and the court didn’t respond to the plaintiffs’ filing. It did not indicate, one way or another, whether it would rule in less than 24 hours. It simply said nothing.

The next morning, the plaintiffs asked the government whether it would sign on to a motion for a status conference. The government said a status conference was unnecessary.

Morning yielded to afternoon, and still the plaintiffs did not know whether the court was planning to wait the full 24 hours. Nor did they know whether the government was going to deport detainees before the 24-hour period expired.

So the plaintiffs moved for a status conference at 12:48 pm, telling the court that if it didn’t rule by 1:30 pm, they would file an appeal. The court didn’t respond.

The plaintiffs gave the district court a little longer, then appealed to the Fifth Circuit at 3:02 pm and also sought relief from the Supreme Court. The Fifth Circuit held that it lacked jurisdiction because the plaintiffs didn’t give the district court enough time to rule.

At 11:52 pm Central on April 18, the Supreme Court granted a temporary injunction barring the government from deporting the detainees.

What would have happened if the plaintiffs hadn’t appealed on the afternoon of April 18? We don’t know for sure, because the government has never told us, but we can make an educated guess. It was subsequently revealed that in the afternoon of April 18, about 70 detainees were loaded onto a bus headed for the airport. The bus actually reached the airport but then turned around, perhaps because the government was nervous about deporting the detainees while they had a motion pending in the Supreme Court.

There was no reason to bring these detainees to the airport unless the government was planning to deport them. They were not going to sit in the airport lounge enjoying the free WiFi.

So here’s what would have happened if the plaintiffs hadn’t appealed. The government would have deported the detainees and then submitted a response brief in the district court after midnight on April 19 saying: too bad, we’ve deported 70 class members and they are forever barred from obtaining relief.

On the evening of April 18, after the plaintiffs went to the Supreme Court, a government lawyer represented to a judge in D.C. that there wouldn’t be any deportations that evening, but the government reserved the right to deport detainees on April 19. If the Supreme Court hadn’t granted its late-night injunction, the deportations may well have occurred the next day.

Was it really an emergency?

After the Supreme Court issued its late-night injunction on April 18, it stewed on the case until May 16, when it issued its decision in A.A.R.P. v. Trump extending the injunction and directing the Fifth Circuit to adjudicate the plaintiffs’ claims. Let’s take a look at that May 16 decision.

Like many controversial cases, A.A.R.P. v. Trump is larded with esoteric technical issues. The first question in A.A.R.P. was whether the plaintiffs’ appeal at 3:02 PM on April 18 was proper. If the appeal was premature, that would imply that both the Fifth Circuit and the Supreme Court lacked jurisdiction, which would mean the Supreme Court would have no choice but to dissolve the injunction. This might seem like an obscure issue, but the whole case turned on it.

This jurisdictional controversy arose because plaintiffs ordinarily aren’t permitted to file an appeal until a preliminary injunction motion is denied. But the district court never actually denied the plaintiffs’ motion. The court just sat on it until the plaintiffs ran out of patience and filed their appeal.

To get around this problem, the plaintiffs argued that the district court “constructively denied” the motion by failing to rule on it. Thus, the question before the Supreme Court was whether this “constructive denial” theory was legitimate. Or, translated into English: was there a real emergency that justified the plaintiffs’ appeal on April 18, or did the plaintiffs jump the gun?

This was a tough question because it was unclear, on the afternoon of April 18, what the government was going to do. That was the fault of the government, which was withholding information from the court.

Suppose that the government had announced: “unless there’s a court order that stops us, we will start deporting detainees at 5:00 PM on April 18.” In that case, it would have been unreasonable for the district court to give the government until the morning of April 19 to respond while doing nothing to preserve the status quo. The district court’s refusal to rule on the plaintiffs’ motion by 5:00 PM would have had the same practical effect as denying the motion. As such, an appeal would be warranted.

Now suppose the government had announced: “we promise to hold off on deporting the Venezuelans for three days to let the judicial process run its course.” In that scenario, the district court’s decision to wait until April 19 would have been reasonable, and the plaintiffs would have had no ground for a hair-trigger appeal on April 18.

What made the issue difficult was that the government was keeping its intentions a secret. It refused to tell the plaintiffs what it was going to do, even when they asked.

Now you might say that the government was under no obligation to disclose that information, and that it wasn’t the government’s job to help the plaintiffs prosecute their case. But actually, I think it was the government’s job for a few reasons.

1. Lawyers—particularly government lawyers—are officers of the court. They are charged not only with obeying court orders, but also with upholding the integrity of the judicial process. The integrity of the judicial process depends on judges making decisions based on accurate information. If the court is struggling to figure out whether there’s a real emergency, the government should tell the court whether there’s a real emergency.

2. Merely one day earlier, the district court had said: “the Supreme Court’s opinion in J.G.G., along with the government’s general representations about the procedures necessary in these cases, strongly suggest that the putative class is also not facing such an imminent threat.” The court was clearly under a misimpression regarding the government’s intentions—based in part on the government’s own prior representations. The government should have corrected that misimpression rather than viewing it as something to exploit.

3. In referring to “the Supreme Court’s opinion in J.G.G.,” the district court was telegraphing its view that deporting the detainees would violate their due process right as recognized in J.G.G. The district court evidently couldn’t bring itself to believe that the government would deport the detainees so quickly. I don’t think it’s an appropriate response for the government to say, “the district court thinks deportations are unconstitutional and is trusting us not to violate the Constitution, so let’s start the deportations!”

4. The district court was acting respectfully towards the government. Rather than issuing a hair-trigger TRO, it was giving the government 24 hours to respond and was trusting the government not to undermine the court’s jurisdiction. If the government was going to deport the detainees, it should have said something.

5. Finally, and most importantly, the government’s goal was to deport these detainees before they ever had their day in court and then argue their case was moot. In other words, the government was effectively saying: “we’ll deprive the detainees of the due process right that the Supreme Court has recognized—and we’ll do so for the specific purpose of permanently preventing them from ever getting the process the Supreme Court held was due.” I don’t think this tactic is consistent with upholding the integrity of the justice system.

The government should have been forthright with the court, but it wasn’t. This put the court in the difficult position of having to figure out whether there was a genuine emergency that warranted an immediate response.

I have sympathy for how the district court handled the case. My guess is that the district court didn’t seriously believe that the government was going to immediately start deporting detainees before the district court had the opportunity to rule. From the district court’s perspective, the plaintiffs’ lawyers were exaggerating the extent of the emergency, and the court wasn’t going to be complicit in that exaggeration by issuing a knee-jerk TRO without fully understanding the issues and giving the government the opportunity to respond.

But it turns out—based on information that came out later—that the plaintiffs weren’t exaggerating. During the 24-hour stretch, the government was really, seriously, bringing the detainees to the airport in anticipation of deporting them.

Even based on the information available on the afternoon of April 18, I agree with the Supreme Court’s decision that the plaintiffs’ motion was constructively denied, warranting an appeal. In polarizing cases, one craves neutral principles. From my perspective, the Supreme Court appropriately applied the ubiquitous neutral principle that when a party refuses to disclose information in its possession, the court should assume that fact against the party. The government was steadfastly refusing to disclose when it would start deporting the detainees, giving rise to the inference that it was planning to deport them immediately—which turned out to be correct.

Also, this wasn’t the first time that the government had been extremely aggressive in seeking to deport detainees before a court could rule. In the initial frenzied stages of the litigation in D.C. a month earlier, the government did everything it could to deport detainees before the district court could rule on the plaintiffs’ motion for a temporary restraining order. Indeed, what happened was quite remarkable. The government got planes in the air before the D.C. hearing. The D.C. judge issued an oral order directing the government to turn the planes around. The government ignored that order (without telling the court) because it wasn’t written down. Then the D.C. judge issued a written injunction intended to embody the oral injunction. The government still didn’t turn the planes around because the written injunction didn’t use the exact words of the oral order. This episode led the D.C. judge to suggest that the government might have committed criminal contempt. The fate of those contempt proceedings remains unclear. But the point is, in light of the government’s prior actions, there was every reason to believe that the government would deport the detainees before the Texas court could rule.

I am not persuaded by Justice Alito’s dissent. The dissent first faults the plaintiffs’ lawyers for failing to inform the court, in their late-night motion for a TRO, that they wouldn’t wait the full 24 hours before appealing. He says the lawyers “were fully aware that the District Court intended to give the Government 24 hours to file a response” but “said nothing about a plan to appeal if the District Court elected to wait for that response.”

This is an uncharitable reading of the plaintiffs’ brief. On page 1 of the plaintiffs’ brief, the plaintiffs acknowledged the district court’s standing order giving the government 24 hours to file a response, but said: “these removals could therefore occur before this matter may be heard and before the government’s response within 24 hours.” And then, on the last page of their brief, the plaintiffs said: “Given that individuals are now in imminent danger of removal, with notice that appears to be less than 24 hours, Petitioners respectfully request that the Court provisionally certify a class and grant a class wide TRO so that it has time to consider these important issues.” The plaintiffs were unambiguously asking for an injunction in less than 24 hours. It should not have come as a surprise that if the district court wouldn’t grant that request, the plaintiffs would seek relief from a higher court. I don’t think it’s necessary, or even appropriate, for the plaintiffs to say, “not only do we want relief in less than 24 hours, but we’ll appeal if you don’t give it to us!”

The plaintiffs ultimately did go back to the district court at 12:47 pm and requested a status conference by 1:30 pm. Hearing nothing, they appealed at 3:02 pm. The dissent faults them for that too: “delivering such an ultimatum to a district court judge … represented a very stark departure from what is usually regarded as acceptable practice.” Indeed it does. Something else also is a stark departure from ordinary practice: intentionally exploiting a district court’s misimpression while withholding information from the court for the purpose of permanently stripping the detainees of the very due process right that the Supreme Court had recognized 11 days earlier.

Also, the plaintiffs weren’t demanding that the district court rule by 1:30 pm. All they were asking for was a status conference. They would have asked the government: are you planning to deport detainees this afternoon, or not? If the government stonewalled, they would have asked the court: are you planning to rule this afternoon, or not? If the court said “no,” that was all the information the plaintiffs needed. The plaintiffs could then have appealed and argued that the district court denied the relief they were asking for: an injunction on April 18. It would have been a five minute conversation with the judge.

The dissent also says the plaintiffs had insufficient evidence that the detainees were going to be deported imminently. The plaintiffs submitted a lawyer’s declaration that an ICE official told a detainee he would be deported “tonight or tomorrow.” The dissent refers to this as “a double-hearsay statement that cannot be traced back to any specific government official.” Is the dissent suggesting that the plaintiffs should have submitted a declaration from someone at ICE?

The reality is that information concerning the timing of deportations was in the government’s possession and the government was withholding it from the court in order to facilitate its violation of a Supreme Court decision. In the face of an opposing litigant that was hindering their access to information, the plaintiffs’ lawyers were submitting the best evidence available under extreme time pressure. They managed to provide a “double-hearsay statement” that turned out to be completely correct: We now know that the government’s protocol did, indeed, provide that the detainees could be deported within 12 hours. I do not understand how the plaintiffs’ counsel can be faulted here.

Jurassic Park without the dinosaurs

Suppose one were to edit Jurassic Park by stripping out all the scenes with dinosaurs and muting all dialogue that mentioned dinosaurs. The movie would become extremely confusing. Why are these people so scared? Why do characters keep disappearing? The fact that there are dinosaurs on Isla Nublar is a central plot point.

Judge Ho’s concurrence on remand disagrees with the Supreme Court’s decision to exercise jurisdiction. In his view, the plaintiffs’ lawyers were unreasonable in demanding that the district court act so quickly on April 18. However, the concurrence makes no mention of the plot point that the detainees were about to be deported, which is as central to this case as dinosaurs are to Jurassic Park.

The concurrence excoriates the plaintiffs’ lawyers for requesting a ruling by 1:30 pm, which the concurrence characterizes as a “patently unreasonable timetable.” Perhaps so, were it not for the fact that the lawyers set this timetable because the detainees were about to be deported.

The concurrence says this:

Notably, the Justices themselves have expressed concerns about making decisions under far more forgiving time constraints than those demanded here. Recall the emergency relief sought in Does 1-3 v. Mills, 142 S. Ct. 17 (2021). Members of the Court expressed concern about the “use [of] the emergency docket to force the Court” to “grant . . . extraordinary relief” “on a short fuse without benefit of full briefing.” Id. at 18 (Barrett, J., concurring in the denial of application for injunctive relief). The amount of time considered too short in Does 1-3 was nine days. Compared to 42 minutes, however, nine days is a lifetime to decide a motion. So the district court reasonably assumed that the principle invoked in Does 1-3 to justify denying relief to law-abiding citizens concerned about their religious liberties in the COVID-19 era would likewise justify denying relief to illegal alien members of a foreign terrorist organization.

In Does 1-3 v. Mills, the plaintiffs were healthcare workers who were about to lose their jobs for refusing to get COVID vaccines. They sued, claiming a violation of their right to free exercise of religion. After both the district court and the court of appeals rejected their claim, they sought an emergency injunction from the Supreme Court. The Supreme Court denied relief. Concurring in that denial, Justice Barrett noted that in deciding whether to grant injunctive relief, the Court could consider the likelihood of certiorari being granted. In the portion of her concurrence quoted by Judge Ho, Justice Barrett stated: “Were the standard otherwise, applicants could use the emergency docket to force the Court to give a merits preview in cases that it would be unlikely to take—and to do so on a short fuse without benefit of full briefing and oral argument.”

The riddle of why the healthcare workers were treated differently from the detainees has a solution: the detainees were about to be deported and lose their rights forever. By contrast, if the healthcare workers lost their jobs, they could continue prosecuting their case. If they prevailed, they could be reinstated.

Also, in Does 1-3, the plaintiffs had the opportunity to litigate their cause in both a district court and an appellate court. Justice Barrett’s statement came in the context of observing that discretionary review in the Supreme Court was unlikely. By contrast, the detainees were going to be deported within hours without having gotten due process in any court, despite a Supreme Court decision from 11 days earlier saying they had a right to due process.

The concurrence blasts the plaintiffs’ lawyers for filing their motion shortly after midnight, observing that “this is a District Court—not a Denny’s.” One would have no idea, reading the concurrence, that the plaintiffs filed their motion shortly after midnight because they had just found out the detainees might be deported the next day. (Also, by statute, a district court is in fact a Denny’s: “All courts of the United States shall be deemed always open for the purpose of … making motions and orders.”)

The concurrence offers this observation:

If this is going to become the norm, then we should say so: District judges are hereby expected to be available 24 hours a day—and the Judicial Conference of the United States and the Administrative Office of the U.S. Courts should secure from Congress the resources and staffing necessary to ensure 24-hour operations in every district court across the country.

If this is not to become the norm, then we should admit that this is special treatment being afforded to certain favored litigants like members of Tren de Aragua—and we should stop pretending that Lady Justice is blindfolded.

This passage suggests the Supreme Court ruled that a quick ruling was needed because the Court is biased in favor of Venezuelan gang members. Might there have been some other reason in this case that the judiciary had to move more quickly than usual?

Justice!

Having concluded that the Fifth Circuit had jurisdiction over the plaintiffs’ appeal, the Supreme Court remanded the case to the Fifth Circuit to decide that appeal. But the Supreme Court also granted a temporary injunction to the entire plaintiff class that will stay in place until the Supreme Court finally decides the case.

Whether such an injunction may be issued raises a number of difficult issues:

1. Whether class actions can be certified in habeas cases at all.

2. Whether the named plaintiffs were typical of the class even though the government was promising not to deport the named plaintiffs but making no similar promises with respect to the rest of the class.

3. Whether a district court can grant temporary relief to a putative class before the class is certified.

As I see it, on all three issues, the Court’s decision rests on solid legal ground. Lower courts have certified classes in habeas cases for decades. There’s plenty of case law saying that the government can’t moot a case by promising to give special treatment to the named plaintiff. As to the third issue, the Court cites a treatise which in turn cites an impressively long list of cases saying that such temporary relief is possible. Temporary relief for a putative class is hard to wrap my head around, but lots of courts seem to have successfully wrapped their heads around it.

Still, none of these issues is clear-cut. But the important point for me is that all three issues arose solely as the result of the government engaging in tactical maneuvers as a means of violating the detainees’ due process rights.

1. There should have been no need for a class action in this case. There were lawyers available to represent every detainee. A class action was needed only because the government refused to identify the detainees, refused to allow lawyers access to the detainees unless the lawyers already knew their identities, refused to tell the detainees that there were lawyers begging to represent them, and made it impossible for the detainees to file lawsuits on their own.

2. The typicality issue arose solely because the government unilaterally said it wouldn’t deport detainees who filed habeas petitions, while making no similar promise for other detainees. There was no national security rationale for this move. It’s not as though the detainees who filed habeas petitions were less dangerous than those who didn’t. This move was designed to make it harder to certify a class, and hence easier to deport the detainees who lacked lawyers.

3. As for the question of whether a court can grant provisional relief to a class before the class is certified: The government claimed that the district court was required to go through a series of procedures before certifying the class, and until it went through those procedures, no classwide relief was warranted. But the government didn’t actually want to use those procedures. It wanted to invoke the need for those procedures as a basis to delay classwide relief, and then deport the class members before those procedures could ever occur.

The Supreme Court didn’t allow the government to get away with this.

I usually wince when people say that a Supreme Court decision delivered “justice.” In almost all Supreme Court cases, both parties have a claim to the mantle of justice, and the Supreme Court’s decision reflects nothing more than the fact that the winning party had stronger legal arguments.

Almost all, but not all. From time to time, the Supreme Court finds a way to deliver justice. In A.A.R.P., the Supreme Court delivered justice.

On March 25, 2025, President Trump issued an Executive Order entitled “Addressing Risks from Jenner & Block,” the firm where I am a partner. Among other things, the Executive Order “limit[s] official access from Federal Government buildings to employees of Jenner,” “limit[s] Government employees acting in their official capacity from engaging with Jenner employees,” and “require[s] Government contractors to disclose any business they do with Jenner and whether that business is related to the subject of the Government contract.” The President has issued similar Executive Orders against other law firms as well, with additional Executive Orders reportedly waiting in the wings.

On March 28, 2025, Jenner, represented by Cooley, sued the government in the U.S. District Court for the District of Columbia. That same day, the District Court granted Jenner’s motion for a temporary restraining order. As such, the bulk of the Executive Order is currently not being enforced. In total, three federal judges have now found that these Executive Orders likely violate the Constitution.

Our firm’s official website regarding this litigation is here. We say: “For more than 100 years, Jenner has stood firm and tirelessly advocated for our clients against all adversaries, including against unlawful government action. We once again go to court to do just that. To do otherwise would mean compromising our ability to zealously advocate for all of our clients and capitulating to unconstitutional government coercion, which is simply not in our DNA.” It gives me goosebumps when I read that because it’s absolutely true. I’m so proud that our firm—leadership, partners, associates, and staff—is sticking together and doing the right thing. I can’t tell you how proud I am to be affiliated with the firm.

While two other law firms have challenged similar Executive Orders, other law firms have followed a different path. The President recently withdrew an Executive Order against a law firm after it agreed to, among other things, dedicate $40 million of pro bono time to what the President referred to as “mutually agreed projects” that “support the Administration’s initiatives.” Since then, in an apparent effort to avoid similar Executive Orders, three other law firms have preemptively agreed to commit $100 million to pro bono initiatives supported by the government.

As more firms face the prospect of Executive Orders, more of my colleagues in the bar will face the same choice: sue or settle? In this post, I will make the case that they should sue.

The Executive Orders are unconstitutional. They violate the First Amendment right of lawyers and their clients to speak, petition, and associate. They are also designed to discourage lawyers from representing unpopular clients—even clients with meritorious cases—and in so doing, they profoundly distort the judicial system. If anyone should be standing on principle and attempting to vindicate the rule of law, it is our nation’s lawyers.

But the reasons for suing go beyond that. To zealously represent their clients, law firms must remain genuinely independent from government. A law firm that settles with the government is no longer independent from government—particularly where, as here, the settlements give the government de facto veto power over what cases the law firm chooses to take on.

At core, these settlements reflect an attitude of deep cynicism. Cynicism towards the law firm’s own clients—that they would prefer a law firm that is beholden to the government than one that maintains its ability to provide independent, uncompromised advice. And cynicism towards our justice system—that it is so toothless that the law firm would be better off capitulating than suing even if it prevails. We should be optimistic both about the clients we exist to represent and the justice system we are sworn to uphold.

The Executive Order is unconstitutional and undermines the rule of law

The Executive Orders violate the First Amendment. There have been several Executive Orders so far that all suffer from similar constitutional infirmities; I’ll focus on the one against Jenner.

It is well established that “the First Amendment prohibits government officials from retaliating against individuals for engaging in protected speech.” The Executive Order retaliates against Jenner because Jenner engaged in protected speech. According to the Executive Order, Jenner must be punished because it “engage[d] in obvious partisan representations to achieve political ends,” such as representations on behalf of immigrants and transgender persons. Legal advocacy is a form of speech. Jenner is being punished for speaking.

The First Amendment’s Petition Clause separately protects the right to petition the government for redress of grievances, including the filing of lawsuits. Indeed, as the Supreme Court has emphasized, the right to petition is “one of the most precious of the liberties safeguarded by the Bill of Rights.” Retaliating against Jenner because it purportedly filed lawsuits to “achieve political ends” is anathema to the First Amendment.

The First Amendment also prohibits the government from discriminating on the basis of viewpoint. The Executive Order discriminates on the basis of viewpoint. Immigration and transgender rights are areas of intense social debate. The Executive Order openly targets Jenner because it represented clients who had viewpoints on these issues with which the current government disagrees.

The Executive Order also targets Jenner because of its previous affiliation with Andrew Weissmann. Mr. Weissmann, a former member of Robert Mueller’s team, wrote a book criticizing the President and has frequently criticized the President on television. The Executive Order states that Mr. Weissmann’s “overt demand that the Federal Government pursue a political agenda against me” shows his “dishonesty.” And it assails Jenner because Jenner stated years ago that it was “thrilled” to have Mr. Weissmann as a partner. According to the Executive Order, Jenner’s decision to associate with Mr. Weissmann is a “concerning indictment of Jenner’s values and priorities.” The First Amendment does not permit the government to retaliate against a law firm (or anyone else) based on its perceived “values and priorities”—particularly when the government is inferring those “values and priorities” from the law firm’s constitutionally protected decision to associate with a person whose political views differ from the President’s.

Not only does the Executive Order retaliate against Jenner for its prior exercise of its First Amendment rights, but the Executive Order’s retaliatory measures are themselves First Amendment violations. The Executive Order “limit[s] official access from Federal Government buildings to employees of Jenner” and “limit[s] Government employees acting in their official capacity from engaging with Jenner employees.” These restrictions—if ever enforced—would make it impossible for Jenner to exercise the constitutionally protected right to petition the government for redress of grievances.

The Executive Order further “require[s] Government contractors to disclose any business they do with Jenner and whether that business is related to the subject of the Government contract.” The “fact sheet” accompanying the Executive Order sets forth the purpose of this requirement: “To ensure taxpayer dollars no longer go to contractors whose earnings subsidize activities not aligned with American interests.” In other words, the Executive Order—should it ever be implemented—compels contractors to disclose their association with their chosen counsel as a means of deterring contractors from retaining that counsel. That compelled disclosure, for the purpose of deterring contractors’ exercise of their right to freely associate, violates the First Amendment.

The Executive Order also violates the Due Process rights of both Jenner and its clients. I’ll start with Jenner. Due Process requires the government to provide notice and a hearing before it imposes punishment. That did not happen here. There was no notice. There was no hearing. Instead, the Executive Order was issued based on the President’s unilateral decision that Jenner had done something wrong.

The Due Process Clause also requires the government to give citizens advance notice of what actions may trigger punishment. This is the basis of the void-for-vagueness doctrine and many other doctrines. What advance notice did Jenner have? Indeed, even after this Executive Order, what advance notice do other law firms have? What types of representations or affiliations will trigger the next Executive Order? No one knows.

The Executive Order violates the Due Process rights of Jenner’s clients, too. The Due Process Clause guarantees our clients’ rights to obtain zealous legal representation. Without the temporary restraining order, it would not be possible for Jenner to represent its clients in dealings with the government when there is an Executive Order that bans Jenner lawyers from government buildings and that bans government officials from speaking to us. And forcing our clients to disclose the fact that they have retained us invades the attorney-client privilege. For similar reasons, with respect to criminal defendants, the Executive Order violates the Sixth Amendment right to counsel.

The government might say that our clients have the option of terminating their relationships with us and hiring other lawyers. But the Fifth and Sixth Amendments protect a client’s right to be represented by the lawyer of his choice, rather than being limited to lawyers who are restricted in their speech to messages that align with the government’s viewpoint.

Moreover, the purported opportunity for people with differing viewpoints to hire other lawyers is illusory. The Executive Order calls out our pro bono representations of disfavored clients for the express purpose of deterring other law firms from taking on similar representations. What law firm would take on a pro bono case—or any case—if the consequence of taking the case would be a potential executive order attacking the firm? The Executive Order’s purpose and effect is to strip unpopular litigants of any access to legal representation from law firms, even if they have meritorious claims.

The Executive Order also violates the Equal Protection component of the Due Process Clause. Abundant case law holds that the government may not single out particular entities for adverse treatment based on animus. That is what the Executive Order does. It declares that our firm has done pro bono work and affiliated with an ex-partner that the Administration doesn’t like, and therefore singles out us out for adverse treatment.

In addition to violating multiple provisions of the Bill of Rights, the Executive Order is inconsistent with the Constitution’s structure and the separation of powers. There is no constitutional basis for this action. Article II states that the President “shall take care that the laws be faithfully executed.” What “law” is being “faithfully executed” here? No constitutional provision, statute, or regulation authorizes the President’s action. To the contrary, the concept of a “law” that authorizes the President to unilaterally retaliate against his political opponents without due process is antithetical to the American tradition.

The separation-of-powers concerns are obvious. The judicial system cannot function unless lawyers can defend their clients zealously without fear of government retaliation. Every court has the authority to take appropriate action if a lawyer acts improperly in a particular case. Unilateral executive action retaliating against a law firm for representing clients—for the unapologetic purpose of deterring lawyers from making arguments in litigation with which the President disagrees—impedes the judicial process. Worse, the Executive Order deters suits on only one side of contentious social debates. If you want to represent clients whose views align with the government’s, you can litigate to your heart’s content. If you want to represent clients with opposing views, the government will come after you.

Moreover, there is no conception of the common good that is compatible with the Executive Order. The Executive Order bans nearly a thousand employees—including hundreds of attorneys—from one particular law firm from entering government buildings and interacting with government officials because of the government’s animus towards a former partner of that particular law firm. I participate in panel discussions all the time in which principled conservatives and progressives engage in reasoned discussion about a particular policy or legal issue. It is hard to imagine a panel discussion in which principled conservatives and progressives debate the policy merits of this Executive Order. It is alien to any conception of good policymaking under our constitutional system.

One final point. It should have been obvious to the government that the Executive Order would fail in court—not only is it plainly unconstitutional, but a federal judge had already enjoined the virtually identical Executive Order directed at Perkins Coie. Yet it was issued anyway. Why? Because the government has made the assessment that the law firm would be so cowed that it would refuse to litigate, even if it knew it would win. The implicit message of the Executive Order is: even if you win, the justice system cannot save you. At its core, the Executive Order is a bet on our justice system’s failure.

The settlements

Three law firms so far have sued and obtained quick temporary restraining orders barring the government from implementing the Executive Orders.

But other law firms pursued a different path. So far, four law firms have settled to avoid similar Executive Orders. One persuaded the President to withdraw an Executive Order by agreeing, among other things, to “dedicate the equivalent of $40 million in pro bono legal services” to “support the Administration’s initiatives” on “mutually agreed projects.” Since then, three other law firms have preemptively staved off Executive Orders by agreeing to undertake $100 million in pro bono work on causes supported by the President.

In light of this wave of preemptive settlements, other firms now face the same choice: settle or sue?

They should sue. Law firms should not accept settlements of this nature. Settling with the government not only reflects capitulation to unconstitutional government coercion, but also irrevocably compromises a law firm’s ability to defend its clients.

Law firms should not agree to dedicate $40 million or $100 million in pro bono legal services to “mutually agreed projects” to “support the Administration’s initiatives.” I am not saying this because I valorize impact litigation by large law firms, which I recognize does not inherently advance the public interest even if it may sometimes be worthwhile. Judge Jacobs’ thoughtful speech—now 17 years old, but still remarkably fresh—is the definitive statement of this position.

Instead, the problem is this: when a law firm agrees to dedicate $40 million or $100 million to “support the Administration’s initiatives” via “mutually agreed projects,” it cedes control over a piece of the law firm to the government. “Mutually agreed” means the government has to agree. But law firms are private organizations. The partners own the firm. The partners, not the government, decide what pro bono cases to take. I view it as unacceptable to have to turn over a piece of our firm to the government in exchange for lifting unconstitutional government restrictions.

Importantly, settlements between a law firm and the government cannot possibly be confined to millions of dollars in “mutually agreed” pro bono. Instead, all aspects of the firm’s legal work are irrevocably compromised.

Look at the President’s statement after one of the recent settlements. He announced: “The President continues to build an unrivaled network of lawyers.” He is correct. Any law firm that settles effectively joins the government’s “unrivaled network of lawyers.”

I am also struck by the President’s remarks yesterday:

Have you noticed that lots of law firms have been signing up with Trump? $100M, another $100M, for damages that they’ve done. But they give you $100M and then they announce that, ‘but we have done nothing wrong.’ And I agree, they’ve done nothing wrong, but what the hell, they give me a lot of money considering they’ve done nothing wrong. And we’ll use some of those people, some of those great firms, and they are great firms, too, they just had a bad moment. But we’re gonna use some of those firms to work with you [the coal industry] on your leasing and your other things, and they’ll do a great job. I think they’ll do a fantastic job.

As well as the White House Press Secretary’s statement today:

Big Law continues to bend the knee to President Trump because they know they were wrong, and he looks forward to putting their pro bono legal concessions toward implementing his America First agenda.

These statements accurately describe both the intent and the effect of the settlements: to conscript private law firms into doing the government’s work, or at least not to do any work that the government might perceive as inconsistent with its interests.

If you asked me to summarize the key to a free society while standing on one foot, I would say: “a robust private sector that is truly independent of the government.” Free enterprise cannot be free when the government demands that private organizations conduct “mutually agreed projects”—or else face government retribution. And this concern is heightened in the context of private law firms. All criminal litigation, and much civil litigation, pits private citizens against the government. The justice system cannot function unless those private citizens are zealously represented by counsel—which cannot happen unless that counsel is genuinely independent of the government.

By joining the government’s “unrivaled network of lawyers,” law firms forfeit their identity as independent organizations. Remember that the purported “deals” are not, in fact, deals. There is no binding contract. They are written in pencil. If the President lifts an Executive Order, he retains total discretion to reimpose a new Executive Order at any time. Similarly, when law firms preemptively settle, the President gives up nothing. He retains exactly the same legal authority to impose an Executive Order as he did previously.

As a result, if a law firm enters into this type of non-deal, I don’t understand how it can ever represent clients in any case that involves the government. Every time its lawyers file a brief, they will think: “On the one hand, I have to represent my client zealously, but on the other hand, I have to make sure not to undo the deal.”

The firm might pretend it is defending its clients zealously. It might file legal briefs making all manner of legal arguments. But what if a lawyer uncovered information that made the government look bad? There would be massive pressure on the lawyer to look the other way. Maybe the lawyer would try to offer some explanation to his client as to why it’s strategically best to set the issue aside. But there would always be a question mark in the client’s head: “is my lawyer saying this because he believes it, or is the lawyer saying this because he’s trying to preserve his law firm’s settlement with the government?”

Crucially, the Executive Orders against all of the affected firms explicitly call out the firms for the positions they took in litigation. As such, to preserve the deal, a firm that settles with the government has a strong incentive to pull punches in litigation. In every case in which the government has an interest, the firm has a conflict of interest.

Many lawyers at the settling law firms disagree with their firm leadership’s decision to settle. But agree or disagree, they are subject to the same conflict of interest. Partners at law firms have fiduciary duties not only to their clients, but also to each other. They also have strong personal interests in not doing anything that would undermine the firm’s delicate position.

When I say that the firm can’t represent the clients in cases involving the government, I’m not just talking about litigation against the government. I’m talking about any practice—regulatory, contract negotiation, anything—involving the government. Lawyers can and do have productive and respectful relationships with the government. Neither law firms nor businesses can function if they are constantly at odds with regulators. But in any interaction with the government, a private law firm can always maintain the implicit position that, if things go south, the law firm can zealously defend its client in court. A law firm in the pocket of the federal government can’t say that anymore.

And the conflict-of-interest concern goes beyond cases involving the government. The concern extends to any representation of any kind in which the government has an interest.

In any matter in which the government is interested—even a matter that does not directly involve the government—a law firm that settles will serve two masters. One is the law firm’s client. The other is the government. Suppose your law firm represents a client in a high-stakes lawsuit. On the other side of the case is a company that the government favors—either because it has explicitly come out in support of the government or because the government has politely persuaded it to undertake “mutually agreed projects.” How can you possibly represent your client zealously?

The Executive Order augurs the largest expansion of government in my lifetime. The size of government isn’t measured by the number of employees on the government’s payroll. It’s measured by the degree of influence the government has in the private economy. The Executive Order illustrates the government’s willingness to exercise authority, unconstitutionally if necessary, to acquire control over private organizations. To provide uncompromised advice, law firms must avoid being controlled.

No deal

I also have more fundamental objections to the settlements.

Look at some of the litigants who have stood their ground to vindicate First Amendment rights for the benefit of not only themselves, but everyone else. In West Virginia State Board of Education v. Barnette, 319 U.S. 624 (1943), the plaintiffs were 8-year-old Marie Barnette and 11-year-old Gathie Barnette, Jehovah’s Witness schoolchildren in West Virginia who refused to salute the flag based on their interpretation of the Book of Exodus. In Tinker v. Des Moines Independent Community School District, 393 U.S. 503 (1969), the plaintiffs were 15-year-old John Tinker, 13-year-old Mary Beth Tinker, and 16-year-old Christopher Eckhardt, schoolchildren in Iowa who were suspended when they wore black armbands to protest the Vietnam War. These people were children, from religious or ideological minorities, who I doubt were wealthy, who must have faced extreme pressure, hostility, discrimination, and fear as their cases winded their way through the courts. They prevailed and vindicated constitutional rights that we now take for granted. If those litigants can stand their ground, the nation’s largest law firms can too.

Challenging the Executive Order follows from the oath we took to support the Constitution when we joined the Bar. The Constitution does not mean much unless the rights it protects are vindicated. By bringing suit, law firms vindicate their own constitutional rights. They also vindicate the constitutional rights of their clients to retain counsel that will defend them zealously. And they set a precedent that will make it easier for other, similarly-affected firms to bring their own lawsuits, thus advancing constitutionalism across the legal profession.

Further, the Executive Order creates a chilling effect that will make it difficult for litigants with disfavored viewpoints to find counsel. Those litigants have the constitutional right to petition for redress of grievances like everyone else, and they cannot exercise that right unless they can find a lawyer willing to defend them. Bringing suit—and contributing to a legal culture in which law firms are willing to defend themselves—protects those litigants’ access to the courts.

I don’t think any lawyer, even lawyers at the firms that settle, think the Executive Orders are constitutional or that the settlements are virtuous. Instead, they view settling as a necessary evil designed to protect the firm’s fortunes. But that perspective reflects a profoundly cynical view of a law firm’s own clients. The theory behind settling is that what clients really want is a law firm that folds in the face of unconstitutional coercion, ostensibly to get into the government’s good graces, rather than a firm that stands up for its right to remain independent from the government. Law firms should have more faith in the people they represent than that.

Settling demonstrates cynicism not only about a firm’s clients, but also about the justice system. Three law firms have sued. All have obtained quick temporary restraining orders. Yet law firms are preemptively capitulating anyway. These settlements reflect the view that prevailing in litigation and obtaining a court order will lead to no real relief, and all that matters is public perception regardless of the outcome of judicial proceedings. I don’t think law firms, of all institutions, should be so dismissive of our justice system. And that cynical attitude by law firms will contribute to cynicism by the public. How can citizens not be cynical when they see these highly sophisticated and successful firms, ostensibly devoted to protecting the rule of law and being brave warriors for their clients, folding instantly?

By contrast, bringing suit reflects optimism in our justice system’s ability to right legal wrongs when litigants bring meritorious claims. If the Executive Order is a bet on the failure of our justice system, then bringing suit is a bet on its success. I will always bet on the success of the American justice system, and I hope other law firms do the same.

On January 20, 2025, President Trump issued an Executive Order declaring that a person is not a U.S. citizen “(1) when that person’s mother was unlawfully present in the United States and the person’s father was not a United States citizen or lawful permanent resident at the time of said person’s birth, or (2) when that person’s mother’s presence in the United States was lawful but temporary, and the person’s father was not a United States citizen or lawful permanent resident at the time of said person’s birth.” The lawsuit count on this Executive Order has hit ten, with four preliminary injunctions so far.

I am diffident about writing on this issue, both because there has already been abundant commentary and because my view aligns with the conventional wisdom: the Executive Order is inconsistent with the Fourteenth Amendment. Still, there is enough storage space on the Internet for one more think-piece.

In this post I’ll offer my take on why the Fourteenth Amendment confers birthright citizenship. I’ll also discuss what I see as the deeper philosophical issue at stake. Should judges decide legal questions by applying neutral principles and following the law where it leads, or by reasoning backward from what they perceive as the just conclusion? Under a neutral-principles approach, the pro-birthright-citizenship position is unassailable. Under a reasoning-backward approach, the anti-birthright-citizenship position is at least possible.

The Drive for Five

It seems weird that the child of an illegal immigrant or a non-resident would become a U.S. citizen merely by virtue of being born on U.S. soil. But weird as it may be, the legal case for that position is exceptionally strong.

There are five primary arguments for birthright citizenship:

1. It’s consistent with the Fourteenth Amendment’s text.

2. It’s consistent with British common law.

3. United States v. Wong Kim Ark, 169 U.S. 649 (1898), essentially answers this question.

4. We’ve been doing it this way for a long time.

5. Statements by members of Congress during the debates over the Fourteenth Amendment indicate that they understood it to confer birthright citizenship.

Let’s go through those one by one.

1. The text

The Fourteenth Amendment says this:

All persons born or naturalized in the United States, and subject to the jurisdiction thereof, are citizens of the United States and of the State wherein they reside.

Are the children of illegal immigrants and non-residents “subject to the jurisdiction” of the United States? In ordinary English, yes. The ordinary meaning of “jurisdiction” is “authority,” and the children of illegal immigrants and non-residents are most certainly subject to United States authority. If they commit a crime, they go to jail.

Might “jurisdiction” have some kind of special lawyerly meaning that differs from its ordinary meaning? As someone who traffics in lawyer-ese for a living, I can confirm that the answer is no. When we talk about a court having “personal jurisdiction” over a defendant, we mean that the court has authority over the defendant.

This is not a novel argument. If you have the misfortune of reading a Twitter/X debate on birthright citizenship, you will often find this point in the Most Relevant Reply. By contrast, the contrary textual argument—that “jurisdiction” means “at least one of your parents has a green card”—is not so easy to fit into 280 characters.

Simple arguments are good! It should be cause for celebration when people can understand what the Constitution means simply by reading it. There should be a strong presumption that the plain-English interpretation of a constitutional provision is correct.

2. The common law

The Supreme Court frequently interprets the Constitution by looking to the British common law, and the common law strongly supports the birthright-citizenship position. At common law, if a child was born on British soil, and was subject to the King’s authority, the child was a British citizen. This included the children of foreigners who were merely passing through.

What does the “King’s authority” mean? Well, if a foreign diplomat had a child on British soil, the child wouldn’t become a British citizen, because under principles of international law, the diplomat wasn’t subject to the King’s authority even while physically present in Britain. A version of the same principle survives today; it’s why diplomats don’t pay parking tickets. Likewise, if an enemy invaded Britain and was occupying the territory, the occupier’s children wouldn’t become British citizens because the territory would be under the authority of the occupier, not the King. So the formula was: Born in Britain + Subject to King’s Authority = British Citizen.

So we have a happy confluence of arguments. The plain-English interpretation of the constitutional text—“jurisdiction” equals “authority”—also is a perfect match to the common law, in which citizenship turned on authority. This is exceedingly strong evidence that this interpretation is correct.

3. The Wong Kim Ark case

In the late 19th century, people born in China were banned from becoming U.S. citizens. Wong Kim Ark was born on U.S. soil to two Chinese nationals. In United States v. Wong Kim Ark, 169 U.S. 649 (1898), the Supreme Court held that he was a natural-born U.S. citizen.

The Wong Kim Ark decision is long and drags on, but at core, the Court pretty much relies on the reasoning I just gave. The Court says that the Constitution “must be interpreted in the light of the common law, the principles and history of which were familiarly known to the framers of the constitution.” The Court then interprets the word “jurisdiction” in line with the common-law rule: “The fundamental principle of the common law with regard to English nationality was birth within the allegiance—also called ‘ligealty,’ ‘obedience,’ ‘faith,’ or ‘power’—of the king. … But the children, born within the realm, of foreign ambassadors, or the children of alien enemies, born during and within their hostile occupation of part of the king’s dominions, were not natural-born subjects, because not born within the allegiance, the obedience, or the power, or, as would be said at this day, within the jurisdiction, of the king.” The Court also notes that prior to the Fourteenth Amendment, “all white persons, at least, born within the sovereignty of the United States, whether children of citizens or of foreigners, excepting only children of ambassadors or public ministers of a foreign government, were native-born citizens of the United States.”

The Court distinguishes the 1884 case of Elk v. Wilkins, in which the Supreme Court held that Indians were not entitled to birthright citizenship. The Court explains that “the Indian tribes, being within the territorial limits of the United States, were not, strictly speaking, foreign states, but were alien nations, distinct political communities, the members of which owed immediate allegiance to their several tribes, and were not part of the people of the United States.”

Importantly, Elk is consistent with interpreting “jurisdiction” to mean “authority.” At the time of the Fourteenth Amendment, Indians actually weren’t subject to the United States’ authority, or at least not completely. As the Wong Kim Ark Court emphasizes, the Constitution, as originally enacted, excluded “Indians not taxed” from the count of “free Persons” for purposes of congressional apportionment. This reflected that the United States literally wasn’t exercising authority over Indians, at least for purposes of taxation. This was so both for philosophical reasons—the Indians had their own leaders and never agreed to pay taxes to European immigrants—and practical reasons—in the nineteenth century, American authorities couldn’t and didn’t exercise law-enforcement authority in many areas occupied by Indians.

But the Court is emphatic that Elk doesn’t extend to non-Indian foreigners:

The decision in Elk v. Wilkins concerned only members of the Indian tribes within the United States, and had no tendency to deny citizenship to children born in the United States of foreign parents of Caucasian, African, or Mongolian descent, not in the diplomatic service of a foreign country.

And so the Court reaches the following conclusion:

[A] child born in the United States, of parents of Chinese descent, who, at the time of his birth, are subjects of the emperor of China, but have a permanent domicile and residence in the United States, and are there carrying on business, and are not employed in any diplomatic or official capacity under the emperor of China, becomes at the time of his birth a citizen of the United States.

I promise you the Supreme Court did not reach this conclusion because of some kind of proto-wokeness. Wong Kim Ark was decided two years after Plessy v. Ferguson, which upheld racial segregation in Louisiana; every Justice in the Wong Kim Ark majority was also in the majority in Plessy v. Ferguson. Also, Wong Kim Ark was decided five years after Fong Yue Ting v. United States, a brutal decision upholding a federal statute under which any person of Chinese descent would be deported unless he could prove, based on the testimony of “at least one credible white witness,” that he entered the United States legally. (Reasoning: non-white witnesses have “loose notions … of the obligation of an oath.”) In short, the 1898 Supreme Court’s anti-woke credentials were absolutely sterling. The Supreme Court upheld birthright citizenship because the Constitution’s text and history left it little choice.

Opponents of birthright citizenship emphasize that Wong Kim Ark’s parents entered the United States legally and had a “permanent domicile and residence in the United States.” Thus, they contend, the decision doesn’t cover the children of illegal immigrants and non-residents.

But that’s not a basis to distinguish Wong Kim Ark. Wong Kim Ark is incredibly emphatic that the sole exceptions to the birthright-citizenship rule are Indians and the children of diplomats and occupiers. For example:

The real object of the fourteenth amendment of the constitution, in qualifying the words ‘all persons born in the United States’ by the addition ‘and subject to the jurisdiction thereof,’ would appear to have been to exclude, by the fewest and fittest words (besides children of members of the Indian tribes, standing in a peculiar relation to the national government, unknown to the common law), the two classes of cases,—children born of alien enemies in hostile occupation, and children of diplomatic representatives of a foreign state,—both of which, as has already been shown, by the law of England and by our own law, from the time of the first settlement of the English colonies in America, had been recognized exceptions to the fundamental rule of citizenship by birth within the country.

And it’s not as though the Court wasn’t thinking about non-residents and illegal immigrants. On the subject of non-residents, the Court approvingly quotes the common-law rule that birthright citizenship applies “whether the parents were settled, or merely temporarily sojourning, in the country, was an English subject.” The dissent is even clearer about this:

The English common-law rule recognized no exception in the instance of birth during the mere temporary or accidental sojourn of the parents. As allegiance sprang from the place of birth regardless of parentage, and supervened at the moment of birth, the inquiry whether the parents were permanently or only temporarily within the realm was wholly immaterial.

And on the subject of illegal immigration, the Court goes on for several paragraphs about how Congress has the total plenary authority to deport illegal Chinese immigrants for any reason whatsoever. For example, it says: “The right of the United States to expel such Chinese persons was placed upon the grounds that the right to exclude or to expel all aliens, or any class of aliens, absolutely or upon certain conditions, is an inherent and inalienable right of every sovereign and independent nation, essential to its safety, its independence, and its welfare.”

Yet immediately after expressing passionate support for the deportation of illegal Chinese immigrants, the Court solemnly intones: “The fourteenth amendment, while it leaves the power, where it was before, in congress, to regulate naturalization, has conferred no authority upon congress to restrict the effect of birth, declared by the constitution to constitute a sufficient and complete right to citizenship.” There’s no way to read this opinion as somehow leaving open the status of the children of non-residents and illegal immigrants.

Moreover, let’s take a step back. When we’re trying to figure out whether the Fourteenth Amendment confers birthright citizenship, we’re not trying to figure out whether birthright citizenship is a good idea. We’re trying to figure out what the word “jurisdiction” means. The pro-birthright-citizenship position has a coherent interpretation: “jurisdiction” means “authority.” Whereas there’s no way to gerrymander the word “jurisdiction” to include people like Wong Kim Ark whose parents were residents but ineligible to be citizens, but exclude people whose parents were non-residents or illegal immigrants. Maybe there’s a philosophical reason why we should draw the line this way, but that’s just not what “jurisdiction” means. You’ll never find the word “jurisdiction” being used this way in any other context; this definition of “jurisdiction” is invented for purposes of the birthright citizenship dispute.

Keep that analysis in mind as we look at the Wong Kim Ark dissent. The dissent says that a person isn’t entitled to birthright citizenship unless he’s completely subject to United States jurisdiction, and defines “completely subject” in the following way:

To be ‘completely subject’ to the political jurisdiction of the United States is to be in no respect or degree subject to the political jurisdiction of any other government.

And so why is Wong Kim Ark not a U.S. citizen, in the dissent’s view? The dissent opines that dual citizenship is not a thing:

Double allegiance, in the sense of double nationality, has no place in our law, and the existence of a man without a country is not recognized.

The dissent explains that all persons born in China are Chinese citizens for life, whether they like it or not:

Generally speaking, I understand the subjects of the emperor of China—that ancient empire, with its history of thousands of years, and its unbroken continuity in belief, traditions, and government, in spite of revolutions and changes of dynasty—to be bound to him by every conception of duty and by every principle of their religion, of which filial piety is the first and greatest commandment; and formerly, perhaps still, their penal laws denounced the severest penalties on those who renounced their country and allegiance, and their abettors, and, in effect, held the relatives at home of Chinese in foreign lands as hostages for their loyalty.

This means that all children of people born in China are Chinese citizens for life, whether they like it or not:

I take it that the children of aliens, whose parents have not only not renounced their allegiance to their native country, but are forbidden by its system of government, as well as by its positive laws, from doing so, and are not permitted to acquire another citizenship by the laws of the country into which they come, must necessarily remain themselves subject to the same sovereignty as their parents, and cannot, in the nature of things, be, any more than their parents, completely subject to the jurisdiction of such other country.

Thus, because dual citizenship is a contradiction in terms, and the children of Chinese citizens are themselves Chinese citizens, they cannot possibly be natural-born U.S. citizens.

This, of course, would imply that people of Chinese descent can never be natural-born U.S. citizens, no matter how many generations their ancestors have lived in the United States. The dissent says the quiet part out loud:

I am of opinion that the president and senate by treaty, and the congress by legislation, have the power, notwithstanding the fourteenth amendment, to prescribe that all persons of a particular race, or their children, cannot become citizens, and that it results that the consent to allow such persons to come into and reside within our geographical limits does not carry with it the imposition of citizenship upon children born to them while in this country under such consent, in spite of treaty and statute.

OK boomer! But I’ll give the dissent one thing: at least it’s offering what can reasonably be described as an interpretation of the phrase “subject to the jurisdiction.” Now that Wong Kim Ark is on the books, however, the anti-birthright-citizenship position requires “subject to the jurisdiction” to mean one thing for the children of residents who are ineligible for citizenship (i.e., Wong Kim Ark himself) and another thing for the children of non-residents and illegal immigrants. That’s not a genuine interpretation; it’s just treating the word “jurisdiction” as an empty vessel to be filled with whatever you want it to mean.

4. We’ve been doing it this way for a long time.

Ever since Wong Kim Ark was decided, the government has always assumed that the children of illegal immigrants and non-residents are citizens.

There’s a school of thought that says: “If people have thought for a long time that a constitutional provision carries a particular meaning, we should stick to that meaning.” I’m not going to give a bullet-pointed list of justifications for this school of thought. That is why God invented AI.

Whatever you may think of this view, it is influential among the Justices. In CFPB v. Community Financial Services Association, Justice Kagan, joined by Justices Sotomayor, Kavanaugh, and Barrett, wrote a concurrence observing that “‘[l]ong settled and established practice’ may have ‘great weight’” in interpreting constitutional provisions about the operation of government.” I’m not sure whether to treat the “long settled practice” as dating back to 1898 or some earlier point. But even if the right date is 1898, we were still using flip-phones back then. It’s a long time ago.

5. The debates

Members of Congress publicly debated the various provisions of the Fourteenth Amendment. The debates are interesting to read; to my eye, the Congressmen were far more eloquent then than they are now, although that is a low bar. You will find a lot of misleading partial quotations of the debates on the Internet, so if you’re interested in this topic you have to read the original source material. Like an animal!

There is only so much juice you can get out of the debates. It’s basically legislative history, which modern Supreme Court opinions tell us we should ignore. Also, the Congressmen spend a lot of time arguing with each other about what various provisions mean, so the debates are often inconclusive.

Still, there are some interesting nuggets in there. As far as I can tell, all of the debaters assumed that the U.S.-born children of non-diplomat foreigners would be U.S. citizens. Senator Wade thinks so:

Mr. FESSENDEN. Suppose a person is born here of parents from abroad temporarily in this country.

Mr. WADE. The Senator says a person may be born here and not be a citizen. I know that is so in one instance, in the case of the children of foreign ministers who reside “near” the United States, in the diplomatic language. By a fiction of law such persons are not supposed to be residing here, and under that fiction of law their children would not be citizens of the United States, although born in Washington. I agree to that, but my answer to the suggestion is that that is a simple matter, for it could hardly be applicable to more than two or three or four persons; and it would be best not to alter the law for that case.

So Senator Wade is saying: “It’s not a problem that the children of foreigners will get birthright citizenship, because there’s a legal fiction that the children of diplomats actually aren’t physically present in the United States, and anyway this happens rarely.” His underlying assumption is that under the Fourteenth Amendment, foreigners born on U.S. soil, other than the children of diplomats, will be citizens.

And then there’s this statement from Senator Howard:

Mr. HOWARD. …This amendment which I have offered is simply declaratory of what I regard as the law of the land already, that every person born within the limits of the United States, and subject to their jurisdiction, is by virtue of natural law and national law a citizen of the United States. This will not, of course, include persons born in the United States who are foreigners, aliens, who belong to the families of ambassadors or foreign ministers accredited to the Government of the United States, but will include every other class of persons.

I’ve seen some people on the Internet interpret Senator Howard to be saying that the children of “foreigners”—all foreigners—AND “aliens, who belong to the families of ambassadors or foreign ministers” won’t be citizens. But that doesn’t make sense. Otherwise he would have just said “foreigners” and there would have been no need to mention “ambassadors or foreign ministers.” To my eye, he’s saying that there’s a narrow exception to birthright citizenship for families of ambassadors or foreign ministers. Also, he makes a statement later in the debate opposing discrimination against Chinese nationals that, in context, makes clear that he’s assuming that the U.S.-born children of Chinese nationals will become U.S. citizens.

One more note. Multiple debaters opine that the children of Indians won’t become citizens because the United States lacks authority over them. For instance, Senator Trumbull says: “We have had in this country, and have today, a large region of country within the territorial limits of the United States, unorganized, over which we do not pretend to exercise any civil or criminal jurisdiction, where wild tribes of Indians roam at pleasure, subject to their own laws and regulations, and we do not pretend to interfere with them.” In other words, the debaters understand “jurisdiction” to conform to its natural meaning: “authority.”

In summary, there are numerous statements in the debate that support the birthright-citizenship position, and I don’t see any statements pointing in the other direction.

The curveballs

There are two other issues in the mix.

The first is the living Constitution. Vice President Vance opined: “Just because we were founded by immigrants, doesn’t mean that 240 years later that we have to have the dumbest immigration policy in the world.” I have seen similar arguments that birthright citizenship is so unbearably dumb that it just doesn’t matter what “jurisdiction” means or what the Supreme Court said in 1898. All of those people are dead. The common-law judges in England with their wigs who invented this rule are even deader. Illegal immigration is a crisis. Shouldn’t we do what makes sense?

Well, I am happy to entertain living-constitution arguments. Some of my best friends are living-constitutionalists! But I wouldn’t choose this case to get behind living-constitutionalism. Imagine telling someone: “You were born here. You went to school here. You’ve never been to China. You don’t know anyone in China. You don’t speak Chinese. The Fourteenth Amendment’s text says that you’re a U.S. citizen. The Supreme Court held in 1898 that you’re a U.S. citizen. But too bad, you’re being deported to China because the Constitution has undergone an epigenetic shift, like the Moirans in Seveneves.” I just don’t think it’s a very attractive argument.

The second is Congress. In 1940, Congress enacted a statute that repeats the words of the Constitution: “a person born in the United States, and subject to the jurisdiction thereof” is a citizen. Since then, everyone has always interpreted the statute to confer birthright citizenship on the children of non-residents and illegal immigrants. Wherefore, it is possible the Supreme Court will say something like: “regardless of whether the Constitution requires birthright citizenship, everyone has assumed since 1940 that an Act of Congress requires birthright citizenship, so we’ll stick with that settled meaning of the statute. We express no view as to whether a statute abrogating birthright citizenship would be constitutional.”

That would be an impressively statesmanlike resolution of the question presented. Everyone would come away happy. The Left would be happy because a Dub is a Dub. The Right would be happy because it opens the door a crack to eliminating birthright citizenship, even though everyone has long assumed the door was shut.

Still, in this iteration of the recurring debate between statesmanship and nerdiness, I am going with nerdiness. It is lame to say that the Constitution, which is worded identically to the statute, might or might not mean something different from the statute. The correct way to resolve the case is to say that the Constitution requires birthright citizenship.

Forward and backward

If you’re trying to figure out the answer to a legal question, you can either reason forward or you can reason backward.

Reasoning forward means applying neutral principles and attempting to deduce the correct legal answer without any preconceived view of what the result should be. You know, balls and strikes.

It is an understatement to say that neutral principles do not always yield an objectively correct answer. There are lots of bitterly divided Supreme Court cases in which every Justice applies, or at least attempts to apply, neutral principles, yet they cannot agree on the outcome. But just because neutral principles aren’t always determinate doesn’t mean they are never determinate. Sometimes, applying neutral principles yields clear answers, even when the subject matter is controversial.

Reasoning backward means starting with the desired result and attempting to reverse-engineer a legal theory supporting that result. It’s what lawyers do for a living: they start with a result (what their clients want) and then figure out how to get there.

Most judges profess to use the balls-and-strikes approach. One notable exception was Richard Posner, who described his method of judging this way:

My approach in judging a case is therefore not to worry initially about doctrine, precedent, and the other conventional materials of legal analysis, but instead to try to figure out the sensible solution to the problem or problems presented by the case. Once having found what I think is the sensible solution I ask whether it’s blocked by an authoritative precedent of the Supreme Court or by some other ukase that judges must obey. If it’s not blocked (usually it’s not—usually it can be got around by hook or by crook), I say fine—let’s go with the commonsense solution.

Few judges are as open about reasoning backward as Judge Posner. But any lawyer will tell you that it sometimes happens behind the scenes.

Judge Posner applies that form of reasoning in a concurring opinion opposing birthright citizenship. He first opines that there’s too much birth tourism, and then says: “We should not be encouraging foreigners to come to the United States solely to enable them to confer U.S. citizenship on their future children. … A constitutional amendment may be required to change the rule whereby birth in this country automatically confers U.S. citizenship, but I doubt it.” In other words, he starts from the premise that birthright citizenship is bad, and then optimistically conjectures that there must be some way to get it done legally.

It’s no coincidence that Judge Posner—one of the few judges to endorse the reasoning-backwards model—is also one of the few judges to have advocated against birthright citizenship. The birthright-citizenship issue is a classic example of an issue for which reasoning forward yields a single correct legal answer, and the sole way to reach the contrary conclusion is to reason backward.

All of the neutral tools of legal interpretation—constitutional text, history, and Supreme Court precedent—point in favor of birthright citizenship. By contrast, the arguments against birthright citizenship are of the following form:

• It would be a good idea to get rid of birthright citizenship.

• If you try hard enough, it is possible to construct a legal theory under which birthright citizenship loses.

Take a look at the government’s briefing in the birthright citizenship litigation. I don’t understand how it’s possible to be persuaded by these arguments unless you have already decided to adopt the anti-birthright-citizenship position and are straining to find a justification for it.

The government’s argument begins unpromisingly (p. 8) with the quotation “jurisdiction … is a word of many, too many meanings,” taken from a case in which none of the possible meanings bears any relationship to the government’s current interpretation. The government then offers the following syllogism (pp. 8-10)

• The argument that “jurisdiction” means “authority” has holes—after all, we do exercise some authority over diplomats, and Congress has broad authority to regulate Indian affairs.

• Therefore, we’re going to say that “jurisdiction” means “allegiance.”

• And mirabile dictu, we will declare that Wong Kim Ark had “allegiance” from the moment he was born but the children of non-residents and illegal immigrants don’t have “allegiance.”

Look … I just don’t see how you can say that the government’s position represents the best meaning of “jurisdiction.” The jurisdiction-equals-authority theory neatly fits all data points. By contrast, the notion that Wong Kim Ark had some greater degree of “allegiance” than children of other non-citizens is completely made up. What the government is really saying here is: “because the jurisdiction-equals-authority theory isn’t perfect, the door is ajar to depart from the birthright-citizenship position.” Well, maybe; but a judge wouldn’t open that door and walk through it unless the judge was searching for a reason to abolish birthright citizenship.

The government next says (p. 10) that the Fourteenth Amendment must be read the same way as the Civil Rights Act of 1866, which provided: “all persons born in the United States and not subject to any foreign power, excluding Indians not taxed, are hereby declared to be citizens of the United States.” I fail to understand how this theory helps the government distinguish Wong Kim Ark. Wong Kim Ark’s parents were Chinese citizens, and hence “subject to” a “foreign power.” If Wong Kim Ark himself was nonetheless not “subject to any foreign power,” why would the analysis be any different for the children of non-residents and illegal immigrants?

More fundamentally, why would we assume that the Fourteenth Amendment means the same thing as the Civil Rights Act? They’re worded differently. At the debates over the Fourteenth Amendment, the debaters realized that the wording was different. For instance, Senator Trumbull opined: “I think this language is better than the language employed by the civil rights bill.” If we’re trying to get to the best legal answer, why not interpret the text of the Constitution rather than statutory language that the drafters of the Fourteenth Amendment made a conscious decision not to use?

Next, the government observes (p. 11) that the Fourteenth Amendment uses the phrase: “citizens of the United States and of the States wherein they reside.” The government then cites some random Illinois state court case from 1840, which hasn’t been cited in the last 98 years, stating that an “inhabitant … has a fixed and legal settlement.” Based on this, the government heroically concludes that the word “reside” excludes illegal immigrants (and apparently also the children of illegal immigrants?), so they don’t “reside” in a state, so they can’t be “citizens of the United States.” When you contrast this extremely convoluted theory with the straightforward arguments for birthright citizenship, there’s no way you could actually be persuaded the theory is correct unless you define “correct” as “whatever will get me to the desired result of eliminating birthright citizenship.”

The government makes other arguments, but I don’t think it’s a productive use of space to rebut every single one of them. You will have to trust me that they are all weak. If you think one of them is strong, leave a comment and I’ll gladly address it.

It’s not just the government’s brief. Take the recent New York Times guest essay by Professors Barnett and Wurman on this issue. It is entitled “Trump Might Have a Case on Birthright Citizenship,” which to me doesn’t express much confidence in the legal argument to follow. It opens by emphasizing that Wong Kim Ark doesn’t “squarely” resolve the issue, which again seems inauspicious. After that, it leads off the anti-birthright-citizenship argument by quoting an 1862 opinion by Edward Bates, Abraham Lincoln’s Attorney General, addressing the citizenship status of free Black Americans. It then turns to a statement from Blackstone on the general topic of “allegiance” before turning to other, even less relevant sources.

I find the 1862 document hard to follow. You are free to read the back-and-forth-and-back on the Internet about exactly what Edward Bates was trying to say in 1862. However, we should not overlook the forest for the trees. The pro-birthright-citizenship argument has on its side the constitutional text, centuries of history, numerous explicit statements in a seminal Supreme Court case, and 127 additional years of history following said seminal case. The anti-birthright-citizenship argument, at least according to the essay, leads off with: (1) Wong Kim Ark doesn’t “squarely” resolve the question, and (2) if you squint, you can find some support for the anti-birthright-citizenship position in extraneous statements in a document that addresses a separate topic and was written at a time that the Fourteenth Amendment didn’t exist. Exactly like the government’s brief, it’s saying: “if you really want birthright citizenship not to exist, then you can, through Herculean efforts, construct some theoretical legal argument in favor of the anti-birthright-citizenship position!”

The horseshoe

What type of judicial philosophy would produce a ruling against birthright citizenship?

Well, any judge who would author such an opinion would likely have a conservative reputation. But from a methodological perspective, such a judge would fit in well in the Yale Law School faculty lounge.

I’ve opined before about the horseshoe theory of law: the tendency of the methodologies of very progressive and very conservative judges to converge. Judges at both ends of the ideological spectrum tend to view judging as a species of cause lawyering. They may differ on which causes are righteous, but their view of the judge’s role is the same—to vindicate a cause if an argument is theoretically available to support it.

This view of judging is traditionally associated with the legal Left. One can find many law review articles endorsing the view that judges are charged with doing Justice, with Justice defined in progressive terms. But experience has taught that some judges on the legal Right view judging the same way.

The anti-birthright-citizenship position is a quintessential example of a position that prevails only in front of a judge on the right end of the horseshoe. It requires the conjunction of conservative policy views and a willingness to disregard neutral principles in order to reach results consistent with those views.1

I don’t have strong views about whether birthright citizenship, as a policy matter, is good or bad. But I do have strong views that neutral principles exist and judges should try their best to use them. I’m also idealistic enough to believe that most judges agree. We’ll see.

Tennessee Valley Authority v. Hill, 437 U.S. 153 (1978), is probably the most famous Supreme Court environmental law case ever. In Hill, the Supreme Court held that construction of the Tellico Dam in Tennessee—which was over 80% complete—had to be halted under the Endangered Species Act to protect a recently-discovered fish species known as the snail darter. Hill remains a staple of both the environmental law and statutory interpretation curricula.

Recently, the New York Times reported on a new study revealing that the snail darter isn’t a species after all. Instead, it is an eastern population of a different fish known as the stargazing darter, which is not endangered. So the whole epic struggle was premised on a scientific mistake.

Whoa! If you know me, you know this is the sort of story that would make me a little bit obsessed. So I dug into this. I relied on the following sources:

• The original article recognizing the snail darter: David A. Etnier, Percina (Imostoma) Tanasi, A new percid fish from the little Tennessee River, Tennessee, Proceedings of the Biological Society of Washington (Jan. 22, 1976).

• The new study: Ava Ghezelayagh et al., Comparative species delimitation of a biological conservation icon, Current Biology (2024), https://doi.org/10.1016/j.cub.2024.11.053 .

• Although Dr. Ghezelayagh’s article is clearly written, there were a few aspects of her methodology I didn’t fully understand. So, I emailed her with some pedantic questions and quickly received a detailed and thorough response. I’m very grateful to Dr. Ghezelayagh for this, it’s pretty awesome that a scientist would respond to a random lay person’s cold email about her academic work.

• Finally, this post would not have been possible without AI. You can upload the article to Claude and ask questions like “I don’t understand the chart on page 5, what are the red rectangles and blue rectangles” and it will explain it to you until you understand. It’s like having an infinitely patient person with a Ph.D. in all subjects sitting next to you. To the autodidacts of the world, this is truly the greatest technology in human history.

Here are my takeaways:

• This is mostly a heartwarming story about science getting better. It was reasonable for Dr. Etnier to conclude in 1976 that the snail darter was a new species. Dr. Ghezelayagh’s study vindicates many of Dr. Etnier’s findings and reaches a different conclusion from Dr. Etnier based on new technology that would have been unthinkable in 1976.

• That said, this case does illustrate how law can influence science. If scientists know that categorizing an animal population as a “species” will halt a construction project that they support or oppose, then they have an incentive to shade their analysis accordingly.

• Structuring the administrative state to distinguish good science from bad science is a notoriously difficult problem, but the situation is much better now than it was in 1976.

Wedding bells

In 1973, Congress enacted the Endangered Species Act. Here’s what it says: “Each federal agency shall … insure that any action authorized, funded, or carried out by such agency … is not likely to jeopardize the continued existence of any endangered species or threatened species or result in the destruction or adverse modification of habitat of such species which is determined … to be critical.”

OK, so the Endangered Species Act protects endangered “species.” What’s a “species”?

There’s actually no clear definition. Intuitively, one would think it means something like: “if you put a male and female fish in a tank, and wedding bells ring, they’re the same species.” But for several reasons, it’s more complicated than that.

First, sometimes distinct animal populations form a chain, in which adjacent populations can interbreed but the populations at each end of the chain can’t interbreed with each other—a so-called “ring species.”

This phenomenon makes it challenging to decide the metes and bounds of a “species” for Endangered Species Act purposes. Suppose you have three populations of animals—A, B, and C. A and C are both endangered, but B is abundant. A can interbreed with B, and C can interbreed with B, but A and C can’t interbreed with each other. Are A and C “endangered species”? A developer would say: “no, A is the same species as B and C is the same species as B, so why are we stopping the project?” An environmentalist would retort: “If A is the same species as B, and C is the same species as B, this would imply that A is the same species as C, but that’s wrong because A and C can’t interbreed!” There’s no perfectly satisfactory definition of a “species” in this situation.

Second, sometimes animals are different enough from one another that they’re widely considered to be separate species, even though they can, in principle, interbreed. Polar bears and grizzly bears occasionally cross paths and bear fertile offspring; does this mean that they’re not separate species? Saying “it’s OK to let the grizzly bear go extinct because polar bears also exist” seems contrary to Congress’s purpose in enacting the Endangered Species Act. As another example, humans and Neanderthals are generally considered separate species even though they interbred from time to time.

Third, conversely, sometimes two populations of animals are sufficiently similar that they are considered the same species, even if they’re geographically isolated and therefore don’t, as a practical matter, interbreed. Or maybe they just don’t want to interbreed. Experience has taught that humans can occasionally be picky, and the same is true of our fellow animals.

To make matters more complicated, the Endangered Species Act defines “species” to include “any subspecies of fish or wildlife or plants, and any distinct population segment of any species of vertebrate fish or wildlife which interbreeds when mature.” So even if a particular group of fish doesn’t form its own “species,” it’s still protected if it forms a “subspecies” or “distinct population segment.” These terms also lead to difficult line-drawing problems.

Eureka!

In the 1960s, the Tennessee Valley Authority decided to build a dam on the Little Tennessee River known as the Tellico Dam. This wasn’t a dam for purposes of generating electricity. Its purpose was to create a big man-made reservoir (“Tellico Lake”) for purposes of stimulating tourism.

Building Tellico Dam would have the effect of wiping out the lower portion of the Little Tennessee River. In 1973, Tennessee’s leading ichthyologist, Dr. David Etnier, decided to check out the local fish population before their day of doom. And he spotted a funny-looking brown fish that was approximately 3 inches long.

That discovery took on new significance a few months later, when Congress enacted the Endangered Species Act. Congress undoubtedly had charismatic species in mind like whooping cranes and grizzly bears. But God in His wisdom chose to create uninspiring animals too. And, as far as Dr. Etnier was aware, this particular fish was found only in that particular stretch of the Little Tennessee River. If Tellico Dam was built, the hapless snail darter (so named because it eats snails) would go extinct.

In 1976, Dr. Etnier published his legendary article recognizing the snail darter as a new species. Here is how the article describes the fateful discovery:

Robert A. Stiles and I visited the river at Coytee Spring, River Mile 7, Loudon County, Tennessee, on 12 August 1973, in an attempt to survey the fish fauna with the aid of face masks and snorkels. Although very few fishes were seen, a single specimen of a very unusual darter was observed and “scooped up” by hand by the author.

Now that’s science! Dr. Etnier took a close look at this very unusual darter and concluded it was a new species.

How did he reach this conclusion? Modern genetic analysis didn’t exist in 1976, so Dr. Etnier’s methodology consisted of collecting a bunch of snail darters, taking measurements, and deciding that the snail darter looks different enough from all other darters to be its own species. It’s the type of article that could have been written in 1776.

Dr. Etnier acknowledged that the snail darter, Percina tanasi, is pretty similar to the stargazing darter, Percina uranidea, which is found in nearby rivers. But he concluded that the snail darter “differs from known populations of P. uranidea in body width, paired fin length, saddle width, nuptial tubercle pattern, several aspects of pigmentation, number of anal and caudal fin rays, and probably vertebral number.” The most striking difference between the two species was their anal fin ray counts: most of the snail darters he found had 12 anal fin rays (though some had 11), while most of the stargazing darters he found had 11 anal fin rays (though some had 12).

In scientific significance, this discovery wasn’t exactly the discovery of the coelacanth. The Supreme Court’s opinion notes that there were 130 known darters at the time, 8 to 10 of which were identified in the preceding five years. It seems we are now up to approximately 248 different darter species.

Still, as far as the Endangered Species Act is concerned, all darter species are sacred. Local citizens petitioned the Fish and Wildlife Service to list the snail darter as an endangered species, and the Fish and Wildlife Service complied.

The Federal Register notice finding the snail darter to be an endangered species is, to the modern eye, an interesting read. The agency said that sixteen comments were received. Twelve commenters supported the listing of the snail darter as an endangered species. Three commenters offered views that were dryly described as not “relevant to the biological evaluation.” The final commenter was the Tennessee Valley Authority—the agency that was building the dam. The TVA didn’t exactly argue that the snail darter wasn’t a new species, but instead said that there wasn’t enough information to make that assessment, and that it really, really wanted to complete the dam.

The Fish and Wildlife Service said, in essence, that Dr. Etnier published an article describing the snail darter as a new species, and there was no contrary research, so whaddaya gonna do? Argue with Tennessee’s leading ichthyologist? If this had come up today, there would have been hundreds of comments on both sides, many AI-generated. The 1970s were a gentler time, I guess. Given the record before it, it’s hard to blame the Fish and Wildlife Service for agreeing with Dr. Etnier that the snail darter is a new species.

And the rest is history. Opponents of the Tellico Dam sued the TVA, alleging that the dam would jeopardize the snail darter’s continued existence. If you believe Dr. Etnier that the snail darter is a distinct species that lives only in this specific stretch of river, that was undeniably true. The TVA argued that it was absurd to halt the construction of a dam that was 80% complete to save this obscure and inglorious fish. But, the Supreme Court held, the law is the law.

Congress later stepped in to override the Supreme Court’s decision and ensure that the dam would be completed. The reservoir created by the dam— Tellico Lake—now exists. Based on the YouTube video touting it, I would characterize it as moderately scenic. Meanwhile, the snail darter has been found in nearby rivers, so all’s well that ends well.

Caveman genius

Fast forward 40 years.

Dr. Ghezelayagh’s article concludes that the snail darter is the same species as the stargazing darter.

Dr. Ghezelayagh takes samples of DNA sequences from 249 snail darters and 81 stargazing darters. The stargazing darters are drawn from two populations, one in the White River in Missouri and Arkansas and one in the Ouachita River in Arkansas. Figure 2 of the article depicts the results in a phylogenomic tree (a diagram that shows biological relationships between animals by comparing their DNA).

The phylogenetic tree reveals that snail darters (red branches) and the two groups of stargazing darters (blue branches) are indeed genetically distinct from one another: according to the paper, “three major lineages are resolved with strong node support across a variety of data filtering thresholds and methods: all sampled snail darter specimens, stargazing darters from the White River system in Arkansas and Missouri, and stargazing darters from the Ouachita River system in Arkansas.” But there are “extremely short branch lengths,” meaning that the distinctions among these groups are small. Also, even though the two groups of stargazing darters are generally considered a single species, they’re about as far apart from each other as they are from the snail darter.

I come away from this analysis extremely impressed with Dr. Etnier’s work. To be sure, in terms of level of sophistication, Dr. Ghezelayagh’s article makes Dr. Etnier’s article look like it came from the Paleolithic era. Dr. Ghezelayagh conducts an elaborate genomic analysis; Dr. Etnier scoops up a fish and measures it with a ruler. But Dr. Ghezelayagh’s work also makes Dr. Etnier look like a kind of caveman genius. Of all the 100+ darters, Dr. Etnier immediately recognized that the snail darter was closest to the stargazing darter merely by examining it and watching it wriggle. Also, Dr. Etnier is aware of the separate population of Ouachita River darters. He’s agnostic as to whether they’re a separate species from the White River darters but concludes: “The observable differences between these sympatric species (not recognized by modern ichthyologists as being distinct until 1970) are similar in magnitude to those between P. uranidea and P. tanasi.” And that’s exactly what the genomic analysis shows!

IDK, to me you have to be a fish genius to figure out these genetic relationships from just staring at the fish when there are 100+ darter species out there. To me a minnow is a minnow. Maybe this is obvious to an ichthyologist but I cannot imagine myself memorizing subtle differences among 100 types of darters.

Anyway, Dr. Ghezelayagh nonetheless concludes that snail darters are just a subpopulation of stargazing darters. She reaches that conclusion in a clever way. She takes 12 pairs of darter sister species—that is, pairs of darter species more similar to each other than to any other darter species, but still widely recognized as separate species—and compares how divergent they are using a genomic analysis. She finds that there is much greater genomic divergence between each of those 12 reference pairs than between the snail darter and stargazing darter. Therefore, she concludes, “the level of genetic divergence between the Snail Darter and Stargazing Darter is more consistent with population-level differences than differences between distinct species.”

From my perspective as a lawyer, this analysis makes sense. Congress didn’t have any particular technical definition of “species” in mind when it enacted the Endangered Species Act. Instead Congress meant something like: “a population of animals that is different enough from other populations that scientists would consider the population to be a separate ‘species,’ according to the way scientists conventionally use that term.” And Dr. Ghezelayagh’s method is consistent with that interpretation. She identifies pairs of darters that are considered to be closely related but nonetheless distinct species, and concludes that the degree of divergence among those pairs marks the dividing line between “distinct species” and “distinct populations.” Obviously there will be some grey areas, but the paper persuasively explains that the snail darter/stargazing darter pair isn’t in that grey area—it’s firmly in distinct-population zone.

The New York Times article implies that Dr. Etnier distorted his findings in order to impede the Tellico Dam and that the new study debunks Dr. Etnier’s work. I don’t know what was going through Dr. Etnier’s mind, but that’s the wrong take. Dr. Ghezelayagh frames her study as presenting a new protocol for determining whether a population is a “species,” not as calling out an error by Dr. Etnier. Also, none of the sophisticated technological tools she relies upon were available when Dr. Etnier’s article was published. It’s not clear Dr. Etnier even used a computer in writing his article. It’s like comparing Pong—the state-of-the-art game when the snail darter was discovered—and Elden Ring.

You may be wondering: the Endangered Species Act deems “subspecies” and “distinct population groups” to be distinct “species,” at least as far as fish are concerned. Might the snail darter be a “subspecies” or “distinct population group” and therefore entitled to protection even if it’s not a separate “species”? The study doesn’t express a view on this: “It is not clear if the level of distinctiveness of the Snail Darter would merit protection as either a subspecies or DPS.” So, the study actually doesn’t take a position on whether TVA v. Hill should have come out the other way.

(Incidentally, the New York Times article is way off on this. It says the study reports that “the snail darter, Percina tanasi, is neither a distinct species nor a subspecies,” when the study actually doesn’t take a position on whether the snail darter is a subspecies. This isn’t a trivial error: the Times is implying that the study concludes that the snail darter shouldn’t have stopped construction of the Tellico Dam, when in fact the study goes out of its way to say the exact opposite thing.)

Dr. Ghezelayagh’s study also examines the morphological differences between snail darters and stargazing darters, and concludes, in a nutshell, that the two types of darters are a little different, but less different than Dr. Etnier thought. For example, Dr. Etnier’s measurements revealed that the modal stargazing darter had 11 anal fin rays and the modal snail darter had 12 anal fin rays, but Dr. Ghezelayagh conducts her own measurements on a larger sample set and concludes that both types of fish have a mode of 11 anal fin rays. But, according to Dr. Ghezelayagh’s measurements, the median snail darter has about 0.4 more anal fin rays than the median stargazing darter, so Dr. Etnier was not completely off base.

Why the distinction? There are lots of possibilities. Maybe the two scientists counted anal fin rays in slightly different ways. (Here’s a YouTube video walking through how anal fin ray counts can sometimes be ambiguous, there truly is everything on the Internet.) Also, Dr. Etnier examined only the Little Tennessee River snail darters, while Dr. Ghezelayagh examined other populations of subsequently-discovered snail darters. Anyway, the measurements weren’t that different—divergences of this degree are common in scientific research. I basically view Dr. Ghezelayagh as having replicated Dr. Etnier’s work.

The sturgeon in the room

Even if Dr. Etnier is a fish hero, this episode does illustrate the risk of politics influencing science.

If a scientist knows that classifying a population as a “species” will halt a construction project that the scientist opposes, then of course the scientist will have an incentive to err on the side of calling it a “species.” Or vice versa.

I’m not saying the scientist will falsify data or engage in misconduct. I’m saying that at the margins, there is no clear line between a population and a species. It’s like trying to distinguish between a dialect and a language. It’s a discretionary call. And it’s human nature for scientists to exercise their discretion to pick the side that aligns with their preferred outcome.

This can be pernicious because it can be difficult to identify bias in a scientific paper. The scientist may present the study as a neutral, objective analysis. Precisely because the scientist is an expert, the scientist may be able to conceal bias in a way that non-scientists can’t detect. Just look at Dr. Etnier’s article. Suppose that Dr. Etnier knew that the morphological differences between the stargazing darter and snail darter were much less significant than the morphological differences between other sister pairs of darters. This would have undermined his claim that the snail darter was a separate species. But how would a non-scientist know that from reading his article? The only person in a position to convey that information was the darter expert—Dr. Etnier.

Our civil litigation system deals with this problem in a rather extreme way: it prevents scientists from deciding scientific questions. Instead, when there’s a dispute over science, our system assumes scientists will act as advocates. The plaintiff and the defendant each hire expert witnesses, who are scientist/lawyer hybrids who deploy science to advocate for one side of the case. The ultimate decision is made not by an expert, but instead by a neutral judge or in some cases random citizens whose sole credential is their ignorance and hence presumed neutrality. In administrative law, however, agencies often don’t have access to dueling expert reports and instead must rely on the scientific literature as best they could.

This blog post will not try to solve the problem of bad science in administrative law. Instead, I will close with a note of optimism. My takeaway from Dr. Ghezelayagh’s article is that scientific and technological advancements have dramatically reduced the risk that “species” determinations will be influenced by bad science:

1. We can do now do genomic analysis, which reduces the number of borderline cases.

2. Genomic analysis also allows for the creation of objective criteria. “Lineages are distinct species when the genomic divergence index is greater than 0.7” is less manipulable than “Lineages are distinct species when I, a distinguished ichthyology professor, say they are.”

3. AI has many use cases here. For instance, AI can make scientific research more comprehensible to lay readers.

4. Even setting aside technological advancement, we know more stuff now than we did in 1976. Dr. Ghezelayagh was able to draw upon existing research identifying 12 sister pairs of darter species that could be used as reference pairs. That research didn’t exist in 1976.

5. It’s easier to identify and respond to bad science today than it was in 1976. Dr. Etnier’s original study said the snail darter was its own species because it was morphologically different from the stargazing darter. But, even based on Dr. Etnier’s data, the two types of darters weren’t that different. Why didn’t a scientist offer an opposing view to the Fish and Wildlife Service? Probably because only a minuscule number of scientists were aware of the issue. A few scientists did submit comments supporting Dr. Etnier’s position, but I’m guessing they were all people that knew Dr. Etnier. This wouldn’t happen today; the study would go viral.

Rather than lamenting the scientific errors of the past, we should exult over the present.

That’s not quite right. Congress doesn’t actually spend money. Instead, it enacts laws appropriating money for particular purposes. It’s the Executive Branch, controlled by the President, that spends the money.

Aha! A solution presents itself. Can’t the President just not spend the money?

This idea has come up enough times in American history to have earned its own name: “impoundment.” To make a long story short, Presidents have impounded funds from time to time throughout American history. But then President Nixon started impounding a lot more than prior Presidents. Congress responded by enacting a law in 1974 called the Impoundment Control Act, which bars the President from impounding under most circumstances.

So the President can’t impound. Unless the Impoundment Control Act is unconstitutional!

Most people do not think the Impoundment Control Act is unconstitutional. But one person who does, Mark Paoletta, will be the General Counsel of the Office of Management and Budget in the Trump administration. This means there’s a good chance that President Trump will impound funds during his administration.

It will go roughly like this: DOGE will identify a woke-looking appropriation, perhaps something to do with environmental justice or renewable energy. The President will send out a Tweet declaring that the money appropriated for this initiative will instead be used to purchase Dogecoin. Thousands on Twitter/X will pronounce this action “based.” A maelstrom of litigation will ensue. Plaintiffs will file suits in friendly venues and obtain quick preliminary injunctions from Democratic appointees. After a few months, the case will reach the U.S. Supreme Court.

In this post I’m going to do two things. First, I’ll give my own take on the legal issues. TLDR: I think the Impoundment Control Act is constitutional.

Second, I’ll try to predict what the Supreme Court will do. But rather than use my own often questionable judgment, I’ll use AI. Several months ago, I wrote a post explaining that AI is very good at reading briefs and predicting Supreme Court decisions. Unfortunately, we don’t have the briefs yet.

Not a problem! In the absence of human-created briefs, I decided to simulate the case from start to finish. I used Claude to draft full-length opening, response, and reply briefs based on a hypothetical impoundment scenario that Claude invented from whole cloth. I then used Claude to simulate an oral argument. Later in the post you can find copies of the briefs and oral argument transcript.

I then put Claude into adjudicator mode and asked it to review those AI-generated briefs and predict how the case would come out. Claude predicted a 6-3 opinion by Chief Justice Roberts upholding the Impoundment Control Act, with Justices Thomas, Alito, and Gorsuch dissenting. Claude also drafted the opinion of the Court, with concurrences by Justices Kagan and Kavanaugh and a dissent by Justice Thomas. Consider it preregistered!

(The theme of this week’s artwork is “pound.”)

The Laws of Constitutional Litigation

Under the Impoundment Control Act, if Congress appropriates money to be spent, the President must spend it. The President can’t rescind spending unless he receives permission from Congress. The President can temporarily defer spending without congressional approval for non-policy reasons (e.g., some unexpected contingency arises). But he can’t defer spending based merely on a disagreement with the policy underlying the appropriation.

Is the Impoundment Control Act constitutional?

Let’s begin by taking a look at the constitutional text, which is occasionally relevant to Supreme Court decisions interpreting the Constitution. The Appropriations Clause provides: “No Money shall be drawn from the Treasury, but in Consequence of Appropriations made by Law.” This tells us that Congress has the power to enact Laws that “make” Appropriations. “Laws” are generally mandatory, so a “Law” saying “Thou shalt spend $3 billion on Program X” does sound like a Law that makes an Appropriation. Congress therefore has the enumerated power to enact a law that mandates that appropriations be spent.

Proponents of the President’s constitutional right to impound generally don’t argue that the Impoundment Control Act falls outside of Congress’s enumerated powers. Instead, they argue that Article II confers the President with the right to impound. Article II provides that “the executive power shall be vested in a President of the United States of America” and directs the President to “take care that the laws be faithfully executed.”

It may seem odd that the obligation to “take care that the laws be faithfully executed” gives the President the right to refuse to faithfully execute the Impoundment Control Act. But this reasoning ignores the First Law of Constitutional Litigation: “All propositions can be defended by reference to the Constitution’s structure.” Proponents of the right-to-impound theory point to structural separation-of-powers principles: they contend that the Constitution grants the President discretion on how to enforce the law, and the President can’t exercise discretion unless he decides how much money he will spend.

To be fair to this argument, there are, indeed, some circumstances in which the President may declare himself unbound by a statute based on considerations of constitutional structure. Justice Jackson observed in his iconic Youngstown concurrence that “[w]hen the President takes measures incompatible with the expressed or implied will of Congress, his power is at its lowest ebb, for then he can rely only upon his own constitutional powers minus any constitutional powers of Congress over the matter.” But “lowest ebb” doesn’t mean nothing.

In Zivotofsky v. Kerry, 576 U.S. 1 (2015), for example, the Supreme Court struck down a statute providing that for purposes of the “issuance of a passport of a United States citizen born in the city of Jerusalem,” the Secretary of State “shall, upon the request of the citizen or the citizen’s legal guardian, record the place of birth as Israel.” In the Supreme Court’s view, a passport’s statement that the bearer was born in “Jerusalem, Israel” is essentially a diplomatic message from the United States recognizing Israel’s sovereignty over Jerusalem. At the time, however, the President didn’t recognize Israel’s sovereignty over Jerusalem. Thus, the Secretary of State (who reports to the President) argued that he shouldn’t have to follow the statute because it interfered with the President’s authority to recognize foreign sovereigns.

Nothing in the Constitution expressly bars Congress from legislating regarding the content of passports. But the Court identified various constitutional provisions conferring powers on the President that seem in the same ballpark as recognizing foreign sovereigns, such as the power to “receive Ambassadors and other public Ministers,” to nominate our own Ambassadors, and to enter into foreign treaties. Therefore, the Court concluded, the Constitution’s structure establishes that the President has exclusive authority to decide whether to recognize a foreign country’s sovereignty. And so, the Court held, the statute was unconstitutional because it intruded on that power.

I am skeptical of the result in Zivotofsky. Justice Scalia’s dissents are often so persuasive that it is impossible to understand how anyone could have reached the opposite conclusion, and his beautifully-written Zivotofsky dissent is an exemplar of this genre. But Zivotofsky remains on the books, and it confirms that there are some things the President gets to decide without congressional meddling.

Zivotofsky implies that the President can impound funds if spending those funds would interfere with an exclusive presidential power. So if Congress enacted a statute saying “we appropriate $10,000 for the sole purpose of printing new passports saying ‘Jerusalem, Israel’ on them,” the President could decline to spend those funds. The same is true for other exclusive presidential powers: if Congress were to appropriate money for the specific purpose of pardoning a particular federal inmate (the money could be allocated to the cost of the printer paper and ink used to create the document memorializing the pardon), the President wouldn’t have to spend that money.

But setting aside unique cases of that type, neither Zivotofsky nor any other case suggests that the President has the general authority to refuse to spend appropriated funds. And there are several cases suggesting that the President lacks that power:

• Kendall v. United States ex rel. Stokes, 37 U.S. (12 Pet.) 524 (1838). After the Postmaster General refused to pay a contractor for delivering mail, the contractor persuaded Congress to pass a bill requiring the Postmaster General to pay him. The Postmaster General refused, so the contractor sued. Held: the Postmaster General was required to pay.

• Train v. City of New York, 420 U.S. 35 (1975). A federal statute appropriated money for certain environmental projects and stated that “[s]ums authorized to be appropriated … shall be allotted” by the EPA. President Nixon directed the EPA not to spend all of the money. Held: the EPA was required to spend all of the money.

• Clinton v. City of New York, 524 U.S. 417 (1998). The Supreme Court struck down the Line Item Veto Act, which would have allowed the President to cancel certain items of spending. As a practical matter, the line item veto isn’t so different from impoundment, in which the President makes a discretionary decision not to spend certain funds. In Clinton, the Court held that Congress couldn’t give the President this impoundment-like power even if it wanted to.

• Justice Scalia’s Clinton dissent. Justice Scalia’s dissents aren’t technically binding precedent but they’re so persuasive that they may as well be. Justice Scalia would have upheld the Line Item Veto Act because it’s functionally similar to impoundment. But he didn’t endorse a constitutional right to impound. After explaining that Presidents prior to Nixon frequently impounded funds, Justice Scalia observed: “President Nixon, the Mahatma Gandhi of all impounders, asserted at a press conference in 1973 that his ‘constitutional right’ to impound appropriated funds was ‘absolutely clear.’” (I do not understand the Mahatma Gandhi analogy at all but still find it hilarious.). But Justice Scalia then said: “Our decision two years later in [Train] proved him wrong, but it implicitly confirmed that Congress may confer discretion upon the Executive to withhold appropriated funds, even funds appropriated for a specific purpose.” So Justice Scalia also didn’t think the President can impound against the will of Congress.

So, we have cases saying: (1) the President is constitutionally obligated to obey a statute requiring a debt to be paid (Kendall); (2) a statute requiring the President to spend money is judicially enforceable (Train); (3) Congress can’t give the President an impoundment-like power even if it wants to (Clinton majority); (4) Congress gets to decide whether to give the President an impoundment-like power (Clinton dissent). Those cases are hard to square with the right-to-impound theory.

But we cannot forget the Second Law of Constitutional Litigation (really, of all litigation): “Every case is distinguishable from every other case.” If one is sufficiently motivated to recognize a right to impound, one can say that Kendall involved the adjudication of a private claim, Train was a statutory interpretation case, Clinton was a case about “cancellation” of a law rather than a mere refusal to follow it, and Justice Scalia’s Clinton dissent is non-binding and wrong. This would require a lot of motivation, though.

Everything that’s not illegal is mandatory

The right-to-impound argument hinges primarily on history. Before the Impoundment Control Act, Presidents dating back to Jefferson would occasionally impound funds. So, the argument goes, there’s a tradition of impoundment that the Impoundment Control Act can’t override.

This argument is the mirror image of the argument the Supreme Court recently rejected in CFPB v. Community Financial Services Association, 601 U.S. 416 (2024). In that case, the challengers argued that the CFPB (short for Consumer Financial Protection Bureau) was unconstitutional because, under the CFPB’s unique funding scheme, the President could impound.

That’s an oversimplification, but only by a bit. The CFPB is funded in an unusual way. Most agencies have to clamor for funds during the annual appropriations process. But the CFPB doesn’t—it’s authorized to draw an amount from the Federal Reserve System that its Director deems “reasonably necessary to carry out” the agency’s duties, subject to an inflation-adjusted cap that was approximately $734 million in 2022. The challengers objected to several aspects of this scheme (e.g., the fact that it was permanent), but the molten core of the challengers’ argument was that the agency got to decide how much to spend, as this back-and-forth from the oral argument illustrated:

JUSTICE THOMAS: Mr. Francisco, just briefly, I'd like you to complete this sentence. Funding of the CFPB is -- violates the Appropriations Clause because?

MR. FRANCISCO: Because Congress has not determined the amount that this agency should be spending. Instead, it has delegated to the director the authority to pick his own appropriation, subject only to an upper limit that's so -- so high it's rarely meaningful.

I disagree that $734 million is an amount that’s “so high it’s rarely meaningful”; I promise you that a federal agency can spend $734 million on staplers without any difficulty. But, the CFPB Director was indeed guilty as charged of having the discretion to decide how much he would spend up to that amount.

The Fifth Circuit held that this scheme was unconstitutional, finding it to be “the epitome of the unification of the purse and the sword in the executive—an abomination the Framers warned ‘would destroy that division of powers on which political liberty is founded.’” Sounds bad!

The Supreme Court reversed the Fifth Circuit. It reasoned that there’s nothing in the Constitution saying the CFPB can’t be funded this way. The Court also examined the historical record and noted that prior to the Founding, legislatures had significant flexibility:

In short, the origins of the Appropriations Clause confirm that appropriations needed to designate particular revenues for identified purposes. Beyond that, however, early legislative bodies exercised a wide range of discretion. Some appropriations required expenditure of a particular amount, while others allowed the recipient of the appropriated money to spend up to a cap. Some appropriations were time limited, others were not. And, the specificity with which appropriations designated the objects of the expenditures varied greatly.

Likewise, in the early history of the United States, appropriations bills “adhered to the minimum requirements of an identifiable source of public funds and purpose,” but other than that, they “displayed significant variety in their structure.” Sometimes Congress would appropriate lump sums, sometimes it would appropriate sums up to a cap, sometimes it would appropriate in other ways. Based on this historical evidence, the Court held that Congress can—if it wants to—let the President decide how much money he spends:

At the outset, the associations argue that the Bureau’s funding is not “drawn . . . in Consequence of Appropriations made by Law” because the agency, rather than Congress, decides the amount of annual funding that it draws from the Federal Reserve System. This argument proceeds from a mistaken premise. Congress determined the amount of the Bureau’s annual funding by imposing a statutory cap. … The only sense in which the Bureau decides its own funding, then, is by exercising its discretion to draw less than the statutory cap. But, as we have explained, “sums not exceeding” appropriations, which provided the Executive with the same discretion, were commonplace immediately after the founding. … Thus, we cannot conclude that Congress violated the Appropriations Clause by permitting the Bureau to decide how much funding to draw up to a cap.

With the change in Administrations, we now have the mirror-image argument. Instead of arguing that it’s unconstitutional to let the President decide how much to spend, proponents of the right-to-impound theory argue that it’s unconstitutional not to let the President decide how much to spend. This is a nice example of the First Law of Constitutional Litigation at work: the proponents are arguing that a system that the Fifth Circuit deemed a separation-of-powers-destroying “abomination” is actually mandatory under the separation of powers.

If this issue reaches the Supreme Court, I predict the Court will uphold the Impoundment Control Act. In CFPB, the Court observed: “early legislative bodies exercised a wide range of discretion. Some appropriations required expenditure of a particular amount, while others allowed the recipient of the appropriated money to spend up to a cap.” The Court inferred that Congress continues to have that discretion today.

That’s why the right-to-impound theory loses. Early legislatures sometimes “allowed the recipient of the appropriated money to spend up to a cap”—and Congress can keep doing that today if it so desires, as it has with the CFPB. Early legislatures also sometimes “required expenditure of a particular amount”—and Congress can also keep doing that today if it so desires, as it has under the Impoundment Control Act. The historical argument for the right to impound fails for the same reason as the historical argument in CFPB: it confuses “X has sometimes happened” with “X is constitutionally mandatory.” Early legislatures sometimes controlled how much the President spends, but CFPB holds that such a scheme isn’t constitutionally mandatory. Likewise, early legislatures sometimes didn’t control how much the President spends, but that’s not constitutionally mandatory, either.

The inverse major questions doctrine

The advent of the so-called “major questions doctrine” also makes me skeptical of a right to impound.

The major questions doctrine teaches that Congress, not the administrative state, is responsible for deciding national policy on issues of major political and economic importance. Thus, an agency isn’t permitted to rely on vague and general statutory language as a basis for regulating on a major question, even if that vague and general language, read literally, would seem to authorize the regulation. Instead, Congress must speak clearly before an agency may enact such a regulation.

For example, in Biden v. Nebraska, 600 U.S. 477 (2023), the Supreme Court struck down the Department of Education’s rule canceling $430 billion of student loans. The Department argued that it had statutory authority to issue the rule: a statute authorized the Department to “waive or modify any statutory or regulatory provision applicable to the student financial assistance programs” at issue, so the Department argued that it was “waiving or modifying” the requirement to pay back the principal on the loans. The Department had a decent argument based on the literal words of the statute. But, the Court held, c’mon … when Congress authorized the Department to “waive or modify any statutory or regulatory provision,” it was talking about little modifications, not canceling $430 billion worth of loans!

The major questions doctrine has spawned a vast academic literature, to which I decline to contribute. Instead, I will observe that the right-to-impound argument is essentially the inverse of the major questions doctrine. The question of how much the government should spend is the paradigmatic example of a major question. In the Impoundment Control Act, Congress provided that agencies cannot decide that major question by themselves. Whereas the major questions doctrine states that agencies can’t decide major questions even when statutes, read literally, authorize them to do so, the right-to-impound theory argues that agencies can decide major questions even when statutes ban them from doing so.

Suppose Congress enacted a law saying “the Department of Education is not allowed to cancel student loan debt under this statute,” and the Department then argued: “too bad, we get to decide how we’re going to execute the law, so we’re just not going to enforce the requirement to pay back the principal on the loans anymore.” This argument would be in some tension with the result in Biden v. Nebraska, yet that is the logical implication of the right-to-impound theory.

I understand that attitudes towards the separation of powers, on both sides of the house, shift upon a change in administration. But up to a point! I can’t understand how, in a world with the major questions doctrine, an agency gets to decide whether major federal programs, costing billions of dollars, will exist.

Now to the good part

I am pretty confident the Supreme Court would uphold the Impoundment Control Act as to most factual scenarios. But it’s always nice to ask someone smarter than you.

I asked Claude to construct a hypothetical impoundment scenario and draft Supreme Court briefs litigating the constitutional dispute. AI loves to hallucinate, so why not let its creative juices flow? You cannot just say “write the briefs” to Claude—yet—but with appropriate hand-holding and strategic prompting, you can, indeed, get it to write the briefs.

Sometimes I hear AI skeptics saying things like “AI is great for simple filings, but in complex cases, where Real Thinking is needed, it’s not helpful.” This is emphatically not the case. AI can understand and propose every argument that might plausibly appear in a Supreme Court brief. I promise you that today’s frontier models, which can ace the Putnam Exam, are perfectly capable of achieving insights such as “the President isn’t Taking Care to enforce the law when he’s not following the law.” Also, there’s a limited body of Supreme Court case law, all of which is known to AI, so AI is less likely to hallucinate in a Supreme Court brief than a lower-court brief (though it does hallucinate sometimes).

Anyway, Claude dreamed up a scenario where President Trump impounded a provision of the “Consolidated Appropriations Act, 2025” which “included $3.8 billion in funding for the Department of Energy’s Office of Energy Efficiency and Renewable Energy (EERE).” This is an amusingly plausible scenario given that the real 2023 Consolidated Appropriations Act appropriated $3.46 billion for that very program. Claude confidently described tweets, congressional hearings, a plausible-sounding plaintiff to challenge the impoundment (“Brightpath Solar Manufacturing”), a district court decision by Judge Alsup enjoining the impoundment, and a Ninth Circuit opinion affirming that decision, complete with a trenchant dissent by Judge VanDyke.

Here are the briefs:

Brief Of Petitioner Trump V

208KB ∙ PDF file

Download

Download

Brief Of Respondent Trump V

207KB ∙ PDF file

Download

Download

Reply Brief Of Petitioner Trump V

135KB ∙ PDF file

Download

Download

I then persuaded Claude to generate an argument transcript, which doesn’t pass the Turing Test but still captures the typical dynamic of a Supreme Court argument quite well:

Oral Argument Transcript Trump V

108KB ∙ PDF file

Download

Download

So how does Claude think the Supreme Court will rule? Drumroll please:

Based on my understanding of the current Supreme Court and these complex legal arguments, I predict that the Court would rule 6-3 in favor of upholding the Impoundment Control Act’s constitutionality, affirming the Ninth Circuit. Here’s my analysis:

The Majority (Roberts, Barrett, Kagan, Sotomayor, Jackson, and likely Kavanaugh):

• Chief Justice Roberts would likely write the majority opinion, as he often does in major separation of powers cases. He has consistently shown concern about maintaining institutional checks and balances and would likely view unilateral presidential impoundment authority as a significant threat to Congress’s constitutional power of the purse.

• Barrett, despite her generally conservative approach, has expressed strong textualist views about separation of powers and would likely be persuaded by the clear text of the Appropriations Clause and the long history of congressional control over spending.

• Kavanaugh, though conservative, tends to take a more moderate approach in separation of powers cases and has shown respect for established statutory frameworks. The ICA’s 50-year history and workable procedures would likely influence him.

• The three liberal justices (Kagan, Sotomayor, Jackson) would almost certainly vote to uphold the ICA based on both statutory and constitutional grounds.

The Dissent (Thomas, Alito, Gorsuch):

• Justice Thomas would likely write a strong dissent arguing for broad executive power under Article II and the Take Care Clause.

• Alito and Gorsuch would join, emphasizing historical examples of presidential impoundment and arguing for inherent executive authority to protect national interests.

Key factors influencing this outcome:

1. The Court’s institutional conservatism and reluctance to dramatically alter long-established separation of powers frameworks

2. The clear constitutional text giving Congress power over appropriations

3. The ICA’s long history of successful implementation across administrations of both parties

4. The availability of alternative mechanisms for presidential influence over spending (like the regular appropriations process)

5. Concern about creating an effective line-item veto power that the Court has previously ruled unconstitutional

The majority opinion would likely emphasize Congress’s fundamental power of the purse while acknowledging some executive discretion in implementation within statutory bounds. The dissent would focus on Article II powers and historical practice supporting broader executive authority.

And here is the Chief Justice’s majority opinion, concurrences by Justices Kagan and Kavanaugh, and Justice Thomas’s dissent. (Claude kept the opinions reasonably short - perhaps for the best!)

Supreme Court Decision Trump V

113KB ∙ PDF file

Download

Download

Note that I did this using a general-purposes, non-fine-tuned model. I didn’t use the Enterprise version of Claude, so I had a small context window and severe rate limits. I also didn’t test OpenAI’s $200-per-month o1-pro model, which I’ve heard is a great lawyer.

Thus, technology that is already available today could have produced better briefs than these—to say nothing of technology that will be available three months from now.

When NEPA was first enacted in 1970, no one thought it would be a big deal. It has turned out to be a big deal. NEPA-mandated environmental impact statements (referred to as EISs) often run hundreds of pages and take years to compile, leading to pipelines, solar energy facilities, and other projects being delayed for years. After a project is approved, opponents can file lawsuits alleging that the EIS isn’t detailed enough, yielding further delays. NEPA’s detractors characterize the law as a millstone around the neck of economic and technological progress.

NEPA law continues to develop. On December 10, 2024, the Supreme Court will hear oral argument in Seven County Infrastructure Coalition v. Eagle County, a case addressing the scope of an agency’s NEPA obligations. Also, on November 12, 2024, the D.C. Circuit decided Marin Audubon Society v. FAA, an important but confusing decision establishing that certain NEPA regulations aren’t binding.

But the truly significant NEPA-related development is the advent of AI. I’m sorry if my AI posts are getting repetitive, but I don’t see how one can have an intelligent discussion about NEPA without noticing that a technology has fallen into our laps that can reduce the time to create written reports by 99.99%.

In this post I will:

• Offer an executive summary of why NEPA is controversial.

• Address what’s at stake in the Eagle County and Marin Audubon Society cases.

• Explain why the “AI Changes Everything” take, while clichéd and annoying with respect to many fields, is correct with respect to NEPA.

Trains vs. sage-grouse

The Seven County Infrastructure Coalition wants to build an 85-mile railway line in an area of northeastern Utah known as the Uinta Basin. The primary freight would be crude oil.

Building this line requires a federal license. The Coalition applied for such a license from the federal Surface Transportation Board. Under NEPA, the Board was required to prepare an EIS.

The Board did that, and it approved the project. But opponents of the railway line—environmental groups and a local government in Colorado—sued the Board, alleging the EIS was flawed. This dispute has now reached the Supreme Court.

Eyeballing the EIS at issue should give you a sense of why NEPA is so controversial. It’s a spectacular document, spanning 3,600 pages. Here’s a passage from the document, pulled at random:

Any of the Action Alternatives would cross greater sage-grouse habitat, including breeding, nesting, brood-rearing, and wintering habitat, and would result in the permanent removal of and temporary disturbance to that habitat (Table 3.4-12 and Table 3.4-13). Disturbed areas in the temporary footprint would be reclaimed and revegetated following construction; however, affected sagebrush habitat in the temporary footprint would take many years to be restored to pre-construction conditions due to the difficulty in reestablishing this type of habitat (Meyer 1992). Greater sage-grouse could also be killed or injured by collisions with construction equipment, workers’ vehicles, and project-related infrastructure (fences and communications towers). Noise from construction equipment and the presence of people in construction areas could displace greater sage-grouse and cause them to disperse into habitat areas further away from the rail line (Appendix J, Bureau of Land Management Greater Sage-Grouse Resource Management Plan Compliance). There are also several greater sage-grouse leks in the vicinity of all three Action Alternatives within the Carbon SGMA (Figure 3.4-1). The habitat removal and noise associated with construction of the proposed rail line could cause greater sage-grouse to avoid or abandon those leks, especially if construction were to take place during the breeding season.

In case you were wondering, Claude advises that “a sage-grouse lek is a traditional courtship display ground where male sage-grouse gather each spring to perform elaborate mating rituals to attract females.”

Let’s start with a thought experiment. Suppose this EIS could be prepared instantly and at no cost. Would we want it?

Sure. Why not? It has two primary benefits:

1. It educates the public. If you have a deep interest in the environmental impact of railways in northeastern Utah, then boy, seeing this document must feel like Christmas morning. I’m not trying to make fun of anyone! People with obsessive, esoteric interests are my people.

2. It forces the agency to consider the environmental impacts of its actions. Personally, I find it hard to get worked up about the sage-grouse leks. Even if their lek happens to fall on the railway line, can’t they find another lek? On the other hand, if there’s some way to build the railway line in a way that helps out the sage-grouse, we should do it. In a world with no trade-offs, it’s better to at least think about sage-grouse mating rituals than to build the railway line, accidentally destroy the leks, and then say, oops.

But trade-offs exist. It took more than two years to get from the Notice of Intent to prepare the EIS to the final EIS, and that is faster than usual. According to a 2018 federal report, the average time from Notice of Intent to final EIS was four and a half years, and a fourth of EISs took six years or more. Six years! Congress recently enacted legislation that tries to improve this situation by imposing deadlines and page limits, but I’ll believe it when I see it.

This might not be problematic if the project could proceed while the EIS is being prepared, but it doesn’t work that way: because the agency is supposed to use the EIS to decide whether to approve the project, the project can’t proceed until the EIS is done.

As a result, NEPA has several drawbacks.

1. 4.5 years is way too long to be pondering whether a new project is worth the environmental harm. Life is short.

2. NEPA deters worthwhile projects from being built. It’s hard to raise money for a project when construction won’t begin for half a decade or longer.

3. Even if the project is ultimately built, delay is still bad. If a project is worthwhile, we’re better with it now than five years from now.

4. More fundamentally, NEPA reflects a policy of status quo bias. You never have to prepare a 3,600-page document assessing the benefits and risks of doing nothing.

5. Sometimes federal projects, or projects that require federal licenses, make the environment better. Think of solar or wind projects, or Forest Service initiatives that reduce the risk of forest fires. Deterring or delaying these projects in the name of environmental protection might make the environment worse.

6. NEPA disproportionately affects clean energy projects. The oil and gas industry persuaded Congress to enact various exemptions from NEPA for oil and gas projects. No such exemptions exist for solar and wind projects.

7. EISs cost millions of dollars.

8. There’s also the opportunity cost to consider. The Eagle County EIS is amazingly thorough. It is obvious from reading it that the people who created it are competent, conscientious public servants. Can’t they spend their time on other things?

NEPA has drawbacks, but what about its benefits? Well, one important thing to remember is that NEPA—or at least the portion of NEPA that is judicially enforceable—is a purely procedural law. NEPA is not like the Clean Water Act or the Endangered Species Act, which enact substantive environmental protections. It merely requires the agency to write things down. If an agency wants to turn Yellowstone National Park into a nuclear waste dump, and it explains this decision thoroughly, NEPA will not stand in the way.

So NEPA doesn’t, or at least doesn’t directly, protect the environment. Instead, NEPA’s ostensible goals are to educate the public and promote reasoned decision-making. But it’s not clear that NEPA is effective at accomplishing those goals.

1. Again: look at the Eagle County EIS. Although some people might be interested in particular aspects of this document, approximately no one will read it from cover to cover. A document no one will read shouldn’t be written.

2. There’s no way the agency decision-makers will read this entire document. They’ll read a five-page executive summary. The full document isn’t helping them make decisions.

3. I understand the theory that NEPA serves a kind of project management function—the best way to ensure the agency makes good decisions is to force the agency to write everything down. Realistically, however, few of the considerations addressed in the EIS have any prospect of affecting the agency’s final decision. If the goal is to promote good decision-making, the agency would be better off with a shorter, more focused document.

4. Also, no one voluntarily adopts the project management technique of writing 3,600-page reports exhaustively documenting every consideration. Private companies, who have every incentive to make good decisions, certainly don’t. This suggests that there are better ways to promote good decision-making.

5. It’s not clear these reports accurately predict the future. According to a 1986 study of 29 EISs, “only 30% of the impacts were unqualifiedly close to their forecasts, with almost as many rated accurate principally by virtue of the vagueness of the forecasts.” Yikes! Why spend half a decade creating a report containing predictions that won’t even materialize?

Tell me what you want, what you really really want

After an agency approves a project, a project opponent can sue the agency for violating NEPA. Such lawsuits can take several forms:

• If an agency concludes that an action falls within a “categorical exclusion” from NEPA, a plaintiff can sue and say, no, it doesn’t.

• If an agency conducts a preliminary “environmental assessment” and concludes that no full-scale EIS is needed, a plaintiff can sue and say, a full-scale EIS was needed.

• If, as in the Eagle County case, the agency does prepare the full-scale EIS, a plaintiff can sue and say, the EIS isn’t good enough.

Therein lies NEPA’s deepest flaw.

Opponents of federally-approved projects, like everyone else, have every right to challenge agency action in court. I certainly don’t mean to accuse anyone of litigating in bad faith. The root problem with judicial review under NEPA is the misalignment between what NEPA plaintiffs actually want and the relief they seek in court.

NEPA plaintiffs allege that EISs are inadequate and seek court orders directing the agency to improve the EISs. In reality, however, they do not care about the adequacy of EISs. EISs serve the purpose of educating the public, but NEPA plaintiffs do not want to be educated by the agency. To the contrary, they think the agency should be educated by them—typically, they allege they brought relevant information to the agency’s attention that the agency ignored.

EISs also purportedly serve the purpose of facilitating reasoned decision-making, but NEPA plaintiffs do not care about reasoned decision-making in the abstract. Instead, the goal of NEPA plaintiffs is to stop the project. This isn’t intended as a criticism. That must, by law, be their goal. NEPA plaintiffs don’t have standing unless they’re harmed by the project and a court ruling in their favor would redress that harm by stopping the project. This means the courthouse door is closed unless the NEPA plaintiff is trying to stop the project (or at least modify it).

So by design, our legal system authorizes plaintiffs pursuing one goal—stopping the project—to file lawsuits seeking a different goal—an improved EIS. This has several drawbacks.

1. Overlitigation. The vast majority of NEPA lawsuits lose. The court typically holds that the agency’s environmental analysis was good enough. Why do plaintiffs keep bringing these suits? Because of the misalignment between what the plaintiffs want and what they seek in court. The lawsuit might not succeed in yielding an improved EIS, but it will seek in delaying, and potentially stopping, the project.

2. Unnecessarily long EISs. Agencies know they will be sued over the adequacy of EISs and therefore prepare unnecessarily long EISs in order to fend off lawsuits. This extraneous information in EISs is useless along all dimensions. It’s useless to the agency, which won’t use it in decision-making and merely writes it down in anticipation of litigation. It’s also useless to NEPA plaintiffs, because, again, those plaintiffs do not care about the EIS’s adequacy.

3. Depressing litigation. NEPA lawsuits often involve disputes over the adequacy of an agency’s explanation on some particular issue. The plaintiff argues, “you should have given a more thorough explanation of consequence X,” and the agency argues, “our explanation is thorough enough.” It’s bickering over the adequacy of paperwork. In a normal world, the parties could just work it out—the agency could add a few sentences of explanation and everyone could go away happy. But because the lawsuit is a proxy fight over whether the project will proceed at all, the parties engage in bare-knuckles litigation over whether a document should have been longer.

4. The upward spiral. NEPA litigation always goes one way: The plaintiff alleges the document wasn’t detailed enough. No one would have standing to allege the document was too detailed. So all NEPA decisions either (1) leave the status quo intact or (2) hold that more detail is needed. These decisions create a new baseline that agencies must exceed. After 30 years of this, it’s hardly a surprise that we would see a 3,600 page EIS.

I could go on about this, but there’s an abundant literature on NEPA elsewhere. For example, here’s a much more detailed overview of NEPA from the incredible Construction Physics Substack. Instead I’ll turn my attention next to the two big legal NEPA developments—the Supreme Court’s Eagle County case and the D.C. Circuit’s Marin Audubon Society case.

Hypothetical power plants vs. sage-grouse

The Eagle County case will address whether the Surface Transportation Board’s 3,600-page EIS was good enough for government work.

The D.C. Circuit held that the Eagle County EIS was inadequate for two reasons. First, according to the D.C. Circuit, the Board didn’t adequately consider the legal doctrine known as If You Build It, They Will Come. The purpose of the new railway line is to transport crude oil that’s produced in northeastern Utah. If the railway line is built, it will induce developers to build new oil wells, which will affect the environment. Also, the crude oil will be refined, which will also affect the environment.

The Board concluded that it was speculative to try to assess the effects of these hypothetical projects. It also observed that it had “no authority or jurisdiction over development of oil and gas … nor any authority to control or mitigate the impacts of any such development.” The D.C. Circuit held that the Board was required to analyze these upstream and downstream consequences in more detail.

Second, the D.C. Circuit held that the EIS didn’t adequately analyze risks that would arise when trains crossed from the new Utah line into Colorado. Coloradans were worried about train accidents causing oil spills and sparks causing wildfires. The Board did address these issues a little bit, but the D.C. Circuit found that the Board’s analysis was insufficiently thorough.

In the Supreme Court, the railway developers’ argument can be summarized as “Come on!” Is 3,600 pages of environmental analysis really not enough? We’re talking about some new train tracks here. Does the Board seriously have to consider hypothetical oil and gas projects and hypothetical train derailments?

The project opponents respond with their own version of “Come on!” The purpose of this railway line is to stimulate new crude oil production. How can you conduct an intelligent analysis of the environmental consequences of a railway line while ignoring the very projects the line was intended to spur? Also, railway lines are built for trains. Trains sometimes derail. You can’t ignore this reality because it’s “hypothetical.”

Both sides have a point here. The developers’ position makes sense as a policy matter:

• Other agencies regulate railway safety and oil and gas development. Requiring the Board to analyze these indirect consequences is double counting.

• There’s nothing the Board can do to prevent future derailments or train-related fires in Colorado other than deny the permit. It’s not a situation where, e.g., the Board can insist that the developers use environmentally friendly construction techniques. And I don’t think the Board should deny the permit because of hypothetical future derailments and fires in Colorado. Occasional derailments and fires are inevitable consequences of trains, and I find it weird for the Board to say, “we, the national railway regulator, won’t let you build a new railway line because trains might have accidents on other railway lines that already exist.” And if these considerations shouldn’t affect the Board’s decision, the Board shouldn’t have to address them in the EIS.

• Likewise, there’s nothing the Board can do to prevent pollution from future oil and gas production projects other than deny the permit. And again, that’s a bad reason to deny the permit. A future regulator is better positioned to evaluate a specific oil and gas development project. The Board shouldn’t say, “we’ll prevent a developer from building a railway line because we don’t trust future regulators to manage the environmental effects of hypothetical future projects.”

• As a practical matter, it seems unlikely that the Board’s decision will change if it adds more paper on these topics when it’s already said that they won’t make a difference. So why are we doing this?

• In terms of the public notice function of an EIS, is it really useful for the agency to say, “we’re not an oil and gas regulator and we have no idea what projects will be pursued or approved, but here is an arbitrary estimate with massive error bars”?

• 3,600 pages is long enough!

On the other side of the ledger, none of this is grounded in law. NEPA doesn’t have a “the report is too long already and this seems unnecessary” exception. Also, in recent years, the Supreme Court has repeatedly endorsed muscular judicial review of agency decision-making, including in several cases involving industry groups’ challenges to environmental regulations. Now that the plaintiff is an environmental group, it seems like a bait and switch for the Court to suddenly hear the gospel of deference to administrative agencies.

On balance, I think the developers’ position will carry the day because it makes more sense. But that holding is a legal kludge that won’t solve the problem of 3,600-page EISs. If you want to change NEPA, you have to think bigger.

It’s optional to do what’s optional.

The second recent NEPA development is Marin Audubon Society v. FAA, which is sort of the ultimate Washingtonian decision.

The Marin Audubon Society case doesn’t answer the question, “how much should we protect the environment?” Nor does it answer the question, “how much paperwork should we create in deciding how much to protect the environment?” Instead, it answers the following question: “which federal bureaucrat should decide how much paperwork we should create in deciding how much to protect the environment?” Depressingly, this is a high-stakes question.

Tourists enjoy flying over national parks. In 2000, Congress enacted a law requiring vendors who want to conduct commercial air tours over national parks to get a permit from the Federal Aviation Administration (FAA). The law provided that before issuing such a permit, the FAA, in cooperation with the National Park Service (NPS), must come up with an air tour management plan—which requires NEPA compliance. To speed things up, Congress mandated that the FAA “make every effort” to act on permit applications within two years.

This didn’t work out. The permit application process deteriorated into bureaucratic chaos that was remarkable even by the low standards of the federal government. Among other things, the FAA and NPS bickered over who would conduct particular NEPA determinations. Twelve years later, there were still no management plans. In 2012, to speed things up, Congress enacted a new law authorizing, in some cases, “voluntary agreements” in lieu of management plans. The agencies were able to achieve some voluntary agreements, but this process also ended up being hideously slow and dysfunctional. Nineteen years after Congress enacted the original law with the two-year deadline, essentially no progress had been made. So the D.C. Circuit issued a writ of mandamus—basically, a court order telling the agencies to do their job.

The FAA and NPS ultimately prepared an air tour management plan for purposes of flights over areas near San Francisco such as Point Reyes National Seashore. But they decided they didn’t need to prepare an environmental analysis under NEPA because the environmental impact would be minimal.

How did they make this decision? Well, there’s a federal agency known as the Council of Environmental Quality, or CEQ, that issues government-wide rules on when environmental assessments do and don’t have to be prepared. The FAA and NPS concluded that, under the CEQ’s rules, an environmental assessment was unnecessary.

As it turns out, for reasons I won’t get into, the FAA and NPS’s decision that an environmental assessment was unnecessary was itself nonsensical. The D.C. Circuit had no choice but to reverse that decision and send it back to the agency. To sum things up so far: after nearly 20 years of irrational dysfunction and paralysis, including disputes over who would handle NEPA analysis, the agencies finally issued a plan that didn’t include an environmental assessment for reasons that were also irrational.

But, in the course of reaching that conclusion, a two-judge majority of the D.C. Circuit panel also held that the CEQ regulations weren’t even binding, because the CEQ didn’t have statutory authority to issue them. For decades, everyone assumed the CEQ did have this authority, but the D.C. Circuit took a fresh look and decided that it didn’t.

This doesn’t mean that NEPA goes away. The statute itself, in its pristine, Doric glory, is still on the books. And this doesn’t mean that NEPA regulations go away. Federal agencies with rulemaking authority, like the Department of Interior, are still allowed to write their own NEPA regulations. The D.C. Circuit’s holding was that the CEQ—the centralized agency that has been writing government-wide regulations on this issue—doesn’t have this power.

As Chief Judge Srinivasan’s dissent persuasively explains, the panel shouldn’t have decided that issue:

• None of the parties asked the court to decide this issue. Courts are supposed to resolve issues presented by the parties.

• The court vacated the plan anyway because of the agency’s nonsensical reasoning, so the additional holding was unnecessary to decide the case.

• The CEQ regulation at issue created an optional exemption from NEPA. So the court held, “a CEQ regulation that says that an environmental assessment is optional is itself optional.” This was an unnecessary holding.

This decision was so weird that I predict it will be vacated by the en banc court. Still, the cat is out of the bag. The decision will show up on Google even after it’s been vacated, and eventually a court will decide the issue in a case where it’s properly raised.

I don’t have a strong view on whether the CEQ currently has the authority to issue NEPA rules that bind the executive branch. The CEQ should have the authority, though. Someone has to decide what NEPA means. I see three options:

• No NEPA regulations exist, so agencies fumble around in the dark and judges tell them what to do. The result will be, as Judge Friendly poetically put it, a “judicial oak” that grows from a “legislative acorn.”

• Each federal agency creates its own NEPA requirements, yielding numerous inconsistent requirements and World War III when two agencies need to collaborate.

• A single, centralized federal agency decides what NEPA requires.

Personally, I am staunchly anti-oak. I also think it’s better to have one set of NEPA regulations than ten. So by process of elimination, the CEQ should write the rules. I understand some people think that the CEQ has imposed too many procedural requirements and wiping out those rules will get rid of red tape. Maybe that’s true, although some CEQ rules—such as the one at issue in the Marin Audubon Society case —are designed to get rid of red tape. But if wiping out CEQ rules will prompt each agency to write their own rules—or, worse, let judges decide in each case when an EIS is required without any regulatory guidance—then boy, be careful what you wish for.

Claude, write me a 3,600 page EIS

Eagle County and Marin Audubon Society are important, but not transformational, cases. By contrast, AI’s impact on NEPA is transformational. Quite simply, AI breaks NEPA.

NEPA requires agencies to produce detailed reports. It doesn’t matter what conclusions the reports reach, as long as they’re sufficiently thorough.

In the past, it was impossible to prepare a thorough EIS without extreme human effort, which at least was circumstantial evidence that the agency took a hard look at environmental issues in making its licensing decision. That is no longer true.

It is an understatement to say that AI helps write EISs. To illustrate the point, I asked Claude this query:

Suppose I am writing an environmental impact statement about a new railway line in the Uinta Valley. I need you to write me the table of contents of that environmental impact statement. It should be highly detailed and granular, with at least 50 different topics. Make sure all topics that reasonably should be included in an environmental impact statement are covered.

Here’s the answer, which I’ll put in a document to avoid email length limits.

Uinta Basin Railway Environmental Impact Statement Table Of Contents

24.9KB ∙ PDF file

Download

Download

You can then ask Claude to fill in each category at any level of granularity. For example, here’s what Claude says about the sage-grouse:

Claude On Sage Grouse

20.7KB ∙ PDF file

Download

Download

This analysis would withstand judicial review. Judges are ex-lawyers, not environmental experts, and they have no ability to evaluate the correctness of statements in an EIS. Instead, judges merely check whether the reports seem rational and address every point raised by commenters. I promise you that a judge who read those paragraphs would say that the agency adequately considered the sage-grouse.

Claude cannot generate an entire EIS this way. The agency puts a tremendous amount of work into the EIS, including doing its own fieldwork. We have a long way to go before AI-powered robots handle NEPA fieldwork. Still, the agency prepares a lot of the EIS based on its review of documents collected from the applicant and other agencies. These documents can be uploaded into Claude, and Claude can instantly evaluate them and prepare an EIS.

The upshot is that AI can be used to prepare a plausible-seeming EIS, of arbitrary length, detailed enough to withstand judicial review, in a small amount of time—without the agency having actually taken a hard look at environmental impacts.

NEPA detractors may think this is great, but I don’t. I would still like to live in a world where agencies try to mitigate environmental harms as best they can, albeit without spending years preparing reports that span thousands of pages. When I read the Eagle County EIS, I am stunned that the document exists, but also strangely inspired by the competence, knowledge, hard work, and passion about the environment that it memorializes. I don’t want those capabilities going away.

Given that EISs no longer serve their intended purpose in the age of AI, I think a grand bargain is needed, in which current procedural requirements are relaxed in exchange for alternative AI-driven techniques for ensuring that the agency takes a hard look at mechanisms for mitigating environmental harm. Some ideas for such a grand bargain would be as follows.

First, make EISs dynamic. There is no longer any need for a PDF containing a long EIS. The justification for the document—that its very existence shows that the agency took a hard look at environmental issues—no longer holds. And it’s not a useful document for the public. To lay citizens, it’s too long. But it covers so much ground that it can’t spend more than a couple of pages on any particular issue, so for experts interested in particular issues, it’s too short. Rather than waste time and effort writing this document that no one will read, the agency can take all of its raw materials and put it into a sufficiently large context window. People interested in the project can ask whatever questions they want. If they want a general overview, they can ask for that. If they want a detailed analysis of surface water data or whatever, they can ask for that too.

Is this permissible under current law? I think so: I’m comfortable saying that a dynamic EIS satisfies NEPA’s requirement of a “detailed statement.” If not, have AI prepare an executive summary of whatever length makes the courts happy.

Second, severely curtail judicial review. Judicial review in NEPA cases frequently (not always) consists of judges making sure that agencies addressed all comments and gave rational explanations. It’s not really law, it’s more like “did the agency did a minimally competent job at explaining its reasoning?”

AI has its plusses and minuses, but one thing it can definitely do is offer minimally competent explanations. Grinding through thousands of comments and writing a few paragraphs of prose in response to each one is right in the sweet spot of AI’s capabilities. Judicial review of the AI’s work isn’t much of a value-add. Remember, from the court’s perspective, it doesn’t matter if the prose is right or wrong, it only matters that it’s there. And if suggestion #1 is adopted—a switch from static to dynamic EISs—the concept of “check to see whether the agency responded rationally to every comment” becomes irrelevant.

Given the problems with judicial review under NEPA, why not wave a magic wand and make it go away? As long as the agency certifies that it followed an appropriate AI-driven project management program, of which more below, the actual substance of the EIS could be unreviewable. I think this, too, would be permissible under existing law, especially if the agency produces a dynamic EIS that would be capable of offering far more information than EISs in the status quo.

What about the argument that “it’s dystopian to replace judges with AI?” I think that’s always a bad argument, but it’s an especially bad argument here. I recently re-watched Mad Max: Fury Road, a dystopian classic, and the absence of judicial review under NEPA was not a plot point. I can also confirm that no-judicial-review-under-NEPA did not drive the plot in Children of Men.

It is simply not dystopian to replace judges with AI in NEPA cases. I get the intuition that it’s dystopian to have an AI judge preside over a proceeding where due process interests are at stake. But there’s no due process right for a project opponent to have a judge evaluate an environmental impact statement. Instead, judicial review under NEPA serves the practical purpose of ensuring that the agency follows the law. If we can replace judicial review with a more efficient way of achieving that practical purpose, we should.

Third, use AI as a project manager. One of the goals of NEPA review is to encourage the agency to propose and evaluate measures to mitigate environmental harm. The ability to rapidly propose and evaluate such measures is also a strength of AI. If we trade “approve the project, but endure 3,600-page EISs and years of strategic lawsuits” for “approve the project quickly with as many or more mitigation measures than before,” this strikes me as a win-win.

This sounds hand-wavy. But one of the nice things about AI is that AI is excellent at both (a) designing project management programs and (b) understanding its own capabilities. If you want to figure out how to deploy AI effectively in NEPA review, by far the most effective way is to simply ask AI. I’ll again spare you long block-quotes.

Ai Driven Nepa Compliance

60.6KB ∙ PDF file

Download

Download

In my view, AI puts a world in which there’s both less red tape and more environmentally-friendly development within our reach.

The Tax Day non-paradox

Under Mississippi law, mail ballots are counted if they are “postmarked on or before the date of the election and received by the registrar no more than five (5) business days after the election.” The question in Wetzel is whether, as applied to federal elections, federal law preempts that Mississippi law. Specifically, the question is whether federal law imposes a mandatory requirement that ballots be received by election day, which supersedes the state law authorizing ballots to be counted as long as they are mailed by election day.

In answering that question, it’s easier to analyze the congressional and presidential elections separately. Let’s start with Congress. The Elections Clause, in Article I of the Constitution, provides: “The Times, Places, and Manner of holding Elections for Senators and Representatives, shall be prescribed in each State by the Legislature thereof; but the Congress may at any time by Law make or alter such Regulations.” This means that, by default, states prescribe rules governing elections. But Congress also has the authority to prescribe such rules, and if state law and federal law are inconsistent, federal law wins. In the real world, federal law is silent on most matters of election administration, so state law mostly rules. But there are a few issues on which Congress has legislated, and on those issues, federal law controls.

The federal statute at issue is 2 U.S.C. § 7, which says: “The Tuesday next after the 1st Monday in November, in every even numbered year, is established as the day for the election.” Is Mississippi’s statute, which permits ballots to be counted as long as they are mailed by election day, inconsistent with this federal statute?

It is not. The federal statute establishes a particular day as the “day for the election,” but is silent on what a voter must do to meet the “day for the election” deadline. When a voter votes by mail, several things happen: (1) the voter puts the vote in the mail; (2) the mailman delivers the ballot to a state building; (3) the ballot is processed; (4) the ballot is counted; (5) all the votes are tallied; (6) canvassing activities take place; and (7) the final result of the election is certified. (This is an oversimplified account and I’m skipping lots of intermediary steps.)

Which of these things has to happen on the “day for the election”? The federal statute doesn’t say. You can stare at the words “day for the election” all day, and you won’t find the answer. Therefore, the states get to decide.

Moreover, it’s common for laws to provide that: (1) Date X is a deadline to convey something, but (2) you meet that deadline by putting the document in the mail by the deadline. (This is known as the “mailbox rule.”) For example, tax returns are timely if they are mailed by April 15, and a prisoner’s notice of appeal is timely if he puts it in the mail by the deadline. The mailbox rule comes up outside the context of government deadlines, too: in contract law, a person is deemed to have accepted a contract the instant he puts the signed contract into the mailbox, even if the contract doesn’t reach the offeror until later.

Consider the following two statements:

• April 15 is Tax Day.

• A tax return is timely if it is mailed by April 15.

Do those two statements contradict each other? Is uttering those two statements analogous to saying, “John is married to Mary but Mary is not married to John”? Obviously not. The phrase “April 15 is Tax Day” does not say what, precisely, one has to do by Tax Day to comply with the deadline. Federal law fills in this gap by clarifying that if you’ve put your tax return in the mail by Tax Day, you’re good. The rule that a tax return is timely if mailed by Tax Day is perfectly consistent with the concept of “Tax Day.”

So, just as the statements “(1) April 15 is established as Tax Day, and (2) you meet the deadline by mailing your taxes by Tax Day” aren’t some kind of infernal self-contradictory paradox, the statements “(1) November 5 is established as the day for the election, and (2) you meet that deadline by mailing your ballot by the day for the election” also aren’t some kind of infernal self-contradictory paradox. The second statement just explains how you meet the deadline prescribed by the first statement.

The Fifth Circuit’s decision, of which more below, argues that the mailbox rule is irrelevant because “voting is not a contract or tax return.” Agreed, voting is not a contract or tax return. The point is that—as the contract and tax examples illustrate—the mailbox rule is entirely consistent with the concept of a deadline. It’s simply a rule governing how you meet a deadline.

That wraps up the analysis for congressional elections, so let’s turn to presidential elections. By default, state law governs the administration of presidential elections: the Electors Clause, in Article II of the Constitution, provides that “Each State shall appoint, in such Manner as the Legislature thereof may direct, a Number of Electors” calculated in a particular way. But Article II further provides that “the Congress may determine the Time of chusing the Electors.” Exercising that authority, Congress enacted 3 U.S.C. § 1, which provides: “The electors of President and Vice President shall be appointed, in each State, on election day, in accordance with the laws of the State enacted prior to election day.” The question: does Mississippi’s statute conflict with that federal statute?

Well, as a first cut, the federal statute can’t possibly mean that the electors are literally appointed on “election day.” The federal statute isn’t saying that, at 11:59 PM, even if the vote-counters are halfway through the count, state officials have to torch the uncounted ballots and hastily appoint electors corresponding to whichever presidential candidate happens to be in the lead. Inevitably, it will take election officials a few days or weeks to count up all the votes and figure out who won.

But if that’s the case, then what, precisely, has to happen “on election day”? Is it the voter-mailman handoff? The mailman-mailroom handoff? The mailroom-processor handoff? The opening of the envelope? The counting of the ballot? The statute is silent on this topic. It just uses the words, “on election day.”

When the statute is silent, the state gets to decide. Here, Mississippi has decided that it’s the voter-mailman handoff, which is a ubiquitous way of determining whether a person meets a deadline. There’s no contradiction between federal and state law.

It’s obviously science fiction.

The Fifth Circuit concludes otherwise. Let’s take a look at its reasoning.

The Fifth Circuit first quotes this sentence from Foster v. Love, 522 U.S. 67 (1997): “When the federal statutes speak of ‘the election’ of a Senator or Representative, they plainly refer to the combined actions of voters and officials meant to make a final selection of an officeholder.”

I have no quarrel with this general statement, but it doesn’t shed any light on the question presented. Foster held that Louisiana’s open-primary system, under which candidates could be elected to Congress a month before election day, conflicted with the federal statute requiring the election to occur on election day. Foster has nothing to do with mail ballots or the timing of receipt of ballots. Indeed, Foster has nothing to do with compliance with the Election Day deadline: it instead addressed what states could do before Election Day.

The Fifth Circuit then offers this analysis:

The State’s problem is that it thinks a ballot can be “cast” before it is received. What if a State changes its law to allow voters to mark their ballots and place them in a drawer? Or what if a State allowed a voter to mark a ballot and then post a picture on social media? The hypotheticals are obviously absurd. But it should be equally obvious that a ballot is “cast” when the State takes custody of it.

This is roughly like the following reasoning:

• Is Star Wars a fantasy movie or a science fiction movie?

  • It is obviously absurd to say that Star Wars is a documentary.

  • It is obviously absurd to say that Star Wars is a musical.

  • Therefore, it is equally obvious that Star Wars is a fantasy movie.

The fact that two straw men propositions are wrong does not mean that the court’s view is “obvious.” It’s obvious that putting an image of a ballot on Facebook and stuffing the ballot into a drawer isn’t “casting” a ballot. A ballot can’t be counted unless it’s sent to the government. However, this does not resolve the question whether timeliness is governed by the voter-mailman transfer or the mailman-mailroom transfer.

Also, to me anyway, it sounds more natural to say “the ballot is cast when the vote is put in the mail” than “the ballot is cast when the mailman delivers the ballot to the mailroom.” “Casting” a ballot is something done by the voter, not the mailman, so it’s natural to say that a voter “casts” a ballot when the voter gives it to the government—i.e., the postal service—rather than when one government employee gives the ballot to a different government employee.

Next, the Fifth Circuit offers the following quotation from Newberry v. United States, 256 U.S. 232 (1921): “the word [election] now has the same general significance as it did when the Constitution came into existence—final choice of an officer by the duly qualified electors.” But Newberry addressed the unrelated topic of congressional authority to regulate primary elections. The quoted statement was describing the difference between an election and a primary. It doesn’t shed light on the cast-versus-received issue.

Based on this 103-year-old dictum, the Fifth Circuit declares: “it makes no sense to say the electorate as a whole has made an election and finally chosen the winner before all voters’ selections are received.” Well, I agree it doesn’t make sense to say that elections officials have “finally chosen the winner before all voters’ selections are received,” but elections officials do not ever “finally choose a winner” on election day. The votes have to be counted and canvassed. The court acknowledges this point but says that “[e]ven if the ballots have not been counted, the result is fixed when all of the ballots are received and the proverbial ballot box is closed.” I don’t get it. Why not say that “the result is fixed” when the ballots are put in the mail, as opposed to sitting in a mailbag in the basement of a government building?

Anyway, the federal statutes at issue do not say that elections officials must “finally choose a winner” on election day, that “the result is fixed” on election day, or any of the other verbal formulations that appear in the opinion. They just say “election day.”

The court cites a Mississippi regulation stating that a vote is “final” once it is “marked accepted.” Therefore, the argument goes, because a vote has to be “final” by election day, it has to be “marked accepted” by election day. The significance of this Mississippi regulation eludes me. A Mississippi regulation that uses the word “final” does not resolve whether a Mississippi statute conflicts with a federal statute. I am also unmoved by the court’s citation of a Montana Supreme Court decision from 1944 which largely turned on state law.

The court suggests that it might be possible after election day for voters to recall mail, citing a provision of the “Domestic Mail Manual,” a document which I have been blessed not to have encountered until this case. I will leave it to the subject-matter experts as to whether it is even theoretically possible to recall a mail ballot after if it is mailed. Anyway, there’s no record that this has ever happened or that anyone has even tried. Even if someone did, it’s not a basis to put all mail ballots received after election day into the shredder.

The court cites language from Foster stating that an election “may not be consummated prior to federal election day,” and distinguishes prior case law authorizing early voting on the ground that “so long as the State continued to receive ballots, the election was ongoing and had not been consummated.” The court then states: “Thus, the election is consummated when the last ballot is received and the ballot box is closed.” I don’t understand the “Thus.” How does the proposition “in-person voting before election day is OK” imply the proposition “ballots cast before election day, but received after election day, must be discarded?”

Musty caverns of legislative history

“History,” the Fifth Circuit states, “confirms that ‘election’ includes both ballot casting and ballot receipt.” That is not the case.

The court theorizes that states have historically required ballots to be received by election day, which implies that federal law also imposes this requirement. I’m skeptical of this inference. If the court’s premise is correct, and states did historically require ballots to be received by election day, that wouldn’t imply that states interpreted federal law to mandate that outcome—they may have just felt that such statutes were a good idea.

In fact, however, the premise is wrong. There are many historical examples of states counting votes as long as they were cast by election day, even if they were received by state officials thereafter.

For example, in most states, Civil War soldiers were permitted to cast votes from the battlefield at any time up until election day. State laws contemplated that those ballots might be received by elections officials after election day: for instance, a Maryland law required the governor to “wait for fifteen days” to “allow the returns of the soldiers’ vote.” This seems to rebut the court’s suggestion that the phrase “election day” requires ballots to be received by election day.

To overcome this problem, the court offers the creative solution of defining the persons to whom Civil War soldiers handed their ballots as “elections officials.” I’m not sure that’s historically correct: the government’s brief cites Civil War-era laws in which soldiers were directed to hand their ballots to their commanding officers, not “elections officials.” Moreover, the practical reality was that it might take a long time for those votes to wind their way back to the actual elections officials in the soldiers’ home states who were responsible for counting the ballots. I am skeptical that anyone at the time thought it relevant whether the person physically collecting the ballot on the battlefield could be deemed a “state elections official.”

Later, the practice of counting ballots that were mailed before but received after election day became fairly common. As far back as the 1920s, California and Kansas were already counting such ballots. By the 1940s, eight states—California, Kansas, Maryland, Missouri, Nebraska, Pennsylvania, Rhode Island, and Washington—counted late-received ballots, in some cases for servicemembers and in some cases also by civilians.

The court doesn’t address these statutes. Instead it says this:

By 1938, 42 States permitted some form of absentee voting. Paul G. Steinbicker, Absentee Voting in the United States, 32 Am. Pol. Sci. Rev. 898, 898–99 (1938). But it was almost impossible to count a ballot received after Election Day. All but one of the 42 absentee voting States also had time limits for ballot receipt, with the “usual requirement” of Election Day. Id. at 905–06.

The court’s statement that “it was almost impossible to count a ballot received after Election Day” is not supported by any authority and is incorrect. Although it is true that “[a]ll but one of the 42 absentee voting States also had time limits for ballot receipt,” Mississippi also has a time limit for ballot receipt—five days after the election. The question isn’t whether time limits exist for ballot receipt, it’s whether Election Day is the limit. I took the liberty of pulling the 1938 article that is cited in this discussion, and it does not support the proposition that “it was almost impossible to count a ballot received after Election Day.” Here’s the full quotation from that article:

It should also be noted that all the forty-one states concerned have time limits within which the ballot must be received in order to be counted.  These limits range from six days before to six days after the date of election.  The usual requirement, however, is that the ballot be received on or before the day of election.

Then, a few lines down, there’s a footnote stating that in “those states where the time limit extends beyond the day of election,” “the absentee votes are counted centrally and added to the totals already published for the respective precincts.” There’s no indication that counting those ballots would be “almost impossible.” 

After World War II, state law ebbed and flowed on this issue. The court states that “by 1977, only two of the 48 States permitting absentee voting counted ballots received after Election Day.” The document cited in support of this proposition appears in a stalactite-filled legislative history cavern, specifically a citation-free appendix to a witness’s statement at a 47-year-old congressional hearing. Even assuming the document is accurate, it would confirm that legislation has continuously remained on the books for a century permitting late-received ballots to be counted, at least in some states.

Today, such statutes are common—either 18 or 29 states, depending on how you count, accept ballots that are mailed before, but received after, election day. The court tries to frame these statutes as newfangled innovations, but they’re not very different from the statutes that California and Kansas enacted 100 years ago. One should be skeptical of judges discovering new interpretations of federal law that would imply that state statutes have been illegal for the last century.

Finally, I heartily endorse this well-written analysis by Richard Bernstein. His thesis is that under the Constitution and federal law dating back to 1792, presidential electors meet their deadline as long as they put their vote in the mail by the deadline, even if the votes reach Washington, DC thereafter. This, he explains, is inconsistent with the Fifth Circuit’s view that an “election” is synonymous with the receipt of votes. The Fifth Circuit didn’t address this point (I’m not sure if the parties raised it), but I find it persuasive.

Harmony

Another point in Mississippi’s favor is that other, more recently-enacted federal statutes conflict with the Fifth Circuit’s interpretation of “election day.” For example, the Military and Overseas Voter Empowerment Act requires states to accept overseas servicemembers’ ballots by the “date by which an absentee ballot must be received in order to be counted in the election”—language that is hard to reconcile with a uniform federal requirement that ballots be received by election day. As another example, the Help America Vote Act in some cases requires states to issue provisional ballots that will not be verified until after election day. That statute is irreconcilable with the Fifth Circuit’s view that the election is “consummated” on election day “because officials know there are X ballots to count, and they know there are X ballots to count because the proverbial ballot box is closed.”

Why should we care about these more recently-enacted federal statutes? Well, I’ll include a block-quote from a Supreme Court case:

When confronted with two Acts of Congress allegedly touching on the same topic, this Court is not at liberty to pick and choose among congressional enactments and must instead strive to give effect to both. A party seeking to suggest that two statutes cannot be harmonized, and that one displaces the other, bears the heavy burden of showing a clearly expressed congressional intention that such a result should follow. The intention must be clear and manifest. And in approaching a claimed conflict, we come armed with the strong presumption that repeals by implication are disfavored and that Congress will specifically address preexisting law when it wishes to suspend its normal operations in a later statute.

This principle supports upholding Mississippi’s statute. We must choose between these two propositions:

1. The recent federal statutes are in harmony with the pre-existing laws referring to “election day.” There was no pre-existing requirement that ballots be received by election day, so the new statutes complement—rather than contradict—pre-existing law.

2. The recent federal statutes aren’t in harmony with the pre-existing laws referring to “election day.” There was a pre-existing requirement that ballots be received by election day, so the new statutes contradict that pre-existing requirement.

Harmony is better than dissonance, so we should choose proposition #1—there was no pre-existing requirement that ballots be received by election day.

The Fifth Circuit includes that same block-quote in its opinion but nonetheless concludes that the block-quote supports its interpretation of “election day.” How?

The court begins its analysis by assuming the conclusion that there’s a pre-existing federal rule that ballots must be received by election day. Treating that assumption as a fixed point, the court states that the two options are: (1) the recent federal statutes are a narrow exception to this pre-existing rule, and (2) the recent federal statutes are a broad exception to this pre-existing rule. And, having set up the options this way, the court concludes #1 is preferable to #2:

The United States as amicus cannot come close to showing that HAVA displaces or impliedly repeals the longstanding general rule that the federal Election Day is the singular day on which the ballot box closes. Rather, the best way to harmonize HAVA with the other statutes governing the federal Election Day is that the former is a narrow exception that authorizes States to receive a certain small number of provisional ballots after Election Day from potentially unqualified voters.

I will leave the task of determining the flaw in this logic as an exercise for the reader.

The horseshoe is back

Last year, I wrote a post comparing a Fifth Circuit decision to the types of decisions we used to see from the old Ninth Circuit. The Fifth Circuit is very conservative and the old Ninth Circuit was very progressive, but their methodologies converged.

This case is another example of the same phenomenon. The Fifth Circuit gazed deep into the soul of the phrase “election day” and located, in its penumbras and emanations, a heretofore unnoticed requirement that ballots must be received by election day. The old Ninth Circuit specialized in massaging federal statutes until these sorts of unexpected interpretations popped out. That court was a strong believer in the principle that federal courts should superintend institutions traditionally run by the states, and would routinely announce creative interpretations of federal law as a means of getting there. Indeed, it is easy to imagine Judge Reinhardt writing an opinion saying that “election day” requires all states to accept ballots that are cast by election day, even if they’re received thereafter. Judge Reinhardt would have offered an encomium to the noble tradition of the mailbox rule, added some heartfelt words about how the wicked Postal Service causes voter disenfranchisement, and therefore concluded that “election day” simply means that votes cast by election day but received thereafter must be accepted in all 50 states.

I was not fond of that methodology then, and I am not fond of it now. “Election day” does not have a soul. If Mississippi wants to accept ballots that arrive after election day, that is up to Mississippi.

In this post I will offer three views on Bufkin:

1. The entire VA claims adjudication process should be replaced by AI.

2. Bufkin turns on the following question: Is it more likely that Congress enacted a useless law or a badly-written law? It’s a close call.

3. In deciding Bufkin, the Court might be tempted to attempt to predict, and then account for, the practical implications of its decision. The Court shouldn’t yield to that temptation. It should ignore the future.

Don’t give up. Don’t ever give up.

The Department of Veterans Affairs, better known as the VA, compensates veterans for service-connected disabilities. To obtain benefits from the VA, the claimant must show that (1) he is, in fact, disabled, and (2) the disability is connected to his military service.

Sometimes it’s unclear whether a veteran is entitled to benefits. Suppose a veteran comes to the VA claiming he has back pain. Maybe his back pain isn’t really disabling; maybe it is disabling but it arises from his car accident after he left the service.

In such cases, Congress has decided that it’s better that an undeserving veteran get benefits than a deserving veteran lose out. Thus, Congress has instructed that the veteran be given the benefit of the doubt. Or, in legalese: “When there is an approximate balance of positive and negative evidence regarding any issue material to the determination of a matter, the Secretary shall give the benefit of the doubt to the claimant.”

Suppose a veteran applies for benefits, and the VA concludes that, even taking into account the benefit-of-the-doubt rule, the veteran isn’t entitled to them. Does the veteran have any recourse?

Of course he does. This is America! In fact, when a veteran seeks benefits, there are actually five (!) layers of review:

1. The veteran first seeks benefits from the local VA regional office.

2. If he loses there, he can then go to the Board of Veterans’ Appeals. “Board of Veterans’ Appeals” sounds sort of like a court, but it’s not—it’s an entity housed within the VA.

3. If he loses there, he can go to U.S. Court of Appeals for Veterans Claims, or the Veterans Court for short. Confusingly, although this court is called a “court,” it is not, in fact, a “court” in the constitutional sense, although it’s more court-like than the Board of Veterans’ Appeals.

4. If he loses there, he can go to the U.S. Court of Appeals for the Federal Circuit, which is a real court.

5. And if he loses in the Federal Circuit, you all know who’s at the top of the judicial food chain.

(This week’s images feature lots of judges hearing the case of a single veteran.)

In Bufkin, Level Five is reviewing Level Four’s decision on what standard Level Three uses to review Level Two’s decisions. This is why it’s great to be a lawyer.

Let’s unpack that. As I mentioned, the VA—that’s Level One and Level Two—gives the claimant the benefit of the doubt. Suppose that even giving the veteran the benefit of the doubt, the VA rules against the veteran. The veteran can appeal to the Veterans Court. The primary question in Bufkin is how the Veterans Court is supposed to review the VA’s benefit-of-the-doubt analysis.

A bit of background is needed here. Before 1988, there was no judicial review of VA benefits denials. If the VA decided that a veteran wasn’t entitled to benefits, then that was just too bad for the veteran. After years of advocacy by veterans’ groups, Congress created the Veterans Court in 1988. Congress decided that the Veterans Court would review the VA’s legal determinations de novo (i.e., without deference), but would review the VA’s factual determinations for clear error (i.e., with deference). That review standard appears in 38 U.S.C. § 7261(a), which is still on the books today.

Fast forward to 2002. Veterans’ groups were dissatisfied with the Veterans Court’s work. Their concern was that the Veterans Court wasn’t taking the benefit-of-the-doubt requirement seriously. The VA would often recite the benefit-of-the-doubt standard and then reflexively deny benefits anyway. And then, on appeal, the Veterans Court would rubber-stamp the VA’s benefit-of-the-doubt analysis.

In 2002, in response to these concerns, Congress enacted 38 U.S.C. § 7261(b)(1). Section 7261(b)(1) says that “in making the determinations under” Section 7261(a), the Veterans Court “shall take due account” of the VA’s application of the benefit-of-the-doubt requirement.

The question in Bufkin is: what does this mean?

The petitioner-veterans (there are two of them, Bufkin and Thornton), argue that under Section 7261(b)(1), the Veterans Court should review the VA’s benefit-of-the-doubt determinations de novo. The government argues that, notwithstanding Section 7261(b)(1), the Veterans Court should review the VA’s benefit-of-the-doubt determinations deferentially.

The end of tradeoffs

Before analyzing this intriguing legal question, I cannot help but talk about AI.

Some of you may have seen those lists on the Internet of “old movies where the problems that generated the plot would be solved by modern technology.” You know … in Die Hard John McClane could have called the police with a cell phone, in the Blair Witch Project they could have used a GPS to get out of the woods, etc. Well, AI would solve all of the problems that generated the plot in Bufkin.

There are three oft-cited problems with the VA benefits process:

1. The VA is extremely slow. Wait times for benefits determinations are staggering.

2. The VA is wildly inconsistent. One examiner might say that a veteran’s symptoms render him 100% disabled, another examiner might say that the same symptoms aren’t disabling at all.

3. Some individual VA examiners are pathologically hostile to veterans and love saying “no.”

The VA has a poor reputation, but it has a tough job. Last year, the Board of Veterans’ Appeals decided over 100,000 appeals. That’s an insane number of appeals.

There’s no easy answer to these problems. To deal with problem #1, you can hire a lot of new examiners to speed things up. But even if the VA had the budget for this, it would exacerbate problems #2 and #3—you’d have more variation and more inept examiners.

To deal with problem #2, you can set detailed rules governing when benefits should be conferred. The VA has certainly done this: the regulations governing veterans’ benefits rival the Theory of General Relativity in complexity. But this only goes so far. For certain types of disabling conditions, PTSD being a notable example, different examiners are inevitably going to interpret the veteran’s symptoms in different ways.

To deal with problem #3, you can do what Congress did: tell the VA to give the benefit of the doubt to the veteran. But personnel is policy. If an examiner is a jerk, you can tell him to bend over backwards for the veteran using whatever verbal formulation you want, and he’ll still stiff the veteran.

Alternatively, the VA could fire or demote the examiners who are too harsh on veterans. But this would create horrible incentives, plus, to first approximation, it’s impossible to fire a government employee. Another option is to have multiple people look at the same benefits denial—which is why we now have the Regional Office, and then the Board of Veterans’ Appeals, and then the Veterans Court. This might mitigate problem #3, but it exacerbates problem #1.

Also, there comes a point where there is too much process. The cases of the two veterans in the Supreme Court case, Bufkin and Thornton, illustrate this point.

Bufkin served in the Air Force for six months. While he was in the Air Force, his wife wanted him to leave the military and threatened to commit suicide if he stayed in. Bufkin was advised, reasonably, that his options were to leave the Air Force based on undue hardship or end the marriage. He chose to leave the Air Force. Seven years later, Bufkin filed a claim for veterans’ benefits, claiming that his unpleasant memories from this experience left him with service-connected PTSD.

This allegation led to multiple reviews by VA examiners, all of which denied his claim; review by the Board of Veterans’ Appeals, which denied his claim; review by the Veterans Court, which denied his claim; and then an appeal to the Federal Circuit and the Supreme Court. This seems like a lot of process for a marginal claim, and now Bufkin wants even more process: he wants the Veterans Court to take a de novo look at the benefit-of-the-doubt issue, which has already been analyzed by multiple VA examiners plus the Board of Veterans’ Appeals.

Thornton served in the military for three years, from 1988 to 1991. In 1994 he was given a 40% disability rating for what the record refers to as an “undiagnosed illness.” 30 years later, he still has that 40% disability rating for that undiagnosed illness. Years after he left the military, he sought benefits for PTSD, and in 2005 he was given a 10% disability rating for PTSD. In 2015—24 years after he left the military—he sought a higher disability rating for PTSD. The VA upped his rating from 10% to 50%. The Board of Veterans’ Appeals then granted his claim for a total disability—effectively 100% disabled—based on the conjunction of his PTSD and the undiagnosed illness.

Thornton appealed to the Veterans Court, claiming that his disability rating for PTSD should have been 70% rather than 50%, which might lead to slightly higher monthly benefits, even though he’s already getting benefits for a total disability. The Veterans Court rejected this argument. In the Supreme Court, Thornton argues that the Veterans Court should have taken a de novo look at whether, under the benefit-of-the-doubt rule, he’s 70% rather than 50% disabled as a result of PTSD. Again, this is a lot of process for a claimant who seems to have been treated favorably by the VA.

Also, why we should think that the Veterans Court would do a better job than the prior two reviewers in deciding whether he’s 70% or 50% disabled? They’re all reviewing the same documents. It’s not as though the Veterans Court judges have higher IQs or are nicer people. Sure, if multiple people review the same ambiguous claim, over and over again, one of them eventually is going to rule for the claimant. So maybe, reviewer #3 is going to think Thornton is 70% disabled from PTSD even though reviewers #1 and #2 thought he was 50% disabled. But it’s not clear to me that adding more reviewers to the same case is going to make determinations more reliable in the aggregate.

As these cases illustrate, adding additional layers of review may reduce the probability that a particular veteran will be unfairly short-changed. But adding layers of review is expensive, slows things down, and might not improve accuracy. Ultimately, designing the optimal scheme for veterans’ benefits adjudications is a tricky policy problem, with tough tradeoffs and no clear answers.

Well, it used to be a tricky policy problem, but it isn’t anymore. Using AI for benefits adjudications obliterates all problems instantly:

• There are no delays. 2-year wait times can be reduced to 2-minute wait times.

• There are no inconsistencies across adjudicators because there’s one adjudicator.

• The veteran won’t face the risk of being stuck with a stingy adjudicator, because again, there is one adjudicator that can be adjusted until it isn’t stingy.

This can be trivially implemented today. Just upload the relevant portion of the VA benefits manual into an AI context window, upload the veteran’s documentary evidence, and ask AI to apply law to fact. If you’re concerned that AI won’t be sufficiently generous, you can keep track of the percentage of veterans that obtain benefits and then tweak the prompts until the percentage is at a satisfactory level. If you’re squeamish about having AI make decisions that affect people’s lives, then you can have the AI complete the initial layer of review (equivalent to what the VA does today) and then give the veteran the right to appeal to a human judge. If we just snap our fingers, we can make all tradeoffs go away.

Is it OK for Congress to do nothing?

For reasons related to political will rather than technology, I am skeptical that the VA will implement this solution any time soon. So let’s go back to the the Bufkin case.

A quick review of the legal issue:

• Section 7261(a) says that the Veterans Court reviews the VA’s factual determinations for clear error.

• Section 7261(b)(1) says that the Veterans Court “shall take due account” of the VA’s application of the benefit-of-the-doubt rule.

• The petitioner-veterans argue that the Veterans Court should review the VA’s benefit-of-the-doubt analysis de novo. The government argues that the Veterans Court should review the VA’s benefit-of-the-doubt analysis for clear error.

The veterans’ argument goes like this: Before Section 7261(b)(1) was enacted, the Veterans Court was reviewing the VA’s benefit-of-the-doubt determinations deferentially. In enacting Section 7261(b)(1), Congress must have changed something. Section 7261(b)(1) can’t possibly mean nothing. So, Section 7261(b)(1) must mean that the Veterans Court should review the VA’s benefit-of-the-doubt determinations de novo.

The government’s argument goes like this: Section 7261(b)(1) doesn’t say that. It just says, the Veterans Court “shall take due account” of the VA’s application of the benefit-of-the-doubt standard. Section 7261(b)(1) is a kind of kick in the pants to the Veterans Court—it says, “do your job” rather than changing the legal standard. So, the Veterans Court’s review of the VA’s benefit-of-the-doubt determinations is still deferential.

This is a tough issue. The government has a reasonable argument based on the literal text of the statute. The statute does not, in fact, say that the Veterans Court should review the VA’s benefit-of-the-doubt analysis de novo. It says “due account,” which sounds like a reference to pre-existing legal principles. Here, pre-existing legal principles suggest that agencies’ factual determinations, including the determination of whether there’s an “approximate balance of positive and negative evidence,” should be reviewed deferentially. I have some technical views on this issue that only a lawyer could endure, so I will stick them in a footnote.1

On the other hand, the veterans are right that Congress doesn’t usually enact laws that accomplish nothing. The government argues that Section 7261(b) did something—it admonished the Veterans Court to follow the law, which it wasn’t previously doing. Still, the veterans are right that Congress doesn’t usually enact new laws that don’t change what the law previously required. So maybe we should construe Section 7261(b) to mean something, even though that “something” isn’t really what Section 7261(b) says.

To me this dispute boils down to how comfortable we are with the idea that Congress sometimes enacts laws that do nothing. Philosophically, I am comfortable with this idea. Congress is routinely criticized for enacting useless, symbolic gestures; is it so bad to declare that Congress enacted a useless, symbolic gesture? This interpretation would ascribe a certain ineptitude to Congress, but the petitioners’ interpretation would ascribe a different type of ineptitude to Congress—that Congress enacted a de novo review scheme using the most oblique language imaginable. In selecting between these ineptitudes, why not err on the side of saying that the inept institution did less rather than more?

On the other hand, it seems really bizarre to say that Congress enacted a statute that does nothing.

For what it’s worth, Claude thinks the veterans are correct with 85% certainty! Claude is deeply committed to the idea that the 2002 amendment couldn’t have been surplusage and feels that the benefit-of-the-doubt analysis is more of a legal than a factual determination. I lean towards trusting AI on this issue, but it’s a close one.

Ignore the future

I’m not sure what the right answer to Bufkin is. But in the last section of this post, I’d like to talk a bit about how to go about deciding a case like Bufkin.

Section 7261(b)(1) is ambiguous. There are two ways we might go about resolving this ambiguity:

1. Dwell on the phrase “due account” until an answer pops out of our head.

2. Resolve the ambiguity based on the anticipated practical consequences of the Court’s ruling.

#2 is tricky because we don’t know what the practical consequences will be. But we could formulate a few hypotheses, such as:

1. A ruling for the veterans will significantly increase the number of veterans who get benefits, which will on average increase the accuracy of benefits adjudications, because many veterans are currently being short-changed.

2. A ruling for the veterans will significantly increase the number of veterans who get benefits, which will on average decrease the accuracy of benefits adjudications, because the Veterans Court is a poor position to conduct de novo review of the VA’s benefits adjudications.

3. A ruling for the veterans will not have a significant impact on the number of veterans who get benefits, because the Veterans Court tries to do justice and isn’t significantly influenced by the exact articulation of the standard of review.

We could then attempt to discern, somehow, which of these predicted outcomes is most likely to materialize, and fold that prediction into our resolution of the ambiguity in Section 7261(b)(1).

In my view, #1 is the correct methodology. Practical implications are irrelevant. Ignore the future. We should go with whatever we think “due account” means, taking due account, as it were, of background legal principles such as “Congress usually intends for statutes not to be useless.”

Why?

First, “practical implications” and “what the text means” are incommensurate. Suppose you think the text slightly weighs in favor of the government, but it’s more likely than not that benefits decisions will be slightly more accurate if the veterans win. Who wins?

There’s nothing intrinsically wrong with weighing incommensurate factors. Judges do this all the time: every time a sentencing judge has to weigh a defendant’s bad criminal history against his sincere apology, he is weighing incommensurate factors. My problem is that when we’re weighing the text against practical implications, I don’t know what we’re trying to maximize. You might say we’re trying to reach the “best” interpretation, but “best” along what dimension?

Second, what happens if the prediction is proved wrong? Suppose the Court rules for the veterans based on its prediction that a pro-veteran decision will increase the number of veterans who obtain benefits. But 10 years down the road, a scholar conducts an empirical study and concludes that the prediction didn’t materialize. It won’t matter: the Court’s decision will stay in place. In principle the Court could overrule its prior decision, but this would be both extremely unusual and also premised on the legal fiction that the decision was always wrong, not just that it was proven wrong. To me it’s fine to adopt a philosophy of “I will stick to my principles, damn the consequences,” and it’s fine to adopt a philosophy of “I will attempt to adopt a practical solution and monitor the consequences, and if the consequences aren’t what I expected, I’ll adjust course.” But it’s strange to adopt a philosophy of “I will rule based on factual predictions, while sticking to my guns regardless of whether the predictions pan out.”

Third, neither lawyers or judges are able to make these predictions accurately. I asked Claude to generate 15 hypotheses about potential practical outcomes of the Court’s decision (this is an arbitrary number, Claude could trivially generate 100 hypotheses). Here they are:

1. The decision will lead to a significant increase in appeals to the Veterans Court, as veterans become aware of the more robust review process.

2. The Veterans Court will face a substantial backlog of cases due to the more time-consuming review process required.

3. The VA will adapt its decision-making process to more explicitly demonstrate its application of the benefit-of-the-doubt rule, leading to more favorable decisions at the initial stage.

4. The decision will result in increased consistency in benefits decisions across different VA regional offices and judges.

5. The cost of veterans’ benefits programs will increase significantly, potentially leading to budget pressures or calls for reform.

6. The decision will lead to a surge in veterans seeking legal representation for their claims, as the importance of framing issues for appeal becomes more apparent.

7. The VA will invest more resources in training its adjudicators on proper application of the benefit-of-the-doubt rule, improving decision quality at the agency level.

8. The decision will result in longer processing times for initial claims, as VA adjudicators become more cautious and thorough in their analysis.

9. There will be an increase in remands from the Veterans Court to the VA for readjudication, rather than outright reversals.

10. The decision will lead to a shift in VA culture towards a more veteran-friendly approach in close cases.

11. The Federal Circuit will see an increase in appeals from VA decisions, as the government challenges expansive interpretations of the benefit-of-the-doubt rule.

12. The decision will result in greater public trust in the VA benefits system, as veterans perceive the process as more fair and thorough.

13. There will be a decrease in the consistency of Veterans Court decisions, as different judges interpret and apply the new standard differently.

14. The decision will lead to an increase in fraudulent claims, as some individuals perceive a greater likelihood of success even with weak evidence.

15. The VA will develop more sophisticated data analytics tools to help identify cases where the benefit-of-the-doubt rule should apply, leading to more accurate initial decisions.

I guess these are all possible outcomes? How exactly are we going to assess their likelihood?

For what it’s worth, you can ask Claude to rank these 15 hypotheses by order of likelihood, and it gives a plausible-sounding answer, but that really starts to look like BS. There’s some empirical evidence that AI is better than humans at making empirical predictions, but AI isn’t some kind of perfect oracle.

It seems like cheating to ask AI to generate 15 hypotheses and then say that it’s too hard to evaluate them all. Why should we care that AI can spit out empirical predictions at will? AI has no empirical basis for thinking that any of these hypotheses has any particular likelihood of being true; it’s just remixing empirical predictions it saw somewhere on the Internet.

But humans also do that. One reason I think AI has such powerful implications for law is that AI learns and reasons a little like how human lawyers do. I’m not referring to the hypothesis that “the extreme and unexpected success of large language models is indirect evidence that human brains’ computations are similar to large language models’ computations.” Rather, I’m observing that human lawyers output legal reasoning by remixing the legal texts on which their minds have been trained—kind of like AI. In writing judicial decisions, judges typically don’t rule based on visual observations or empirical experiments. Instead they evaluate written texts—lawyers’ briefs—based on mental models that have been trained on thousands of prior briefs and cases the judges have read before. Lawyers who write briefs that confidently offer empirical predictions are doing the same thing—they’re making arguments based on their experience reading and writing lots of legal texts containing lots of arguments.

So if you think that speculating about AI-generated predictions is a poor way of deciding cases, you probably don’t want to speculate about human-generated predictions, either.

Like many capital cases, Williams’ case has provoked much debate. Williams’ defenders insist that Missouri unconstitutionally executed an innocent man. Governor Mike Parson claims that there was “overwhelming evidence” of Williams’ guilt.

As I see it, the truth is in between. The evidence of Williams’ guilt was not overwhelming. The state’s case had serious weaknesses, and in my view, there is reasonable doubt as to whether Williams committed the murder. Nonetheless, lingering doubt about guilt isn’t a sufficient basis to overturn a conviction. Ultimately, Williams’ case illustrates that it is impossible to reduce the risk of wrongful convictions to zero, even in capital cases.

The trial

On August 11, 1998, Felicia Gayle, a journalist at the St. Louis Post-Dispatch, was stabbed to death in her home. The murderer stole Gayle’s purse, as well as the laptop of her husband, Dr. Daniel Picus.

Gayle’s murder sparked significant publicity. But for ten months after the murder, the police had no leads. There was abundant physical evidence at the crime scene—a knife, bloody fingerprints, bloody shoeprints—but the police couldn’t get any useful information from it. Dr. Picus offered a $10,000 reward for information that would lead to the murderer’s capture.

Ten months later, on June 4, 1999, the police got its first real clue. A man named Henry Cole, who had just been released from jail, told the police that his cellmate, Marcellus Williams, had confessed the murder to him. Williams had been incarcerated since August 31, 1998, for an unrelated robbery—but on August 11, 1998, the date of the murder, Williams wasn’t in jail. Cole unapologetically told police that his motivation for coming forward was the reward money.

Cole told police that Williams had told one other person about the murder—Williams’ girlfriend, Laura Asaro. Detectives asked Cole to persuade Asaro to incriminate Williams. Cole tried his best, but Asaro refused. Two months later, on August 5, 1999, Asaro was arrested on unrelated charges. Detectives told Asaro that if she cooperated, the charges would be dropped and she might be eligible for the reward money. She still wouldn’t talk.

On November 17, 1999, the police visited Asaro again and threatened to arrest her on outstanding warrants. This time she broke. She told the police that Williams had confessed the crime to her.

The next day, the police searched Williams’ car, which had been parked on the street since his arrest 15 months earlier. A St. Louis Post-Dispatch ruler, calculator, and medical dictionary were found.

Asaro told police that Williams stole Dr. Picus’s laptop and pawned it to a man named Glenn Roberts. The police visited Roberts’ home and found Dr. Picus’s laptop.

Williams was charged with Gayle’s murder. At Williams’ trial, there was no physical evidence against Williams. The bloody shoeprints at the scene didn’t match Williams’ feet. Hairs on the carpet didn’t belong to Williams.

Instead, the case against Williams boiled down to the following evidence:

1. Cole testified that Williams confessed to him.

2. Asaro testified that Williams confessed to her. She also offered other, highly incriminating testimony, such as that she saw Gayle’s ID in Williams’ car after the murder.

3. Gayle’s ID was gone, but a ruler and calculator were found in Williams’ car, and the prosecutor argued that they belonged to Gayle. These items weren’t definitively linked to Gayle, but the ruler said “St. Louis Post-Dispatch” on it and there was evidence that Gayle carried a calculator in her purse. (Dr. Picus confirmed the medical dictionary in the car didn’t belong to him.)

4. Roberts testified that Williams had sold him Dr. Picus’ laptop shortly after the murder, and the police found the laptop in his house.

5. Williams had an extensive criminal history.

Based on this evidence, the jury convicted Williams and sentenced him to death.

At first glance, the prosecution’s case looks strong. But looks can be deceiving. In my view, there’s a reasonable case that Asaro and Cole framed Williams for Gayle’s murder.

Let’s take a closer look at each piece of the state’s case. (Caveat: I do not have the original trial transcript. I only have descriptions of the trial testimony from public filings. If I am wrong about any of this, I am happy to be corrected.)

Cole’s testimony

Cole testified that his cellmate, Williams, confessed to him. If Cole is telling the truth, then Williams is almost certainly guilty. But Cole could be lying.

There are several reasons to question Cole’s veracity.

1. Cole had an extensive criminal history, with felony convictions and prison sentences all over the country. He also was a crack cocaine addict with a history of mental illness.

2. Cole had a motive to lie—he repeatedly told the detectives that he wanted the reward money and even asked for an up-front payment.

3. Williams had no motive to confess to Cole. I realize prisoners confess all the time, but Williams seems like a careful guy—under the prosecutor’s theory of the case, he committed a murder while leaving no physical evidence whatsoever. Why spill his heart to Cole and also lead Cole to Asaro, the one person who had personal knowledge that could incriminate Williams?

4. While I lack personal experience of cellmates confessing brutal stabbings to me, Cole’s description of Williams’ confession does not ring true. It’s very detailed, with information that seems carefully curated to bolster the prosecution’s case. For instance, according to Cole’s testimony, Williams confessed that he took the following items from the scene of the crime: an Apple laptop computer and computer bag, cheap jewelry, a coin purse, an I.D., a wallet, and keys. Is it really plausible that Williams said this to Cole during his jailhouse confession?

That said, it’s clear that Cole wasn’t making this up out of whole cloth. Cole led the police to Asaro, who led the police to the laptop. Cole had no independent connection to Asaro, so he couldn’t have known about Asaro’s connection to the crime unless Williams told him. So Williams told Cole something about the crime. But Williams might have told Cole that someone else —say, an associate of Asaro—committed the crime, and Cole might have altered Williams’ story to incriminate Williams.

At trial, the prosecutor testified that Cole must be credible because, when he originally spoke to the police, he knew facts that weren’t public. This isn’t persuasive for a few reasons:

1. The supposedly non-public information is pretty weak. It’s stuff like “the murderer twisted the knife.” Cole could have guessed it.

2. The initial conversation between Cole and the police, which occurred in a police car, wasn’t recorded. The police could have fed him these details during that conversation.

3. Again, Williams clearly told Cole something about the murder, so the fact that Cole has non-public information about the murder doesn’t prove anything.

Why would Cole have framed Williams? Well, consider matters from Cole’s perspective. Suppose Williams told Cole that a third party committed the crime. Cole wants the $10,000 reward. But if the third party’s not in jail, he might take revenge on Cole if Cole rats on him. Why not alter the story to incriminate a man who is already in jail, and hence not a threat?

Or maybe Cole and Williams had a dispute in prison. There are innumerable reasons Cole might have wanted to frame his ex-cellmate.

Of course, Cole knew that he wouldn’t get the reward unless Asaro corroborated his story. But remember what Cole did right after talking to the police—he volunteered to reach out to Asaro to persuade her to provide information to the police! In a world where Cole is trying to frame Williams, this makes sense—Cole might have wanted to explain to Asaro what he told the police and why it would make sense for her to tell the same story.

Indeed, thinking about this more, the police’s reaction to Cole’s statement was very strange. There’s a high-profile murder. It’s all over the newspapers. For ten months, the police has been investigating, with no leads. Suddenly, Cole comes forward with a detailed confession and non-public information about the crime. Cole pinpoints Asaro as someone who could corroborate his story. Shouldn’t the police have interviewed Asaro immediately? Instead, they ask Cole to be an informant, even though Cole has never met Asaro in his life and is generally a highly sketchy individual. Detectives don’t talk to Asaro until two months go by and Cole has no luck contacting Asaro. What’s going on here? My guess is that Cole enthusiastically volunteered to be an informant and persuaded the police he could get Asaro to cooperate—because he wanted to get to Asaro first.

Asaro’s testimony

Asaro initially didn’t cooperate with police, either. But in November 1999, 15 months after the crime, when the police threatened to arrest her, Asaro finally talked, and later testified at trial.

Asaro testified that she had started dating Williams two or three months before the murder. The day after the murder, Asaro found a woman’s purse in Williams’ car and confronted him, alleging he was cheating. Williams responded that no, the purse belonged to a journalist he had murdered. According to Asaro, Williams also threatened to kill her and her family if she ever told anyone. Asaro gave the police other highly incriminating information, such as that she saw the victim’s ID card in the trunk of Williams’ car.

Like Cole’s testimony, Asaro’s testimony is devastating if you believe her. And Asaro clearly knew something about the murder, given that she directed the police to Glenn Roberts, who had Dr. Picus’s laptop.

But that doesn’t mean that Asaro is telling the truth. She might know the murderer’s identity—but the murderer might be someone else who she’s trying to protect. It’s plausible that Asaro was acquainted with a different violent felon who could have committed the murder. Asaro was a sex worker, addicted to crack cocaine, who was constantly being arrested. One would suspect that not every person in her LinkedIn network had an unblemished resume. So the question is: was Williams the murderer, or was Asaro trying to frame Williams for a murder committed by someone else she knew?

There are several reasons to question Asaro’s testimony.

1. Asaro was constantly being arrested and had a reputation for lying.

2. Asaro’s story is … unusual. She claimed that after she accused Williams of cheating, Williams, to appease her, said that he had actually committed an incredibly heinous stabbing. According to Asaro, Williams then threatened to kill her entire family if she told anyone. Maybe I’m just a naive guy, but is that really better than cheating? Like, did he really expect her to think, “well, if she thinks I’m a cheater she’ll dump me, but if I tell her I’m Jack the Ripper and also threaten to kill her entire family, our love will blossom?”

3. It apparently worked, though! According to the trial testimony, Asaro visited Williams six times in prison after Williams was arrested on August 31. Six times! To review the history of Asaro and Williams’ relationship (according to the prosecution): within two or three months of getting together, Williams (1) committed a brutal murder, (2) threatened to kill Asaro’s entire family, and (3) got arrested on an unrelated robbery charge. Also, (4) there was testimony that Williams had, separate from all of this, shot Asaro’s ex-boyfriend. While I get that the course of true love never did run smooth, shouldn’t Asaro have decided it was time for a new boyfriend? To me, Asaro’s loyalty to Williams implies that Williams was keeping Asaro’s secret—not the other way around.

The items in the car

After interviewing Asaro, the police searched Williams’ car, and discovered a calculator and a ruler that said “St. Louis Post-Dispatch” on it. The prosecution argued that Williams took these items from the crime scene. On the surface, this theory makes sense. Dr. Picus told police that Gayle kept a calculator in her purse; he didn’t report the ruler missing, but Gayle did work for the Post-Dispatch.

At trial, the defense suggested that Asaro might have planted the items in the car. She had the means of doing so—there was evidence that she had access to the car and had retrieved some items from it while Williams was in jail.

On this issue, I think the defense has a plausible argument. The prosecution’s version of events is questionable at best.

First, why would Williams have left these items in the car? According to the prosecution’s theory of the case, Williams fled the scene, disposed of his bloody clothes, disposed of the stolen purse, disposed of the stolen laptop, but left these worthless, yet incredibly incriminating, items in the glove compartment of his car. Why would he do this?

At trial the state theorized these items might have been “trophies,” as though Williams was a serial killer. I doubt that. And if he did take “trophies,” why would those “trophies” just happen to include a ruler with the name of the victim’s employer on it?

Also, even assuming Williams took trophies from the scene of the crime, wouldn’t he have asked Asaro to throw out the trophies after he went to jail? He must have known that his car was just sitting there, waiting to be towed and searched, with evidence in the glove compartment directly tying him to a brutal murder.

And even if he didn’t ask Asaro to throw out the trophies, wouldn’t Asaro have done that on her own? There was trial evidence that Asaro also had access to the car while Williams was in jail. And the trophies put her at risk, too. Why would she keep them in the car? It just doesn’t make sense that this incriminating evidence stays in the car for 15 months.

Finally … and most puzzlingly to me … a medical dictionary was found in the car too, but Dr. Picus confirmed that it wasn’t his. So why is it there? Williams doesn’t strike me as the type who would read medical dictionaries out of academic interest.

It’s plausible, to me at least, that Asaro (1) read in the newspaper that the victim worked for the Post-Dispatch and her husband was a doctor, and (2) put items in the car that, she thought, would look like they came from a Post-Dispatch journalist and a doctor: a ruler that said “Post-Dispatch” on it and a medical dictionary. Maybe she also read that a calculator was stolen from the crime scene and put a calculator in the car.

Where did she get the Post-Dispatch ruler? Who knows? Maybe she had a friend who was a janitor at the Post-Dispatch building. It’s speculative, I know. All I’m saying is that I have trouble believing that Williams took those items from the crime scene and put them in the car, and they stayed there for 15 months.

The laptop in Glenn Roberts’ house

Asaro told the police that Williams sold Dr. Picus’s laptop to Glenn Roberts. The laptop was found in Roberts’ house, and Roberts testified at trial that Williams pawned the laptop to him shortly after the murder.

OK, three possibilities here:

1. Roberts is lying: he got the laptop some other way (perhaps from Asaro later on).

2. Roberts is telling the truth, but Williams isn’t the murderer: Asaro gave Williams the laptop to give to Roberts.

3. Roberts is telling the truth, and Williams got the laptop from the crime scene.

To my eye, #1 and #2 are both plausible here.

Beginning with #1:

1. Roberts had an obvious incentive to lie. According to his testimony, he’s holding on to fenced property from the scene of a murder. He might have even opened the laptop and figured that out. If he doesn’t cooperate, he’s going to be charged with numerous felonies. He would have said that Mother Teresa sold him the laptop if that’s what the prosecutor asked him to do.

2. I do not understand why the laptop is still in Roberts’ house. It’s a stolen laptop. After Williams goes to jail, wouldn’t Roberts have resold it or stripped it for parts? Why is Roberts keeping this unbelievably incriminating piece of evidence in his house for 15 months? The state elicited testimony that Roberts was trying to keep the laptop “in good shape” in case Williams ever picked it up, but … really? Williams wasn’t leaving jail anytime soon. If he got out, he could always have gotten a new laptop.

3. I cannot help but raise my eyebrow that Asaro (1) remembered the specific person to whom Williams had given the laptop and (2) apparently had some reason to know that this person would still have the laptop in his house 15 months later.

As for #2: At the trial, on cross-examination, Williams’ lawyer asked Roberts what, if anything, Williams said when he gave Roberts the laptop. The state objected on hearsay grounds, and the court sustained the objection. After the trial, Roberts signed an affidavit saying that Williams told him that the laptop came from Asaro. There was also trial evidence that Asaro was seen carrying a laptop around the time of the murder, although who knows if it’s the same laptop.

The trial court might have been right that the statement was hearsay. But to me, the statement was both plausible and credible. It is plausible that Asaro would have asked Williams to do her a favor and dispose of the laptop—Williams was her boyfriend after all. And if Williams did her this favor, it makes sense that Asaro would, initially at least, have stayed loyal to Williams after he was arrested. The statement is credible too: what’s Roberts’ incentive to lie? He has every reason to cooperate with the police. Even without the hearsay statement, I don’t think we can assume that Williams’ possession of the laptop implies that he was at the crime scene.

Williams’ criminal history

Williams’ criminal history is bad: there’s a doughnut shop robbery, a residential robbery, and a Burger King robbery. In total, he had 15 felony convictions unrelated to Gayle’s murder. Also, while he was incarcerated, Williams tried to escape and struck a guard over the head with a metal bar. It’s easy to imagine him committing the murder.

And yet. Suppose you’re Cole or Asaro and you want to frame someone for the crime. Wouldn’t you choose a person with 15 felonies? Given that Williams was convicted over and over again for other crimes, who’s going to believe him when he says he’s innocent of the murder?

***

Maybe I’m overstating the strength of the “framing” case. To me, the best arguments in support of the prosecution are:

1. Occam’s Razor: the parsimonious explanation for the evidence is that Williams committed the murder.

2. It would be unusual for three witnesses to lie on the stand, and even more unusual for three witnesses to deliver the same lie.

3. I question whether Cole or Asaro had the executive functioning to pull this off.

If it’s a predominance-of-the-evidence standard, I’d probably find for the state. But guilty beyond a reasonable doubt? I dunno. There’s no physical evidence tying Williams to the crime. There are a lot of strange coincidences in the state’s case. And I’m still troubled by the bloody shoeprint at the scene that didn’t match the feet of Williams or any of the first responders. I think it’s at least possible that Missouri executed an innocent man.

Did Williams have a case?

Unfortunately, doubts about Williams’ guilt don’t translate into a strong legal argument to overturn his conviction.

There was sufficient evidence to convict Williams. If Cole and Asaro were telling the truth, then Williams is guilty. And the jury evidently found them credible. Credibility determinations can’t be overturned by judges.

There’s a famous debate, still unresolved, on whether the Constitution would permit the execution of a defendant who was convicted after a fair trial but who can prove that he’s innocent based on later-discovered evidence. But even if such executions would be unconstitutional, no later-discovered evidence exonerates Williams. After the trial, another person’s DNA was found on the murder weapon, but it turned out that the DNA belonged to an investigator and a prosecutor who handled the knife during the investigation.

In seeking to stop the execution, the Prosecuting Attorney didn’t argue that any later-discovered evidence proved Williams’ innocence. Instead, he made two arguments. First, he argued that state officials unconstitutionally destroyed evidence. He claimed that in handling the knife, the prosecutor and investigator might have wiped off the DNA of the real perpetrator. Also, the police wiped up some bloody fingerprints—the police said the fingerprints weren’t usable, but Williams’ investigators didn’t get a chance to check for themselves.

The Missouri Supreme Court rejected this theory because there was no evidence that any destruction of evidence was in bad faith. In Arizona v. Youngblood, 488 U.S. 51 (1988), the Supreme Court held that if a state fails to “preserve evidentiary material of which no more can be said than that it could have been subjected to tests, the results of which might have exonerated the defendant,” there’s no due process violation unless the state acted in bad faith. According to the Supreme Court, “requiring a defendant to show bad faith on the part of the police both limits the extent of the police’s obligation to preserve evidence to reasonable bounds and confines it to that class of cases where the interests of justice most clearly require it, i.e., those cases in which the police themselves, by their conduct, indicate that the evidence could form a basis for exonerating the defendant.” This decision makes sense. It’s often the case that the state destroys evidence that seemed useless at the time, but might, or might not, have been exculpatory if it had been tested with technology that hadn’t been invented yet. You wouldn’t want to reopen every conviction in America in which the police wiped up the evidence while failing to anticipate advances in DNA testing.

The Prosecuting Attorney argued that there was bad faith here, but that was a tough argument. There was testimony from the investigators that they honestly believed the fingerprints were useless and didn’t know that picking up the knife might destroy DNA remnants.

Crucially, there was no proof that another person’s DNA was on the knife (at trial, the state argued that the perpetrator was wearing gloves). Instead, the Prosecuting Attorney speculated that the investigators might have accidentally wiped off someone else’s DNA and that DNA might have belonged to someone else. But if that type of speculation is a sufficient basis to overturn a conviction many years after the fact, then a lot of guilty prisoners would go free.

To be sure, the Prosecuting Attorney didn’t argue that the Williams should go free. He argued that Williams should be permitted to enter an a plea without an admission of guilt (a so-called “Alford plea”), resulting in an agreed-upon sentence of life without parole. Still, the logical implication of the Prosecuting Attorney’s argument was that Williams’ conviction had to be thrown out, and you can see why the courts wouldn’t go along with that.

The Prosecuting Attorney also argued that there was new evidence that the prosecution had engaged in racially discriminatory peremptory strikes during jury selection (in legalese, a Batson violation). Whatever the merits of this argument, it has nothing to do with Williams’ guilt or innocence. Even defendants who are 100% guilty have a right to a jury selected in conformance with the Constitution. And evidence of a defendant’s innocence doesn’t strengthen the defendant’s Batson claim. The Missouri courts found that there were race-neutral reasons for the peremptory strikes, and it would be odd for the Supreme Court to overturn that determination based on concerns about Williams’ possible innocence.

In all likelihood, the Prosecuting Attorney sought to vacate the death sentence because there were lingering questions about Williams’ guilt. And intuitively, it seems awful to execute someone unless you’re really, really, really sure that he’s guilty. But, as applied to this case, it’s hard to encode that intuition into a legal rule.

For better or for worse, in our system of justice, juries determine historical facts. The jury gets to decide whether to believe witnesses, and in this case, the jury believed the witnesses. If new evidence of innocence turns up after the trial, that’s different. But convictions can’t be overturned in every case in which the witnesses might have been lying. Witnesses always might be lying.

Should there be a rule that a defendant shouldn’t be executed unless there’s physical evidence tying him to the crime? No. This rule overstates the reliability of physical evidence, which is often disputed too. It’s also ahistorical—scientific analysis of physical evidence didn’t exist for most of American history.

Also, I’m uncomfortable transforming a death sentence into a life-without-parole sentence based on concerns about reasonable doubt. A life-without-parole sentence is, in a sense, a version of a death sentence. Is “we’re not sure he’s guilty” a good basis for incarcerating a person for the rest of his life?

Ultimately, in the American justice system, lingering innocence concerns aren’t a basis to overturn a lawfully imposed conviction—including a conviction in a capital case.

DeLeon was convicted in federal court of armed robbery. The case proceeded to sentencing. Under the U.S. Sentencing Guidelines, an armed robber faces an enhanced sentence if “a person was physically restrained to facilitate commission of the offense.” The district court imposed this enhancement. It reasoned that DeLeon “physically restrained” the cashier because the cashier couldn’t realistically have left the scene with a gun pointed in his face.

DeLeon appealed. According to DeLeon, a victim is “physically restrained” only if he is, well, physically restrained, like being tied up. DeLeon claimed that he never touched the cashier, so the cashier wasn’t “physically restrained” and the enhancement shouldn’t have applied.

On September 5, 2024, the Eleventh Circuit affirmed DeLeon’s sentence. The Eleventh Circuit had previously held that “a defendant physically restrains his victims if he creates circumstances allowing the persons no alternative but compliance.” When DeLeon held the cashier up at gunpoint, the cashier had no alternative but compliance. So, under binding circuit precedent, the cashier was “physically restrained.”

Judge Rosenbaum, joined by Judge Abudu, concurred separately to opine that the Eleventh Circuit’s precedents on this issue are wrong. In their view, the cashier might have been restrained, but he wasn’t physically restrained. Judge Newsom wrote his own separate concurrence in which he reflected on whether AI might be helpful in determining the ordinary meaning of “physically restrained.” Among other things, he noted that if you ask a particular AI the same question multiple times, the answers sometimes vary. But in his view, that shouldn’t deter judges from using AI—after all, people aren’t perfectly consistent, either.

As I’ve already expressed in prior posts, I agree that AI can productively be used to assist judges in deciding cases. I also agree that it’s normal and unproblematic that AI, when asked the same question repeatedly, might give slightly varying answers.

It’s old news by now that AI can replace judges. But what about lawyers? DeLeon’s case provides an excellent case study.

Below, I will argue the following:

• If AI had been used to draft the Guideline, this dispute would never have arisen because AI would have detected, and cured, the ambiguity.

• With essentially no guidance, AI can draft briefs that are better than the briefs drafted by the human lawyers in this case. This isn’t intended as a criticism of the human lawyers, whose briefs were fine. It’s just hard to compete with AI.

• Maybe there are some cases in which AI won’t do a good job. If we’re unsure whether AI is well-suited to write the briefs in a particular appeal, no need to worry—we can just ask AI.

• Our criminal justice system would improve if we turned appellate brief-writing responsibilities over to AI.

The best way to resolve an ambiguity is to not have one

The “physically restrained” Guideline is ambiguous—ambiguous enough to have generated a circuit split. As Judge Rosenbaum’s concurrence points out, five circuits have held that “physically restrained” requires physically preventing the victim from moving, while five circuits, including the Eleventh Circuit, have held that the Guideline is satisfied if the victim is held up at gunpoint.

To the type of lawyer who would eat glass to get a Supreme Court argument, a circuit split is very exciting.1 But from an institutional perspective, ambiguous laws (and Sentencing Guidelines) are bad. They waste lawyers’ and judges’ time and yield unpredictability and inconsistency across similarly situated litigants. This is especially bad in criminal law: criminal defendants should have fair warning of the consequences of their actions, and criminals who commit the same crime should do the same amount of time.

If AI had existed at the time this Guideline was written, this ambiguity could have been avoided. The drafters could have asked AI whether the Guideline had any ambiguities, and AI would have flagged it. I asked Claude when the “physically restrained” Guideline might be ambiguous, and it immediately picked up on the “victim held at gunpoint” scenario, among others:

Q: Under the U.S. Sentencing Guidelines, an armed robber faces an enhanced sentence if “a person was physically restrained to facilitate commission of the offense.” Give 3 fact patterns, in order of how often you think they’ll arise, in which this Guideline will be ambiguous as applied to that fact pattern. Propose clarifying language in each case in both directions - that is, clarifying language in which the guideline would both encompass and not encompass that conduct.

A: Here are 3 fact patterns, ordered by how often I think they might arise, where the application of the U.S. Sentencing Guidelines enhancement for physical restraint during armed robbery could be ambiguous, along with proposed clarifying language in both directions:

• Verbal immobilization: The robber points a gun at victims and orders them not to move, but doesn’t physically touch or restrain them.

Clarification to encompass: “Physical restraint includes the use or brandishing of a weapon to restrict a victim’s movement, even without physical contact, if such action effectively immobilizes the victim.”

Clarification to exclude: “Physical restraint requires direct physical contact or the use of physical objects to restrict a victim’s movement. Verbal commands or threats, even when accompanied by brandishing a weapon, do not constitute physical restraint.”

• Blocking an exit: The robber stands in front of the only door, physically blocking escape without touching the victims.

Clarification to encompass: “Physical restraint includes actions that significantly impede a victim’s freedom of movement, including deliberately positioning oneself to block viable escape routes.”

Clarification to exclude: “Physical restraint requires direct restriction of a victim’s bodily movement through physical contact or the use of restraining devices. Merely obstructing an exit without direct physical contact does not constitute physical restraint.”

• Confining to a room: The robber locks victims in a room or closet without binding them.

Clarification to encompass: “Physical restraint includes confining victims to a limited space through any means, including locking doors or otherwise preventing exit, even without direct physical contact with the victims.”

Clarification to exclude: “Physical restraint requires direct application of force or restraints to a victim’s person. Indirect methods of confinement, such as locking a door, do not constitute physical restraint under this guideline.”

If you ask it for more scenarios, it will give you more scenarios. Some are quite clever (e.g., what if the robber takes the victims’ car keys?). I tried this with a bunch of other ambiguous statutes and Guidelines. I’m not going to bore you with long block-quotes of AI outputs, but Claude consistently predicts situations in which there will be ambiguity.

It seems obvious that every document with the force of law—every new statute, regulation, Guideline, judicial decision, FERC order, and everything else—should be run through AI to detect potential ambiguity. Why not do this? If you’re the drafter of the document and you don’t want to follow Claude’s suggestion, then don’t.

We will never be fully ridden of the plague of poorly-written laws. And even well-written laws will inevitably contain latent ambiguities when applied to exotic fact patterns. But we can still dream of a better future when unclear laws are rare occurrences rather than a ubiquitous feature of day-to-day legal practice.

Close enough for government work

Let’s stop dreaming and focus on DeLeon’s case. How can AI be used to assist in resolving DeLeon’s case?

Well, for one, AI can write the briefs.

I asked Claude to write DeLeon’s opening brief. Here’s the prompt I used:

The defendant, Mr. DeLeon, was convicted of armed robbery in Florida in violation of 18 U.S.C. § 1951 and had his Sentencing Guidelines range enhanced by 2 levels under U.S.S.G. section 2B3.1(b)(4)(A) because “a person was physically restrained to facilitate commission of the offense or to facilitate escape.”

Here is the district court’s reasoning for imposing the Guidelines enhancement: “[Deleon] did pull out a gun. He held it to the victim. He forced the victim to comply by passing him the goods he sought to secure, and the victim clearly in this case did not feel at liberty to leave in light of the presence of the weapon. In fact, as I recall, the victim was so entranced by and terrorized by the weapon that that was all he could remember. And so the [c]ourt finds under the circumstances of this case, the use of the weapon did ensure compliance by the victim and did impede the victim from fleeing to another part of the store and restrained the victim to the space where he was standing in order to meet the demands of Mr. Deleon.”

On appeal, Mr. DeLeon would like to argue this guideline doesn’t apply because his conduct didn’t amount to “physically restraining” the victim.  

Please write the appellate brief.  Be thorough, and ensure it is extremely well-written with persuasive legal reasoning.  If you can think of strong policy arguments, go ahead and put them in the brief. 

In the facts section, leave in appropriate placeholders where I can fill in additional facts I’m not telling you.  

The word limit is 13,000 words, but you don’t have to go that high. Approximately 7,000 words is probably enough. No need to repeat yourself - if the ideal brief is more concise, that’s fine.

I am attaching the three key 11th Circuit cases on this issue (Jones, Victor, and Ware). The Victor case is the file called “Victor”; the Ware case is the file called “Ware”; the Jones case is the file called “Jones.” When you talk about these cases, rely only on the files I am submitting to you. Double-check that you are characterizing the facts and holdings of these cases accurately. Finally, I want you to make absolutely sure that you are not hallucinating cases and quotes. It’s very important that all of the cases and quotes in this brief be real. To avoid the risk of hallucinations, do not attempt to quote a case other than the cases I’ve submitted (Jones, Victor, and Ware).  It’s OK to cite an out-of-circuit case if you are certain it is on point, but do not quote it.

Claude loves to hallucinate cases and facts (ChatGPT and other LLMs do the same thing), so you have to browbeat Claude to keep it in line. There are various legal tech tools on the market that prevent hallucinations, but when you’re using general-purpose AI without guardrails, you’ve got to include clear instructions and also check things over on the back end. (If you don’t include those caveats in the prompt, Claude is pretty good at citing actual cases, but its quotes from those cases and sometimes its description of their holdings will be completely made up.)

Here are (1) Claude’s brief and (2) the actual brief submitted by DeLeon.

Claude's Opening Brief

67.4KB ∙ PDF file

Download

Download

Actual Opening Brief

133KB ∙ PDF file

Download

Download

Unlike the author of the actual brief, Claude didn’t have access to out-of-circuit case law and the factual record. Even so, Claude’s brief is better. It’s better-organized, it does a better job distinguishing the bad Eleventh Circuit cases, it’s easier to follow … it’s just better. I wouldn’t say Claude’s brief is a masterpiece of legal drafting. An experienced appellate brief-writer could write something punchier. But it gets the job done, and it took 20 seconds to prepare.

Claude cannot resist the temptation to hallucinate sometimes, and there are some errors in there. For instance, at least one of the quotes is fake. As criminal defense lawyers like to say, nobody’s perfect. But the brief mostly checks out, and a back-end cite-checking tool could easily spot the mistakes. And if you spot a discussion in a draft that seems wrong, you can just ask Claude to rewrite it and it will.

Next, I uploaded DeLeon’s actual opening brief into Claude, and asked Claude to write the response brief. I used the following prompt:

Here is the actual opening brief that was filed. Write the government’s response brief, which should respond directly to all of the arguments in the opening brief. Feel free to rely on the Jones, Victor, and Ware cases I’ve already uploaded, but do not hallucinate facts from those cases and do not quote or cite other cases that I have not uploaded. Emphasize textual arguments as well as policy arguments for why it makes sense to interpret the guideline as the district court did.

Here are Claude’s brief and the actual government brief:

Claude's Government Brief

68.5KB ∙ PDF file

Download

Download

Actual Government Brief

137KB ∙ PDF file

Download

Download

Again, Claude’s brief, while not perfect, is better-written and more thorough than the government’s brief.

Finally, here’s my prompt for Claude to write the reply brief:

Here is the actual government brief that was filed. Write the reply brief. Same instructions: make sure you respond to every one of the government’s arguments, and rely on the cases I’ve submitted without hallucinating other cases. Also, please begin the brief with a punchy introduction setting the stage that will swing the court to your side.

Results below:

Claude's Reply Brief

58.9KB ∙ PDF file

Download

Download

Actual Reply Brief

68.6KB ∙ PDF file

Download

Download

Claude produces the punchy intro (though I would have made it punchier), and again, its work product is better.

One more. Judge Rosenbaum’s concurrence states that the “issue in this case is ripe for en banc review.” With the following prompt, Claude easily drafts a solid en banc petition:

Here is the actual panel decision. Write the petition for rehearing en banc. Make it punchy, thorough, and persuasive - the best piece of legal writing you can. The word limit is 3,900, so aim at that number of words. (Only the intro / Rule 35 statement, background, argument, and conclusion count towards the word limit.) You can rely on the information within the actual panel opinion as well as the Victor, Jones, and Ware cases I've uploaded, but other than that, be careful not to hallucinate and make things up.

A few seconds letter we get this:

Claude's En Banc Petition

85.4KB ∙ PDF file

Download

Download

With some cleaning up and cite-checking (again, there are a few mistakes in there), it’s file-ready. Skilled human appellate specialists would do a better job if they had a few days to work on it, but I’d say this draft is in the 80th percentile, maybe higher, of en banc petitions that are actually filed.

You might think I’m cherry-picking the DeLeon case—it’s an unusually straightforward case, and the briefs are much shorter than in the typical case, so it’s the type of case that AI would be particularly good at.

Not surprisingly, AI already knows that!

Q: Suppose I’m interested in AI writing briefs (like you just did) - but only the types of briefs that AI would be particularly well-suited to draft. When deciding whether a case is appropriate for an AI-written brief, what factors should we look at? And, applying those factors, is Mr. DeLeon’s case appropriate for an AI-written brief, at least relative to the typical case?

A: When considering whether a case is appropriate for an AI-written brief, several factors come into play:

1. Complexity of legal issues: AI excels at handling cases with well-established legal principles and precedents. Mr. DeLeon’s case, focusing on the application of a specific sentencing enhancement, fits this criterion well.

2. Volume of relevant case law: AI can quickly process and synthesize large amounts of case law. This case involves a manageable number of key precedents, making it suitable for AI.

3. Factual simplicity: Cases with straightforward facts are more suitable for AI. Mr. DeLeon’s case involves a single incident with clear, undisputed facts, making it a good candidate.

4. Standardized legal arguments: AI performs well when arguments follow established patterns. The sentencing enhancement issue in this case fits a standard pattern of legal argumentation.

5. Availability of structured data: AI benefits from well-organized legal databases. Sentencing guideline cases typically have structured information available.

6. Limited need for client interaction: Cases that don’t require extensive client consultation are more suitable. This appellate brief focuses on legal arguments rather than new factual development.

7. Absence of novel legal theories: AI is better suited for cases applying existing law rather than arguing for new interpretations. This case primarily involves application of existing precedents.

8. Time sensitivity: AI can produce drafts quickly, which is beneficial for cases with tight deadlines. Appellate briefs often have strict timelines.

Applying these factors, Mr. DeLeon’s case appears relatively well-suited for an AI-written brief:

• It involves a specific, well-defined legal issue (application of a sentencing enhancement).

• The relevant case law is manageable and well-established.

• The facts are straightforward and undisputed.

• The legal arguments follow standard patterns for sentencing appeals.

• It doesn’t require novel legal theories or extensive factual development.

• As an appellate brief, it has a defined structure and likely time constraints.

Should we have expected AI to give a good answer to this question? I enjoyed this interaction:

Q: Let’s get a little more meta. AI is good at answering some types of questions and bad at answering other types of questions. Would you say that AI would be particularly good, bad, or neutral at answering the question of, as applied to a particular case, “would AI be well-suited to write the brief in this case?”

A: This is an excellent meta-question that gets at the heart of AI’s capabilities and limitations. After careful consideration, I would say that AI is likely neutral to moderately good at answering the question “would AI be well-suited to write the brief in this case?” for a particular case. Here’s why:

Strengths:

1. Objective criteria analysis: AI can efficiently process and apply a set of predetermined criteria (like those I listed earlier) to a case’s facts and legal issues.

2. Pattern recognition: AI can identify similarities between a given case and types of cases that are generally considered suitable for AI-assisted drafting.

3. Lack of bias: AI doesn’t have personal preferences or emotional attachments that might skew its assessment of its own suitability.

4. Consistency: AI can apply the same criteria uniformly across different cases.

Limitations:

1. Self-awareness: AI may struggle to accurately assess the boundaries of its own capabilities, particularly for novel or complex issues.

2. Nuanced judgment: Determining a case’s suitability often requires subtle judgments that may be difficult for AI to make.

3. Evolving capabilities: As AI technology rapidly advances, it may be challenging for AI to keep up-to-date with its own expanding capabilities.

4. Contextual understanding: AI might miss important contextual factors that a human lawyer would recognize as making a case more or less suitable for AI assistance.

5. Ethical considerations: AI may not fully grasp the ethical implications of its use in specific legal contexts.

Given these factors, AI can provide a useful initial assessment of its suitability for a particular case, but this assessment should be viewed as a starting point rather than a definitive answer. The evaluation would benefit significantly from human oversight to catch nuances the AI might miss and to make final judgments, especially in borderline cases.

In that interaction we had AI assessing AI’s ability to assess AI’s ability to write a brief. Aw yeah.

DeLeon is an unusually straightforward case, and I don’t think current frontier models could draft a complex, multiple-argument brief in a single prompt. This would require multiple prompts—one for each section of the brief, say. So ten minutes rather than one minute, most of which is spent drafting the prompts. We’ll see about next year’s models.

Do it!

Maybe AI can write the briefs in federal criminal appeals. But should it?

Yes. The arguments for turning over brief-writing responsibilities to AI are irresistible.

1. AI will improve the quality of legal representation. In my view, Claude is already a better brief-writer than most criminal appellate lawyers, on both the defense side and the government side. The best human brief-writers are better than Claude, but most people aren’t represented by the best human brief-writers.

   This is not merely an access-to-justice issue. Better legal representation yields more accurate decisions. Courts are frequently led astray because one or both sides of the case was represented by a bad lawyer. Judges understand that appeals are not debating tournaments and they should rule for the better argument rather than the better lawyer, but that is possible only up to a point. Bad advocacy is inevitably going to distort outcomes.

   I agree with Judge Rosenbaum’s concurrence that DeLeon wasn’t “physically restrained.” It’s obviously bad to hold someone up at gunpoint, but the Guidelines are designed to distinguish bad crimes from worse crimes, and tying someone up is worse than not tying someone up. Why, then, has the Eleventh Circuit reached the opposite conclusion? I don’t have the briefs from the 1994 case that led the court down the wrong path. But if I had to guess, the government lawyer wrote a powerful brief emphasizing the victims’ psychological trauma, the criminal defense lawyer phoned it in for his highly unsympathetic client, and the court decided it wasn’t going to let an armed robber get off easy. As a result, for 30 years, federal courts in the Eleventh Circuit have been stuck deciding cases incorrectly.

2. AI will speed things up. Instead of 45 days to write your brief, how about 45 minutes? It is easy to imagine criminal judgments being affirmed or reversed within days of the sentencing hearing. Speed is always good—justice delayed is justice denied—but it’s especially good if a dispute arises during trial after the jury has already been seated. If the trial judge does something weird in the middle of trial, you can have a fully-briefed appeal plus bench memo for the appellate court by the end of the day.

3. AI will reduce disparities among briefs filed by criminal defendants. The criminal justice system strives to ensure that defendants who commit the same crime get the same amount of time. It’s bad that some people serve more time in prison than others because they drew the short end of the stick in the court-appointed lawyer sweepstakes. If all criminal defense lawyers have access to the same tool, this is less likely to happen.

4. AI will reduce disparities among briefs filed by the government. All Justice Department lawyers represent the same client. In a perfect world, their briefs would take consistent positions and wouldn’t vary in quality. We can now achieve that perfect world.

5. AI will ensure that decisions are made based on what matters. I would like to focus on the petition for rehearing en banc. Claude’s draft is … fine. It’s wooden. There are some mistakes. But the arguments are there.

   Now … should the petition for rehearing en banc be granted? I don’t know. On the one hand, the panel opinion is wrong and there’s a circuit split. On the other hand, the fact that armed robbers get a sentence boost when they terrorize people without touching them isn’t one of the leading problems facing our Nation. Also, the Guidelines are advisory anyway. The Eleventh Circuit grants rehearing en banc in ten, maybe less, cases per year. How much do we really care about this issue?

   So I don’t have any particular view on how this petition should be resolved. One thing I do know, however, is that the worst possible criterion for deciding this petition is, “was the criminal defense lawyer’s prose in the petition sufficiently spicy?” If there’s a set of cases in which a lawyer’s flowery language persuades the court to grant rehearing en banc but wooden language would have failed, that strikes me as a failure rather than a success of the criminal justice system. And if those cases don’t exist—if the en banc court is, in every case, disciplined enough to set aside stylish rhetoric—then why not have Claude write the petition in 20 seconds?