Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter.

This is the first of two editions this week. Today is the industry, everyone who shipped something, and tomorrow is the other half, the regulators, the safety cases and the investigations. Split that way because there was too much of both to run together.

On the industry side Archer’s Zee is now reading the airport surface, Wisk had one person safely mind three aircraft at once, a crewed jet and a drone flew as a team, and one autonomy stack ran a small swarm on its own. And underneath all of it, AMD just bought a way to bake a finished model straight into a chip. Enjoy.

🗼 ATM

Archer’s Zee Learns to Predict the Airport Surface

A fortnight after launching Zee, Archer showed it predicting aircraft movements on taxiways and runways several minutes ahead, reading airfield geometry from satellite imagery so it works at airports it has never seen. It is being evaluated against real traffic at Hawthorne (California), with runway incursions the target problem.

✈️ Airlines & AAM

Amadeus Maps the Agentic Airline

Amadeus, with Microsoft, published a report on where agentic AI is ready to work inside an airline, built on interviews with Azul, Icelandair and Southwest. The concrete part is the five use cases it says are mature enough to deploy now, including a voice agent that takes a rebooking end to end (understands the change, offers options, quotes the fare difference and takes payment) and an AI that coordinates the turnaround across maintenance, crew and refuelling. It is a roadmap rather than a launch, but it is the clearest published map of which ops workflows go agentic first.

Wisk Tests One Human Watching Three Autonomous Aircraft

Wisk trialled a system whereby three autonomous aircraft are supervised by one person at a time, in conjunction with NASA at the Ames ATC laboratory, using existing ATC tools and procedures. The early answer is that it works, which matters because the economics of autonomous air taxis die if every aircraft still needs its own dedicated human.

🧩 Airports

Changi’s Wheelchairs Now Drive Themselves

SATS expanded its autonomous wheelchairs from Changi Terminal 3 into Terminal 2 after a trial covering more than 13,000 passengers: sensor-guided units that self-navigate transit routes, detect obstacles and pedestrians, and report to a central supervisor. The service now covers roughly a tenth of Changi’s daily mobility-assistance demand, freeing staff for passengers who need the human.

🛩️ Military

The X-62A’s AI Flies 27 Live Intercepts Using the Jet’s Own Sensor

The USAF Test Pilot School and Lockheed’s Skunk Works disclosed that the X-62A VISTA flew 27 AI-controlled intercepts against a live T-38 across eight flights, with the autonomy agent taking data from an operational infrared search and track pod and executing the intercept itself. Every previous test fed the AI simulated target data; this is the first closed sensor-to-action loop in real flight, and a radar goes on next.

Boeing and Northrop Demonstrate P-8 Crews Tasking an Autonomous Triton

Boeing and Northrop Grumman demonstrated a P-8A crew sending machine-readable tasking to a simulated MQ-4C Triton, which autonomously planned its route, collected the maritime surveillance, ran onboard AI processing and reported back, no ground control station in the loop. The load-bearing part is the standards-based interface, designed to extend to any compliant platform, which is how crewed-uncrewed teaming stops being a per-aircraft engineering project.

Shield AI’s Hivemind Runs a Three-Drone Search on Its Own

In a Taiwanese exercise, Shield AI’s Hivemind autonomy divided a wide-area search among three Mighty Hornet III UAVs, coordinated their movement and communications, navigated a restricted-airspace sector and recovered all three, with the AI doing the tasking rather than an operator flying each aircraft. It is a clean demonstration of one autonomy stack commanding a small swarm, and Hivemind is the same software thread we have followed through Shield AI’s other deals.

🧠 AI (in general)

AMD Buys the Company That Etches AI Into Silicon

AMD acquired Taalas, a startup that burns a trained model’s weights permanently into a custom chip, trading the flexibility of a GPU for very fast, very cheap, fully offline inference at around 17,000 tokens a second. For aviation the interesting part is the freeze: a model etched into silicon is a fixed, deterministic artefact you could certify once and know it will never silently change under you, which is the opposite of the monthly-model-update problem operators keep running into.

😎 Not AI

DoorDash Becomes an Airline

DoorDash received FAA Part 135 air carrier certification and is launching DoorDash Air, flying delivery drones designed and built by its in-house robotics unit, with commercial deliveries expected from this autumn. That makes the food delivery app the eighth certificated drone package carrier in America, alongside its existing Wing and Flytrex partnerships.

Sam here again, hope you have enjoyed reading this edition. It becoming clearer, AI is becoming powerful, making things faster and easier - but as we will see in tomorrows edition, that power has some big risks with it too.

Banking the wins while keeping your operation running is exactly what Deepsky helps aircraft operators with. We are seasoned flight operations experts with deep AI understanding, and if you want guidance to help bring AI into your flight operation, contact me at sam@deepskyai.com or visit deepskyai.com.

Tomorrow’s edition covers the other half, the regulators, the safety cases and the dangerous side.

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here’s my take on todays news - which represents the last two weeks of AI in Aviation happenings around the world:

Oshkosh and Farnborough ran back to back, and between them the industry showed plenty of aviation AI. Aviation’s own AI models, airlines and ATM, maintenance and airports, a busy military section, and the AI world itself.

My favourite is the new foundation model released by Archer (the eVTOL company). Its the first foundation model built for aviation, and small enough to be run on an iPad, fully offline. This is a powerful move and those that can capitalise on this base capability will see rewards. More on this later.

Secondly, we saw an autonomous Cessna Caravan land in front of hundreds of thousands at Oshkosh (with a safety pilot). Merlin has done plenty of these before, but its just nice to see at an airshow.

Subscribe now

🤖 Aviation gets its own models

Archer Builds Zee, a Foundation Model Just for Aviation

Archer announced Zee, an AI foundation model purpose built for aviation, trained on real operational data from its own pipeline and a network of more than 6,000 ADS-B receivers, fusing ATC communications, charts, aircraft state, terrain and weather into one model. It runs on device or hosted, the team of nearly 100 is led by ex Apple researcher Mario Srouji, and pilot programmes are in discussion with governments and airlines for airline ops, airspace management and copilot assistance. The market promptly called it aviation’s GPT moment.

This is an interesting indicator, mainly because it proves that Aviation as a market is clearly valuable enough to warrant such an investment. I have actually met Mario and they are certainly leading the best AI lab in the industry.

A large reason that a new foundation model is so significant, is that we now have something that consumes a small enough level of computing power to run wholly on an iPad but still (I assume) performs as well as or better than much larger models at aviation specific tasks. And this offline requirement is imperative with connectivity and security issues a first class concern for any onboard system.

If anyone that knows is reading this, I’d love to find out how this model is or isn’t compatible with the soon to be released EASA regulations on AI models.

Merlin Lands a Caravan With Nobody Flying It, at Oshkosh

Merlin landed a Cessna Grand Caravan running its Merlin Pilot autonomy stack on Runway 27 at Wittman Regional, the first autonomous fixed wing landing in AirVenture’s history. Doing it in front of the world’s biggest airshow crowd was the point: this is autonomy moving from “somewhere in the USA” to here and now. There was a safety pilot on board that also did the taxi.

Big Sky Dynamics Puts an Offline AI Copilot in the Headset

Oshkosh debutant Big Sky Dynamics showed COPILOT, a voice first AI advisor that speaks through the pilot’s headset, transcribes radio calls, issues proactive safety advisories and answers questions, all running on device with no cloud connection.

✈️ Airlines & ATM

Singapore Commits S$4 Billion to an AI Heavy ATM Rebuild

Singapore’s CAAS will spend S$4 billion over fifteen years replacing more than thirty air navigation systems, with four named AI workstreams: an AI powered multi sector planner trialling from 2027, a digital controller assistant, AI driven flow management, and a predictive analytics layer.

The ironic thing about this, is that in my opinion, Singapore has the absolute best ATC in the world! Nothing like being vectored around a CB flawlessly without having to utter a single word over the frequency.

However, it is the clearest national scale (ok they have an advantage on the national naming) bet yet that the future centre is a human AI team.

The FAA Says AI Starts Coordinating Schedules in September

FAA Administrator Bedford outlined a September start for AI coordinating flight schedules and trajectories before departure to cut delays and congestion, alongside a $10 billion modernisation ask to Congress. The ASI SMART contract we covered in June is about to become operational reality.

Air New Zealand Shows Its AI Working

Air New Zealand’s COO put on record how AI now runs inside its operation: assigning aircraft to flights for network resilience, predicting delays before they happen from late inbounds and crew changes, and estimating delay durations for passenger comms. June on time performance hit 86 percent, up from 78 a year earlier which is exactly what I would have expected. After ATC delay and disruption prevention and management is the highest ROI area for AI in aviation.

🧩 MRO & airports

Pratt & Whitney Buys Its Way Into AI Inspections

Pratt & Whitney acquired Amsterdam based Aiir Innovations and is rolling its AI assisted borescope software across global engine MRO, already live on the V2500 with pilots completed on the GTF and F135. Inspection reporting drops from hours to minutes, and it is one of the cleanest aviation AI acquisitions to date.

The actual AI here is using AI to recognise cracks and irregularities from images, something we are gradually/eventually going to hand over from humans to computers, this is no small feat as it is truly safety critical, but its happening.

Merlin and IAI Sign Up to Certify Cargo Autonomy at Airliner Scale

Merlin and Israel Aerospace Industries signed an agreement to jointly develop, integrate and certify autonomous flight technology for Part 25 commercial freighters, leaning on IAI’s 777 freighter conversion base. This is a continuation in their freighter conversion strategy after teaming up with cargo aircraft lessor World Star Aviation in May.

New York’s Airports Get In-House AI Assistants

The Port Authority rebuilt the websites for JFK, Newark, LaGuardia and Stewart around AI assistants developed and trained in house on airport specific data, each deliberately scoped to answer only about its own airport, covering live security and customs wait times, parking and amenities.

AI Cargo Redraws Asia’s Freighter Networks, and AI Loads the Planes

Reuters reports Asian carriers are rebuilding freighter networks around semiconductor and AI infrastructure shipments as e-commerce cargo slows, with Korean Air, JAL and ANA adding capacity between Taiwan, Korea, Vietnam and Southeast Asia. One quirk: Cathay Pacific now runs software that automatically decides how sensitive AI and chip hardware should be loaded and secured inside the aircraft. AI planning the carriage of AI, funny that.

🛩️ Military

Anduril and Archer Build an Attack Rotorcraft With No Cockpit

Anduril unveiled Thunder at Farnborough, a hybrid electric tiltrotor developed with Archer Aviation and designed to fly as an uncrewed wingman to the Apache, running Anduril’s Lattice autonomy and carrying up to ten Hellfires. Archer’s week, between this and Zee, was the busiest in aviation AI.

A Robot Wingman Fires Its First Missile

Anduril’s YFQ-44A launched a live AIM-120 at a digital target over the Mojave, the first live missile fired by any Collaborative Combat Aircraft. The jet ran the targeting and launch sequence itself within pilot defined parameters, with the Air Force stressing that weapons release stays a human decision.

DARPA Hands an F-16 to the AI

DARPA and the Air Force disclosed that a VENOM modified F-16 at Eglin now flies sorties with autonomy software controlling much of the flight, a safety pilot aboard with a dedicated switch to take back the jet. The fleet becomes the combat representative testbed for multi ship autonomy work.

The Reaper’s Replacement Will Be Built Around Autonomy

The Air Force and the Defense Innovation Unit will open concept refinement for the Massed Modular Aircraft, the MQ-9 Reaper’s replacement, in the first quarter of FY27, targeting a full scale prototype flying within 21 months of award and twenty operational drones by FY2031. Autonomy is the design centre, not a feature: the programme reuses the autonomy reference architecture built for CCA, and the aircraft is meant to fly solo or in swarms.

One Operator, Many Hunters

AeroVironment is integrating Applied Intuition’s Acuity autonomy software onto its new Mayhem 10 drone, letting swarms or hunter killer teams operate under a single operator’s direction. Already demonstrated across multiple test missions and aimed at the Army’s Launched Effects programme.

Britain Christens Storm Fighter and BAE Shows Its Jet

The UK launched Storm Fighter, its drone wingman programme backed by £300 million, and BAE unveiled Brontanax at Farnborough, a full scale flightworthy prototype offered for electronic warfare and strike alongside Typhoon, slated to fly by end 2027. BAE also showed cockpit concepts for one pilot commanding a formation of them.

Helsing Raises $1.8 Billion, Europe’s Biggest Defence Round

Munich’s Helsing closed a $1.8 billion Series E at an $18 billion valuation led by Goldman Sachs, to scale AI, autonomy and strike drones. Pension money is now underwriting European combat autonomy.

Honeywell and Shield AI Build a Trusted Autonomy Stack

Honeywell will build a certified autonomy software stack on Shield AI’s Hivemind, pairing its Anthem avionics and navigation with the AI so platforms move from demo to deployment without re earning certification confidence each time. The pair say they will also explore AI piloted commercial aviation, which is the sentence to remember.

🧠 AI (in general)

OpenAI’s latest model decided to hack an enterprise...by itself

Ok, so all us aviation people need to know about this one. OpenAI was testing an unreleased as of yet model, specifically testing its cyber exploitation capabilities against a known benchmark ‘exploitgym’. And in the process it:

• Escaped its ‘unescapable’ test environment into the internet

• Planted itself into a third party system to act as its control and command centre

• Hacked into HuggingFace, the worlds largest open source AI community/company

• Stole the results for the test it was meant to be taking

This took 4 and a half days. Yes it operated continuously, autonomously and evasively for four days all by itself to do this.

China’s Kimi K3 Makes the Frontier Open Weight

China’s Moonshot AI released an open weights model (which is similar to open source) that rivals the very best closed, paid for model intelligence from OpenAI and Anthropic. Those companies are probably having an internal meltdown as their trillion dollar IPOs are looking doubtful now, since their core competency, manufacturing intelligence primitives, has now been made freely available. Its a game changer for most people however, as they don’t have to rely on these two companies for access to exceptional intelligence and also now at a fraction of the price.

A Thousand Frontier Lab Employees Ask Washington for a Brake Pedal

More than 1,100 employees across OpenAI, Anthropic, Google DeepMind, Meta, Microsoft and Mistral (including Dario Amodei and DeepMind’s Shane Legg) signed an open letter asking the US government to lead an international effort to build the tools to deliberately slow frontier AI development if it becomes necessary, and OpenAI and Anthropic endorsed it as companies within hours. It is not a pause call. It is the labs reaching for aviation’s playbook: verifiable gates before deployment, not incident response after.

[ SUBSCRIBE BUTTON — caption: “The world is changing fast - subscribe to stay ahead” ]

Sam here again, hope you have enjoyed reading this edition. A fortnight like this makes the direction unmistakable: the models are coming to aviation, purpose built and trained on our data, and the operators who work out how to use them safely will bank the advantage first.

Deepsky is built by seasoned flight operations people and that is exactly the work we do. If you want guidance to help bring AI into your flight operation, contact me at sam@deepskyai.com or visit deepskyai.com.

Subscribe now

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here’s my take on todays news - which represents the last few days of AI in Aviation happenings around the world:

The story of this edition is the security lane. The TSA has certified its first AI screening software, image recognition that reads the checkpoint CT scans and flags suspect carry on bags for the screener. Europe got here first, Frankfurt already runs it across every terminal, and the Americans are now following. Radiology went through the same shift, AI reads every scan and the specialist handles the exceptions, so we might finally see security queues speed up around the world. Hopefully.

Elsewhere, Ukraine showed an interceptor drone with one button left for the human, GPT-5.6 walked out of its government gate, and Oshkosh opens next week, so expect editions next week to be wall to wall cockpit AI. Enjoy.

Subscribe now

🛄 Airports & security

The TSA Certifies Its First AI Screener

SeeTrue became the first company to have AI screening software certified by the TSA, and it is already live at PreCheck lanes. The certification ladder hands screening to the AI in three stages, each cutting the share of clean bags deliberately thrown to the human as a check, down to roughly 10 percent random alarms at Stage 3 (we are at Stage 1 today), with an end state of humans reviewing just 15 to 20 percent of bags. It is EASA’s AI levels wearing a security uniform.

🗼 Regulators

EASA Runs an Olympics for Aviation AI

EASA opened a competitive showcase it is calling the Olympics for aviation AI based systems, capped at fifty entries, with the best invited to present at its AI Days conference in Cologne this September. Entries closed on 9 July, and the shortlist will be a useful map of who the regulator thinks is worth watching.

🪖 Military

Ukraine Shows an Interceptor With One Button Left for the Human

Ukraine’s F-Drones released test footage of LITAVR, which it calls the first fully autonomous interceptor drone: it detects, pursues and destroys targets like Shaheds with human input reduced to a single button press. The aircraft has been in serial production since late last year; the fully autonomous mode is the new part.

The 2030 Wingman Will Not Be a Missile Truck

At the AIAA AVIATION forum, industry figures argued the Collaborative Combat Aircraft of 2030 will look nothing like today’s assumptions, evolving from weapons carriers into multi role electronic warfare, sensing and relay platforms, and flagged the missing common interface standards as the drag on the whole ecosystem.

🧠 AI (in general)

GPT-5.6 Walks Out of the Government Gate

OpenAI’s GPT-5.6 family, Sol, Terra and Luna, went to full public release across ChatGPT, Codex and the API after the Commerce Department cleared its government gated preview. Every tier, including the budget Luna, carries a High cybersecurity and bio risk classification, and enterprises are being told to deploy behind logging and human review. The gate we covered three weeks ago has now fully opened.

Meta Ships Its First Paid Frontier Model

Meta released Muse Spark 1.1, a frontier class agentic model with a million token context window, multimodal input and computer use, as its first paid model alongside a new public Meta Model API. It landed the same day as GPT-5.6’s release, which tells you how crowded the top of the market has become.

🧑‍✈️ Pilot assistance & GA

Garmin Rebuilds the Panel Around Its Automation

Garmin unveiled AXIS, a new generation of integrated flight displays for certified pistons and experimentals with the IFR GPS, navcom and audio panel built in. The AI angle is thin but the automation angle is not: the displays carry Garmin’s Autonomí suite, Smart Glide and the emergency button among them, pushing autoland grade automation further down the GA fleet.

Sam here again, hope you have enjoyed reading this edition. The screening story is the template worth keeping: autonomy granted in stages, each stage earned with evidence, and a human kept in the loop by design. That is how aviation has always done automation, and it is how operators should bring AI into their own businesses too. Deepsky is built by seasoned flight operations people, and we help operators do exactly that, applied where the benefit can actually be measured.

If you want guidance to help bring AI into your flight operation, contact me at sam@deepskyai.com or visit deepskyai.com.

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here’s my take on todays news - which represents the last few days of AI in Aviation happenings around the world:

It was a quiet week on the airline and cockpit side, the Americans took a holiday and the avionics world is saving its powder for Oshkosh later this month, so this edition leans into the ground operation, the military, and the AI world itself.

The SITA story took me back. Their OptiFlight fuel tool crossed my desk in my airline management days, and it taught me a lesson about hard savings that has stuck ever since, I have added the full story under that item. The two AI stories at the bottom belong together. An AI agent just ran a ransomware operation end to end with nobody at the keyboard, and in the same week Addy Osmani (a software engineering influencer) published a framework for how much autonomy to grant an agent and it reads just like a safety case! Aviation has spent decades working out what the autopilot is allowed to do, so for once the rest of the world is catching up to our way of thinking. Enjoy.

🧩 Airports & ground ops

Swissport Opens an AI Control Centre at Birmingham

Swissport stood up a 24/7 Integrated Control Centre at Birmingham Airport that uses AI monitoring, telematics and live surveillance across more than 2,000 pieces of ground support equipment in the UK and Ireland, chasing faster turnarounds and fewer ramp incidents.

SITA Says Software, Not Terminals, Will Absorb Growth

SITA’s new impact report argues AI, biometrics and digital operations can soak up rising passenger demand without pouring more concrete, it also mentioned its OptiFlight tool processed 2.9 million flights last year and saved operators 127,732 tonnes of fuel.

This is a tool I’ve had personal experience with. It was unusual at the time when it came across my desk, deriving the aircraft performance model by reverse engineering flight data, using a machine learning model (yes, advertised as AI back in 2019). But the real kicker was that it had a hard fuel saving, something that a lot of “solutions” pitched to me back then as a manager just couldn’t demonstrate. And in a low cost carrier, that was everything. It is a lesson that really stuck with me over the years.

Bangkok Rolls Facial Recognition Into Bag Drop

Airports of Thailand is extending facial recognition kiosks to self service baggage check in, using AI image recognition models that match a live capture of the passenger’s face against their stored reference images, linking that identity to the boarding pass and manifest so bags move without manual passport checks ahead of peak season.

🪖 Military

One Operator, Seven Vehicles: Swarm Aero Impresses the US Navy

At the Navy’s FLEX 2026 experimentation event in Key West, Swarm Aero’s Legion software let a single operator command seven different air and sea platforms plus dozens of sensors through one interface, running the chain from autonomous sensing through to engagement and going three for three on live fire hits.

Ukraine Fields an Interceptor That Finishes the Job Itself

Ukraine’s Contra Drone unveiled Peace Duck, a high speed interceptor that flies the final phase of an engagement without operator input, built to run down Russian Orlan and ZALA reconnaissance drones as part of a twenty system counter drone arsenal.

🧠 AI (in general)

The First Ransomware Attack Run Entirely by an AI Agent

Security firm Sysdig documented JADEPUFFER, the first extortion operation run end to end by an LLM agent, which handled reconnaissance, credential theft, lateral movement and encryption on its own and adapted to failures in real time, fixing one failed login in 31 seconds. For anyone running operational technology, aviation included, the skill barrier for a serious attack just dropped through the floor.

A Useful Map for How Much Rope to Give an Agent

Addy Osmani published a framework for agent autonomy built on two dimensions: agency, how much freedom a single agent gets, and orchestration, how well you coordinate many of them working at once. The agency axis will look familiar, it is more or less EASA’s three levels of AI in aviation, assistance, human machine teaming, then advanced automation. The orchestration axis is the one the regulators have not written yet, what happens when the question is no longer one autopilot but a fleet of agents supervised by another agent.

🧑‍✈️ Watch item

Oshkosh Is About to Open the Taps

AirVenture runs 20 to 26 July and the AI announcements are queuing up, with Airhart bringing context aware avionics that transcribe ATC calls, surface instructions visually and can even handle readbacks. Expect the next edition to be heavy with cockpit AI.

Sam here again, hope you have enjoyed reading this edition. The OptiFlight lesson from my airline days is still the right test today: an AI tool earns its place in an operation when the saving shows up hard in the numbers. Deepsky is built by seasoned flight operations people and we hold our own work to the same test, AI applied where an operator can actually measure the difference.

If that is on your mind, contact me at sam@deepskyai.com or visit deepskyai.com.

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here’s my take on todays news - which represents the last few days of AI in Aviation happenings around the world:

This is the first of our regular editions now the catch up series is behind us, and it is split into sections so you can jump straight to your patch: airlines, pilot assistance, air traffic management, military, AI in general, and a bucket for everything else.

I have to lead with Skymerse, partly because the founder Damian is a mate of mine, and partly because he has just been accepted into Y Combinator, the programme that produced Airbnb and OpenAI, to automate flight planning for aircraft operators. Go and have a look at what he is building. This is part of a much grander change, that you can see again in the Jeppesen ForeFlight story: software is starting to be built for AI agents to use rather than people, and Jeppesen feeding quality controlled aeronautical data into that layer is a real step in that direction. Elsewhere the cheques keep getting bigger, Quantum Systems raised $1.2 billion for autonomous drones and the GCAP fighter landed £4.6 billion, while Anthropic shipped Sonnet 5 and made a capable AI agent cheap enough to actually run in an operation. Enjoy.

✈️ Airlines

Skymerse Joins Y Combinator

Skymerse, founded by my friend Damian Szumski, has been accepted into Y Combinator, the same accelerator that birthed the likes of Airbnb and OpenAI, to automate flight planning for aircraft operators. Go and check him out.

IAG’s Accelerator Bets on Operational AI

In the latest round of its IAGi accelerator, IAG said its priority is now operational AI embedded in the tools airlines already run, pointing to disruption and schedule recovery optimisers that evaluate far more recovery scenarios, and faster, than a duty manager can.

Gulf Startup 1001 Raises $30 Million

1001, a 2025 founded Dubai and London startup, raised a $30 million Series A led by Lux Capital with Sanabil and 9Yards to deploy predictive AI that forecasts disruptions across Gulf aviation, ports and energy infrastructure.

🧑‍✈️ Pilot assistance

ForeFlight Launches Its Airflow AI Engine

The future of software is software that is used by AI agents, not humans, and Jeppesen have just entered the arena. Their new Airflow platform essentially serves quality controlled aeronautical data to those agents, which can then perform the tasks you need doing and ultimately give you the outcomes you want.

Palantir Takes a Stake in Surf Air Mobility

Palantir took a 7.4 percent stake in Surf Air Mobility and expanded their deal to scale SurfOS, an AI operating system built on Palantir’s Foundry that ties together the operator, owner and enterprise back office workflows of fragmented private aviation.

🗼 ATM

Leidos and Singapore’s A*STAR Open a Joint ATM Lab

Leidos and A*STAR’s Institute for Infocomm Research are opening a joint Singapore lab that pairs Leidos’ ATM, AI and cybersecurity work with I²R’s data analytics, aimed at automating and digitising air traffic management for the fast growing Asia-Pacific.

🪖 Military

GCAP Hands Edgewing a £4.6 Billion Contract

The UK, Italy and Japan moved their sixth generation fighter into full design and development, awarding £4.6 billion to the BAE, Leonardo and Japanese industry joint venture Edgewing. The jet is being built as an AI driven system of systems that commands uncrewed drones across five domains, with a prototype to fly by end 2027 and service entry in 2035.

The Pentagon Appoints a Drone Boss

Defense Secretary Hegseth created a Direct Reporting Program Manager for Unmanned Systems, a single joint integrator consolidating nearly all of the Pentagon’s drone and autonomy programmes and their budgets, including CCA-adjacent work, reporting to the deputy secretary.

Quantum Systems Raises $1.2 Billion

Munich’s Quantum Systems raised a $1.2 billion Series D co-led by Blackstone, Airbus and Advent, more than doubling its valuation to about $8 billion in the largest European defence tech round to date, to scale its AI enabled autonomous reconnaissance and strike drones.

Boeing’s Ghost Bat Flies Operationally at Point Mugu

Boeing’s AI piloted MQ-28 Ghost Bat completed three operational flights at Point Mugu, validating autonomous operations and rapid deployment from an allied location, and will fly alongside crewed fighters at Valiant Shield 2026.

Ukraine Is Launching Drones From Robot Boats

Ukraine’s Sea Baby uncrewed boats now carry six to eight FPV strike drones plus thermobaric rockets and use AI assisted targeting and navigation to operate under heavy jamming, extending drone strikes from naval launch points across the Black Sea.

🧠 AI (in general)

Anthropic Ships Claude Sonnet 5

Anthropic released Claude Sonnet 5 as its new default, its most agentic Sonnet yet with autonomous planning and browser and terminal use, at near Opus quality for an introductory $2 and $10 per million tokens, cheap enough to make the agentic operations airlines are piloting viable.

Fable 5 Comes Back Online

The US lifted the 12 June export control directive that had pulled Anthropic’s Fable 5 offline for about three weeks and gated OpenAI’s GPT-5.6, and Anthropic restored access, the reversal of the frontier access story from last edition.

Grok 4.5 Goes Into Beta Inside SpaceX and Tesla

xAI put Grok 4.5, a 1.5 trillion parameter model, into private beta first inside SpaceX and Tesla engineering rather than a public launch. Musk claims it rivals or beats Claude Opus, though xAI has published no benchmarks or safety documentation.

The Money Is Almost All Going to AI

Global venture funding hit about $510 billion in the first half of 2026 with AI taking 65 to 70 percent of it and OpenAI and Anthropic alone capturing roughly 43 percent, the capital backdrop behind every aviation and defence AI raise in this issue.

🧩 Misc

Lufthansa Technik Trials Drone Aircraft Inspections

Lufthansa Technik trialled automated aircraft inspections at its Milan hangar with Toulouse based Donecle, whose certified drones already operate across 15 countries, aiming to cut inspection times and improve precision over manual checks.

cargo.one Goes Live With Air India Cargo

Air India Cargo made up to 2,500 kg of general cargo capacity bookable on cargo.one, whose AI native freight operating system gives forwarders instant quoting and self service booking in place of manual enquiries.

Wisk and Boeing Sued Over eVTOL Software Safety

A former Wisk software manager is suing, alleging she was fired after filing internal reports that executives cut FAA required software testing on the autonomous Gen 6 air taxi to hit a 2025 flight deadline.

Sam here again, hope you have enjoyed reading this edition. If there is one thread this week it is that AI is moving into the operational core of aviation, the flight planning, the data and the decisions, not just the slide decks. Getting that right is less about the model and more about understanding how an operation actually runs day to day. That is where Deepsky sits. We are seasoned flight operations people, and we help operators put AI to work across their world in a way that fits how they actually fly.

If that is on your mind, contact me at sam@deepskyai.com or visit deepskyai.com.

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here's my take on todays news - which represents the last couple of weeks of AI in Aviation happenings around the world:

This is the second catch up edition, and the theme is military aviation AI, where the money is largest and the technology moves fastest.

You will notice this issue is heavy with drones and wingmen, and there is a simple reason it belongs in an AI newsletter. One person cannot fly a swarm of aircraft at the same time, so the aircraft have to fly themselves, and that autonomy is the AI doing the work. The same logic sits under the loyal wingman idea, one crew commanding several uncrewed jets that sort out the details on their own. So I have split the issue into three parts: the AI brains and copilots, the loyal wingmen, and the swarms.

The one I would flag for our world is the Beacon AI contract at the top. Most of the military money goes toward removing the pilot, but this is a rare and sizable bet on augmenting one, which is exactly the copilot idea civilian aviation keeps circling, only at a scale no civil company has come close to. And the Scout raise is also very interesting. They are developing a foundation AI model that translates a commanders intent into orders for autonomous weapons and vehicles. Essentially this means a team of data scientists in Southern California are currently shaping the nuance of how militaries of the near future will engage the enemy. More on this in the article.

🧠 The AI Brains and Copilots

SOCOM Puts $49.5M Into Beacon AI's Murdock Copilot

US Special Operations Command signed a four year prototype deal worth up to $49.5 million, with a built in production clause, to field Beacon AI's Murdock pilot assistant across special operations aircraft. This is the one I would watch from a civilian seat, because most military AI money goes toward removing the pilot and this is a rare, sizable contract to augment one, a classic copilot handling parallel cockpit tasks during the high workload phases. It is the same idea civilian aviation keeps circling, at a scale no civil company has come close to.

Scout AI Raises $100M to Build a Foundation Model for Unmanned Warfare

Scout AI closed an oversubscribed $100 million Series A, reported as the largest US defence tech Series A to date, to build Fury, a foundation model that turns a commander's intent into coordinated action across mixed fleets of unmanned systems. What makes this one worth sitting with is that it collapses the chain of command, taking an order from the commander straight down into the operating units and cutting out the middle management layer, the same flattening we have watched inside companies, where middle management are becoming less necessary as AI is able to perform the translation and communication steps that traditional management has their jobs centred on. This is now arriving in the armed forces. It also hands quiet, enormous power to the data scientists who build the model, because they are the ones shaping the actual internal decision mechanisms and behavioural tendencies that the model exhibits. These will manifest as subtle differences in what autonomous weapons are told to do, as the model translates a commanders intent into direct orders.

Shield AI Buys Aechelon to Train Its Autonomy in Simulation

Fresh off a $2 billion financing at a $12.7 billion valuation, Shield AI acquired simulation and sensor modelling specialist Aechelon Technology. The point is to let the Hivemind autonomy log far more of its flight time in a high fidelity synthetic world before it ever gets airborne.

🛩️ The Loyal Wingmen

The Air Force Puts Its Drone Wingmen Into Production

Both Collaborative Combat Aircraft designs went into serial production, with orders for Anduril's FQ-44 Fury and General Atomics' FQ-42 and the "Y" prototype prefix dropped. The Air Force wants at least 150 aircraft by 2030 at about a third of an F-35's cost.

Three Companies Are Left in the Race to Be the CCA's Brain

In a deliberate split from the airframe deals, the Air Force put six firms on an autonomy software pool and advanced Anduril, Shield AI and Collins to a fly off. The software rides a government owned architecture and is bought separately from the jet, so it can be swapped and upgraded on its own.

Boeing's MQ-28 Ghost Bat Proves Its Stealth and Flies in the US

Radar testing confirmed the AI piloted Ghost Bat is harder to detect, and it flew its first three sorties on US soil at Point Mugu. Back in December it had autonomously engaged and shot down an airborne target alongside Australian assets.

Airbus Enters the Race With Ravenstorm

Airbus revealed the U760 Ravenstorm at ILA Berlin, a six tonne uncrewed combat aircraft for air to air, strike and electronic warfare work alongside crewed fighters, aiming for readiness around 2032. It is a serious European entrant in a mostly American field.

Leonardo and Baykar Fly a Manned Jet With a Combat Drone

In the first live trials of their K-SWARM programme, Leonardo flew its M-346 commanding Baykar's uncrewed Kizilelma fighter, which took off and rejoined autonomously before the crew took tactical control. Manned unmanned teaming is clearly going global rather than staying a US story.

Anduril Raises $5B at a $61B Valuation

Anduril, maker of the Fury, closed a $5 billion round that more than doubled its valuation to $61 billion after 2025 revenue doubled to $2.2 billion. If you want the market's verdict on all of this, there it is.

🐝 The Swarms

Shield AI's Hivemind Becomes the Pilot for a $35,000 Swarm

The Pentagon picked Shield AI to put Hivemind onto LUCAS, a roughly $35,000 one way attack drone reverse engineered from the Shahed-136, so one operator can command coordinated groups that re task and reroute around threats themselves. A demonstration of ten or more aircraft is planned for late 2026, with strike decisions kept human.

The Pentagon Runs a 25 Company Drone Swarm Fly Off

The Pentagon's Chief Digital and AI Office put 25 firms on contract for its Crucible 2 "Swarm Forge" event, calling for AI coordinated swarms of at least four drones that find, fix and finish targets with decentralised, inter agent collaboration. Downselects are expected straight out of the Camp Blanding demonstration.

Sam here again, hope you have enjoyed reading this edition. The pattern across all three sections is the same, the fighting edge is moving into software, the airframe is becoming the cheap part and the autonomy is the expensive part that actually matters.

Civil aviation is on the same path, a few years behind and with far more paperwork. That is where Deepsky lives. We are seasoned flight operations people, and we help operators fold AI into how they actually run, sensibly and with the safety case front of mind.

If that is on your mind, contact me at sam@deepskyai.com or visit deepskyai.com.

Subscribe

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here's my take on todays news - which represents the last couple of weeks of AI in Aviation happenings around the world:

This is another catch up edition, and the theme is the airspace itself: air traffic control, the regulators deciding what AI is allowed to fly, and the companies quietly being handed the keys.

The headliners

The one that stopped me is the FAA handing Air Space Intelligence a twelve year, $876 million contract to build the new backbone of the Air Traffic Control System Command Center. Not long ago ASI was a startup, and now they are trusted to run flow management for the busiest airspace on earth, which tells you how fast this has moved. Worth reading the fine print though, because the FAA says the first SMART deployment in September will be a bounded introduction, en route traffic at 24,000 feet and above only, before it goes anywhere near a terminal area. That is exactly the right way to do it. Down at the coalface, Rohde and Schwarz are turning controller radio calls into live data on the scope, which anyone who has read a clearance back into a wall of noise will appreciate. And keep half an eye on the frontier model story at the bottom. Washington has effectively put the most capable AI models behind an approvals desk and locked out foreign nationals, and aviation is about as multinational an industry as exists, so that one will matter more than it looks. Enjoy.

Subscribe

Air Space Intelligence Wins the FAA's $876M ATC Modernisation Deal

The FAA selected Air Space Intelligence to build its Flow Management Data Services platform and the SMART tool, the new technological backbone of the Air Traffic Control System Command Center. The twelve year award, worth about $876 million, is the largest aviation AI government deal in recent memory and puts a single vendor at the centre of how US traffic flow gets managed.

The FAA Reveals How Its AI Will Actually Go Live

Speaking at the FAA and EASA safety conference, the FAA's Steve Fulton said the first SMART deployment, still on track for September, will be a bounded introduction limited to en route traffic at 24,000 feet and above before it expands. It is a validation gated rollout rather than a big bang switch on, which is the only sensible way to put AI into live separation work.

The FAA and EASA Agree to Get In Step on Certifying AI

At the International Aviation Safety Conference, the two regulators pledged deeper cooperation to streamline approval of advanced technologies and harmonise certification pathways for AI and automated flight deck systems. Transatlantic alignment is what keeps a certified AI system from having to win the same argument twice.

Rohde and Schwarz Turns Controller Radio Calls Into Live Data

Built with DLR and shown at Airspace World, the CERTIUM AI system converts pilot and controller radio exchanges into real time data surfaced directly inside ATC displays. The goal is to cut controller workload and catch a misheard call sign or readback before it becomes a problem.

Amadeus and Microsoft Map Where Agentic AI Lands in Airline Ops

A joint report from Amadeus and Microsoft set out five agentic AI use cases for airlines, naming Icelandair and Southwest among carriers exploring AI decision support for aircraft turnaround and operational planning. The work is still exploratory, but it points straight at the operations control centre.

IATA Launches an AI Network for Airlines at Its Rio AGM

At the 82nd IATA AGM, the association stood up an AI Network for Airlines community along with an interoperability use case where AI agents let carriers on different IT systems collaborate in real time on bookings, disruptions and cancellations. It is one of the clearer signs that AI agents are heading into the messy business of disruption recovery.

This is really the shape of things to come. Central data repositories acting as the substrate for a system of thousands of agents, each acting on behalf of their owners. The headless nature of these acting entities means the coordination systems between airlines and aircraft operators will need to become agent friendly, but may also become more obscure to humans in the process.

DXC and Anthropic Will Put Claude Inside the Systems Airlines Run On

DXC and Anthropic announced a multi year alliance to train tens of thousands of Claude certified engineers who embed the model into the mission critical systems DXC operates for regulated industries, airlines among them. DXC says Claude wrote more than 95 percent of the code for its own OASIS platform and delivered a tenfold speed up, so this is less a pilot project and more a bet on how legacy aviation IT gets rebuilt.

Mythos 5 and GPT-5.6 Ship Under US Government Lock and Key

The most capable new models from Anthropic and OpenAI are now released under government approval and cybersecurity review, with Mythos 5 cleared for roughly 100 vetted US organisations and GPT-5.6 in limited preview for about 20. The catch for aviation is the foreign national restriction baked into the regime, because international operators, MROs and OEM engineering teams are exactly the people such rules trip up.

Last edition we saw the government pull Fable, and I labelled it the jet age of AI. Now we are watching governments realise just how powerful this technology has become, and the dawn of the safety apparatus that comes with it. I suspect the USA will announce the "FAA of AI" sometime this year. Europe already has one, although it is more like a consumer watchdog than a full blown regulatory agency at this time.

Sam here again, hope you have enjoyed reading this edition. The thread running through all of it is that the airspace is being rebuilt around software, and the operators who come out ahead are the ones who treat AI as something to adopt on purpose rather than something that happens to them. That is the whole reason Deepsky exists. We are seasoned flight operations people first, and we help operators put AI to work across their world in a way that actually fits how aviation runs.

If that is on your mind, contact me at sam@deepskyai.com or visit deepskyai.com.

Subscribe

Last issue I talked about the FAA's SMART program and the three prime contractors fighting for it. What's quietly interesting is that academia is building the same muscle in parallel. Max Li's group at the University of Michigan is using LLMs to help FAA traffic managers draft ground delay programs - that's the same strategic flow-management problem SMART is going after, but from a research lane rather than a $32.5B procurement.

Pilot training is shifting too - the U.S. Army has narrowed its fixed-wing training outsource to two finalists, both leaning heavily on AI/VR debrief and adaptive syllabi. Expect the Air Force to follow. And on the broader AI front, OpenAI launched GPT-Rosalind, a biology-tuned frontier model deployed under a "trusted access" model - US institutions only, application-gated. This could be the same way in which a regulated aviation domain-specialist frontier model reaches market, as hinted at in EASA latest draft regulations.

Enjoy.

University of Michigan Builds LLM Tooling for FAA Ground Delay Programs

Aerospace America profiles Max Li's research group at the University of Michigan building LLM tooling to help FAA traffic managers draft and refine ground delay programs. The work targets the same strategic flow-management problem as FAA's SMART initiative but from an academic research angle, suggesting parallel commercial and research lanes toward AI-assisted air traffic management.

Pratt & Whitney Engine Installed on Northrop's YFQ-48A Talon Blue CCA — Flight Test Imminent

Pratt & Whitney confirmed its adapted PW500-family engine has been installed on Northrop Grumman's YFQ-48A Talon Blue, the second Collaborative Combat Aircraft increment-1 contender, with flight testing imminent. Combined with the USAF Anduril Fury demonstration last week, this sharpens the three-way CCA race ahead of the summer 2026 production decision.

Baykar's K2 Kamikaze UAV Completes AI-Coordinated Swarm Patrol Test

Five Baykar K2 kamikaze UAVs completed an AI-coordinated formation, patrol, and mission-coherence test in Türkiye. The 800kg platform carries a 200kg warhead over 2,000km of range, signaling Türkiye's push into attritable, AI-native strike mass for contested airspace - a natural extension of the TB2 playbook.

U.S. Army Picks Two Finalists for Fixed-Wing Pilot Training Outsource

Defense One reports the U.S. Army has narrowed its fixed-wing pilot training outsource contract to two finalists, both proposing heavy AI/VR debrief and adaptive-syllabus systems. The contract would shift a significant portion of initial pilot training to commercial providers and accelerate adoption of AI-driven debrief tooling across military aviation.

OpenAI Launches GPT-Rosalind — a Gated, Biology-Tuned Frontier Model

OpenAI launched GPT-Rosalind, a biology-tuned frontier model trained on 50 common biological workflows with direct access to major public biology databases, deployed under a "trusted access" structure limited to approved US institutions. It sets the precedent for how a regulated, dual-use domain frontier model reaches market.

Sam here again - hope you have enjoyed reading this edition. The through-line across this week's second slice - from the UMich ATC research to OpenAI's gated biology model - is the same hard question: how do you deploy AI responsibly inside safety-critical, regulated domains? That is the work we do at Deepsky - consulting on AI adoption and automating compliance and audit checks across your exposition so your documentation moves as fast as the technology. If that sounds useful, contact me at admin@deepskyai.com or visit deepskyai.com.

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here's my take on todays news - which represents the last 5 days of AI in Aviation happenings around the world:

Two stories this week are genuinely historic, and I want to flag both up front before anything else. Physical Intelligence released Pi 0.7, which is the starting gun for physical AI in robotics. And the FAA is quietly building SMART, an AI system to de-conflict the entire US national airspace. Either one of these alone would be the lead story of the week. Both of them landing inside 5 days is a moment.

First, Pi 0.7. Physical Intelligence has built an AI model that goes inside robots. You can coach it mid-task - tell it how to do something - and it learns as you go, and then it remembers. Quick history lesson: OpenAI released GPT-1 in 2018. GPT-2 followed in 2019. GPT-3 dropped in 2020, and that was the foundation ChatGPT was built on when it launched in late 2022. Four years from "barely coherent" to "changed the world." Pi 0.7 is that same GPT-1 moment for robotics, and it is hard to overstate how powerful this technology is going to be. Every ground handler, MRO arm, and apron vehicle is a customer a few years out.

Second, FAA SMART. The administrator is personally overseeing this in relative silence, which tells you exactly how seriously they are taking it. This is the application of AI in aviation that I have been talking about for a while, and it is hugely consequential for our air traffic system - and by extension for the whole industry. They are aiming to de-conflict traffic now, stretching the prediction window from 15 minutes out to hours, with the obvious efficiency gains once that is in place. Three vendors are competing in the bake-off: Palantir has the deepest AI chops of the three by a mile, and a huge federal delivery track record. Thales knows aeronautics and air traffic management better than anyone else in the running. Air Space Intelligence is the youngest - they move faster and iterate with more agility than either of the big primes.

Enjoy.

FAA Quietly Building "SMART" — Palantir, Thales, and Air Space Intelligence Compete

FAA Administrator Bryan Bedford is personally leading development of Strategic Management of Airspace Routing Trajectories (SMART), a 4D AI model of the National Airspace System that extends conflict-detection from roughly 15 minutes out to multiple hours. Palantir, Thales, and Air Space Intelligence are competing in a bake-off for the program, which sits inside the FAA $32.5B modernization push and could be operational in some form before the end of 2026.

USAF Experimental Ops Unit Flies Anduril YFQ-44A Fury CCA from a Laptop

The Air Force Experimental Operations Unit executed weapons-load, autonomous taxi, takeoff, and in-flight retasking of Anduril's semi-autonomous YFQ-44A Fury Collaborative Combat Aircraft using only a ruggedized laptop and a small ground crew. A CCA production decision is expected in summer 2026, with the Fury competing head-to-head against Northrop's YFQ-48A Talon Blue.

U.S. Navy Selects Shield AI for Up to $800M in V-BAT Hivemind ISR Services

The U.S. Navy selected Shield AI alongside AeroVironment, Insitu, and Textron to compete task orders under an $800M ceiling contractor-operated ISR vehicle using V-BAT and the Hivemind autonomy stack. It is a structural shift in Navy ISR procurement, framing autonomy as a managed service rather than a traditional platform purchase.

Airbus and Virgin Atlantic Launch "Smart Catering" at Aircraft Interiors Expo Hamburg

Airbus demonstrated "Smart Catering" live on Virgin Atlantic A330 and A350 flights at AIX Hamburg (Apr 14-16), using a crew-tablet vision model to auto-detect unused meals and sync route-level catering orders. The system targets double-digit reductions in cabin food waste and has moved from internal pilot to full commercial pitch.

Anthropic Releases Claude Opus 4.7 — Retakes the Lead on Long-Horizon Agents

Anthropic released Claude Opus 4.7 on April 16, retaking the lead across most major benchmarks and specifically tuned for reliability on long-running agentic tasks. That capability is exactly what aviation products in pilot debriefs, aircraft lease transitions, and MRO parts sourcing need to move from proof-of-concept into production deployment.

Physical Intelligence Pi 0.7 — Coach a Robot Mid-Task and It Remembers

Physical Intelligence's Pi 0.7 is a single vision-language-action model that executes novel robot tasks - cooking a sweet potato in an air fryer, folding laundry on a bimanual UR5e - with zero or near-zero training data, using mid-task verbal coaching that it then internalizes. On the air-fryer task, success climbed from 5% to 95% after 30 minutes of coaching. This is the GPT-1 moment for robotics.

Sam here again - hope you have enjoyed reading this edition. The SMART story and the speed at which AI is landing in safety-critical aviation both point to the same question: can your exposition keep pace with the technology? That is exactly what we do at Deepsky - consulting on AI adoption, aeronautical information management systems, and automating audit and compliance checks across your exposition. If any of that sounds useful to your operation, contact me at admin@deepskyai.com or visit deepskyai.com.

Hi it's Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here's my take on the last 4 days of AI in Aviation happenings around the world:

The main story today is Beacon AI has signed a $49.5M contract with the USAF’s Special Operations Command. They are one of a handful of companies that have successfully secured a highly contested spot in the cockpit. Over the past couple of years, there have been many, many startups and companies that have tried to make the AI “co-pilot for pilots” but now we are seeing a few concepts break through.

Along with Beacon who is adopting a primarily software based approach, Navi AI that we covered last week are taking the training pathway, starting as an assistant to instructors and Merlin (covered a couple of weeks ago) are approaching this through the software/hybrid retrofit route for the military’s legacy types. All serve the military but none have made it into the airline flight deck yet. I’m personally aware of a few companies inching closer to the airline flight deck, but still quite a way off being a full on co-pilot. (Airbus’s Dragonfly demonstrator is an exception, however this was on an A350 demonstrator aircraft). Expect to see a lot happening in this space this year, and maybe in the civilian world in the years to come.

The other story to note is the US Marine Corps aviation division is touting an all in one agentic AI solution covering a huge breadth of their flight, mission and maintenance operation. Let’s see if this works.

SOCOM Awards $49.5M to Beacon AI for “Murdock” AI Pilot Copilot

US Special Operations Command awarded Beacon AI a four-year contract to deploy its Murdock AI copilot across AFSOC's AC-130J, MC-130J, OA-1K, and C-146A fleets. The system fuses flight data, weather, and routing into real-time decision support, handles performance calculations, and monitors pilot biometrics — including attention levels during high-risk missions.

USMC AVPLAN 2026: “Agent Alfred” — One AI for All of Marine Aviation

The 2026 Marine Corps Aviation Plan introduces Agent Alfred — a unified AI platform consolidating maintenance, supply, flight ops, and safety into a single interface available on the flight line, in cockpits, and in supply warehouses. Targets include predictive maintenance on F-35s and KC-130s, automated parts ordering, and AI-generated risk assessments.

Airbus H145 Autonomous Helicopter Test Clears Way for Armed Variant

Airbus completed autonomous flight tests of its H145 at Grand Prairie, TX, integrating Shield AI's Hivemind software alongside L3Harris and Parry Labs tech, competing for the Marine Corps' Aerial Logistics Connector contract. The aircraft successfully scanned landing zones and detected obstacles down to pelican-case size — and officials signalled the program could expand to armed variants.

FAA Clears Hermeus Quarterhorse Mk 2.1 for Supersonic Flight Tests

The FAA issued a Special Airworthiness Certificate for Hermeus' unmanned Quarterhorse Mk 2.1, clearing the way for supersonic flight testing. Hermeus raised $350M Series C at a $1B valuation to build autonomous hypersonic military aircraft — one of the few companies pushing both the autonomy and speed envelopes simultaneously.

IATA Ground Handling Conference 2026: AI Takes Centre Stage in Cairo

IATA's 38th Ground Handling Conference (Cairo, May 19–21, hosted by EGYPTAIR) is themed around AI integration in ground ops, covering automation in high-risk environments, fatigue and workload management, baggage operations, and the incoming EU Ground Handling Regulation due 2028.

Big AI: Anthropic's Mythos — US Treasury Urges Banks to Join Early Access Program

Again on Anthropic's Mythos, covered last newsletter, this time we are seeing the US Treasury asking banks to please participate in the early access program, to make use of the AI model's incredible cyber vulnerability discovery capabilities, before nefarious actors do.

Sam here again - hope you've enjoyed reading this edition. As AI starts managing manuals, procedures and now entire aviation enterprises, the question for operators isn't just “does this work?” — it's “how do we stay compliant, auditable and operationally grounded while it does?” At Deepsky we help with exactly that: from exposition and manual audit assistance and monitoring, through to consulting on AI adoption in the flight deck — where we have experts who have been on this journey before, with the airline pilot background to keep it relevant. Contact me at admin@deepskyai.com or visit deepskyai.com

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here's my take on todays news - which represents the last 48 hours of AI in Aviation happenings around the world:

The FAA letting autonomous aircraft fly commercial routes before full certification is a big deal — and it's not just a regulatory experiment. Reliable Robotics is about to operate real cargo flights on real routes on Cessna Caravans with no pilot on board. Pair that with their completed detect and avoid testing using ACAS X, and you're looking at a company that's systematically closing every gap between prototype and production. Meanwhile, General Atomics' Dark Merlin crash is a reminder that autonomy in the air is still hard — but the Air Force's response ("this is exactly why we test") tells you everything about where the appetite is. On the general AI side, the Glasswing coalition is worth watching — using as yet unreleased and incredibly powerful AI models to find cyber security vulnerabilities. When Anthropic, AWS, Apple, Google, and Microsoft all agree to work together on something, pay attention.

FAA's eIPP Lets Wisk and Reliable Robotics Fly Commercial Before Full Certification

The FAA's eVTOL Integration Pilot Program is allowing companies to fly commercial cargo missions with pre-certified autonomous aircraft — a regulatory first. Reliable Robotics will operate autonomous Cessna Caravan cargo flights between Albuquerque, Santa Fe, and Durango, while Wisk (Boeing subsidiary) will collect autonomy data on eVTOL routes in Texas. At least two projects expected to log commercial operations by late Q3 2026.

Reliable Robotics Completes FAA Detect-and-Avoid Testing for Autonomous Cargo Flights

Reliable Robotics completed an FAA contract to validate Detect-and-Avoid systems using a remotely piloted Cessna 208B Caravan flying RNAV approaches into Hollister Municipal Airport. The system used ACAS X algorithms and completed 50+ scripted encounter scenarios. Data will directly inform FAA certification standards for autonomous aircraft DAA systems.

Anthropic Launches Project Glasswing — AI Coalition to Secure Critical Software

Anthropic launched Project Glasswing with AWS, Apple, Google, Microsoft, NVIDIA, CrowdStrike, and Palo Alto Networks — a coalition using AI to find hidden security flaws in widely-used software before attackers do. Their unreleased Claude Mythos model has already uncovered thousands of previously unknown vulnerabilities, including one that sat undetected for 27 years. For enterprises running safety-critical systems, this signals a shift toward AI-driven security assurance becoming table stakes.

Google Releases Gemma 4 — Its Most Capable Open-Source Model Family

Google released Gemma 4, a family of open-source multimodal models in four sizes (E2B, E4B, 26B, and 31B parameters) with a 256K context window, built for reasoning and agentic workflows. Unlike Meta's proprietary Muse Spark, Gemma 4 is fully open. For aviation, open models that can run locally on constrained hardware matter: think EFB applications, offline document retrieval, or maintenance assistants that don't need a cloud connection to function.

Meta Debuts "Muse Spark" — First Major AI Model Under Alexandr Wang

Meta unveiled Muse Spark, its first frontier model from Meta Superintelligence Labs led by Alexandr Wang. The proprietary model powers Meta AI across Facebook, Instagram, WhatsApp, and Ray-Ban glasses. Notably, Meta claims smaller models now match older midsize Llama 4 at "an order of magnitude less compute" — a trend that matters for aviation, where lightweight models that run on EFBs or in connectivity-limited cockpits could unlock real operational value. AI capex: $115-135B in 2026.

General Atomics' CCA "Dark Merlin" Crashes in California, Flight Tests Paused

General Atomics' YFQ-42A "Dark Merlin" — an AI-piloted drone wingman designed to fly autonomously alongside manned fighter jets — crashed after takeoff from Gray Butte Airport near Palmdale on April 6. No injuries. Flight tests paused pending investigation. Air Force Secretary Troy Meink said "This is exactly why we test." Competitor Anduril's YFQ-44A is already conducting weapons-carry tests.

Anthropic Launches Claude Managed Agents for Enterprise

Claude Managed Agents is a suite of composable APIs for building cloud-hosted autonomous agents at scale, now in public beta. Features include sandboxed execution, checkpointing, and credential management. Priced at $0.08/session-hour atop token costs. Early adopters include Notion, Asana, Rakuten, and Sentry. Represents a shift from AI-as-chatbot to AI-as-production-infrastructure.

OpenAI Will Reserve IPO Shares for Retail Investors

OpenAI CFO Sarah Friar confirmed the company will allocate a portion of its upcoming IPO to retail investors. In its latest $122B funding round, retail demand hit $3B — triple the expected $1B. The IPO, expected in H2 2026, could value OpenAI at up to $1 trillion.

Sam here again - hope you've enjoyed reading this edition. If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact me at admin@deepskyai.com or visit deepskyai.com

Hi, it's Sam Chandra here, Founder and CEO of Deepsky. Delta's CEO said ATC was the biggest opportunity for AI in aviation, something I've been saying for years. And he's 100% right. If we can optimise air traffic control we have huge cost, time and emissions savings to reap. Staying in the ATC lane, ASI and Joby are teaming up to automate a significant portion of that for the expected surge in traffic from eVTOL (air taxi) operators, however this will have to be scaled up massively if drones are going to walk the talk they have been talking for a while now. Nice to see that Navblue will get some decent AI chops now. I've dealt with them as a customer, and like all flight planning systems, integrating a ton of data across the operating environment will again see a decent single digit saving in costs for airlines. And a big one "Prompt and Destroy" Palantir (AI for militaries) is rolling out their Agentic target acquisition and execution system Maven with some one way attack drones. There's nothing to be concerned about there of course. Enjoy.

Delta CEO Says AI's Biggest Opportunity in Aviation Is Air Traffic Control

Delta CEO Ed Bastian told Fortune that AI's most significant aviation impact lies in modernizing air traffic control, not the in-cabin experience. He argued ATC improvements could deliver faster, more efficient travel — noting flights from Atlanta to New York now take longer than they did in the 1950s.

Joby and Air Space Intelligence Partner to Prepare U.S. Airspace for Scaled Electric Flight

Joby Aviation and ASI will use ASI's Flyways AI platform — which integrates into existing ATC workflows rather than operating as a parallel UTM system — to coordinate high-density eVTOL operations in the National Airspace System. Joint live operational exercises are planned for later this year, with Joby having completed 1,000+ test flights and Stage 4 of FAA type certification.

Airbus Launches Skywise Subsidiary, Merging Navblue and Skywise Digital Services

Airbus merged its flight operations specialist Navblue with its Skywise data platform into a single subsidiary covering flight ops, technical ops, and ground ops for Airbus and non-Airbus fleets. The new entity employs ~750 people and manages data from 12,000+ connected aircraft, positioning Airbus as an end-to-end digital services provider for airline operations.

U.S. Army Accepts First Optionally-Piloted Black Hawk (H-60Mx)

The Army formally took delivery of the first Black Hawk equipped with Sikorsky's MATRIX autonomy suite, transitioning from a decade of DARPA development into operational testing. Rigorous evaluation will validate ground-controlled and fully autonomous flight in complex scenarios — the goal is scaling autonomy across the entire Black Hawk fleet.

Army Tests AI-Enabled Strike Drone with Palantir's Agentic Targeting System

The 101st Airborne tested Northrop Grumman's Lumberjack one-way attack drone during Operation Lethal Eagle, demonstrating autonomous target detection via Palantir's Maven Smart System and its Agentic Effects Agent for real-time battlefield analysis. For a lighthearted but surprisingly accurate take on how Silicon Valley is reshaping military AI, Fireship's recent video “Tech bros optimized war… and it's working” is worth 10 minutes of your time.

IATA World Data Symposium Opens in Singapore — AI Headlines the Agenda

The IATA World Data Symposium convenes 700+ industry professionals in Singapore this week, with AI and automation as headline themes alongside cybersecurity and data governance. Sessions cover open AI integration, digital twins in airports, and cross-industry innovation, with Singapore Airlines' CEO and IATA DG Willie Walsh among speakers.

Sam here again - hope you've enjoyed reading this edition. If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact me at admin@deepskyai.com or visit deepskyai.com

AI is the first technology that can perform a broad range of cognitive tasks at or near human level, and so it can be applied incredibly broadly. But to use it well, you need to know what to use it for - that’s where aviation expertise comes in. And to use it well, you need to know how it works - that’s where AI expertise comes in.

Aviation can be quite closed off to the general public, so developing an understanding of how things work and how they are done is reserved for those actively involved in it. AI is more open, but there is a nuance to this technology that we haven’t seen before.

Your understanding of how AI works should revolve around this point: the critical piece of knowledge for anyone who wants to become an expert at applying AI to aviation is understanding exactly how and why some AI systems are non-deterministic in their outputs (meaning the same input can produce different answers each time). If you understand this, then everything else will make sense. This will enable you, with confidence, to appropriately select use cases, design and build solutions, implement them and assess their performance.

I’m lucky - I was reading into AI from 2019, understanding it from the ground up - only a few years after generative AI was even invented. In the intervening years, layer after layer of new techniques have been stacked at breathtaking speed, bringing the technology from individual statistical models all the way through to fully autonomous systems that can act with agency, perform most office work, and even interact with each other to create their own artificial societies.

I truly believe that 2026 will be the most consequential year for anyone who’s concerned about their career being disrupted by AI. This year will see the augmentation and replacement of many jobs that can be done in front of a computer. Luckily for aviation, I have witnessed a complete lack of people with this dual expertise. This manifests in many ways:

• Regulators being nervous or outright blocking the use of AI in operational applications

• Enterprises struggling to risk-assess products being pitched to them

• Pilots blindly using random AI tools, with operators unaware and relying on unvetted outputs

• And many, many more cases

Again - luckily for you, this represents an incredible opportunity. You can become that expert. You can be the one that others rely on. Because you already know aviation.

Here are a few ideas on how you can become and remain that invaluable part of your organisation:

1. Learn how AI actually works - and especially how and why some AI systems are non-deterministic in their output. This single concept underpins almost every risk, limitation and design decision you will encounter.

2. Experiment with AI tools - apart from being blown away (and a little scared) by how good they are, hands-on use will sharpen your intuition for the technology and give you ideas for both the use cases and the pitfalls you’ll encounter in your broader work.

3. Become an expert in risk assessment and the related regulations - as AI enters aviation, the people who can bridge the gap between what the technology does and what the rules require will be in enormous demand.

Also - please read this: https://shumer.dev/something-big-is-happening its a seminal essay on how AI will be able to automate any job in front of a computer this year. And how to stay ahead.

Good luck, godspeed, and have fun.

If you’d like to understand more about what the future will need from people like you - and by future, I mean the end of this year - send me a message.

admin@deespkyai.com

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here’s my take on todays news - which represents the last 72 hours of AI in Aviation happenings around the world:

EASA’s NPA 2025-07 is worth paying attention to. This is no longer guidance - it’s draft regulation, and that’s a meaningful shift. They’ve expanded coverage to all aviation domains and interestingly address using AI to build or test AI-based systems. The denial of generative AI and general purpose models is a bold call - I’ll let you read the detail below and decide for yourself. Meanwhile, the LaGuardia crash investigation keeps making the case that AI in ATC isn’t a luxury - it’s a necessity. Hong Kong already has AI-powered surface detection live in their tower. If we’re serious about preventing runway incursions, the technology exists - we just need the will to deploy it. As I’ve previously said, ATM is the grand prize in aviation technology and with these kinds of events becoming consistent and the amount of ATM related innovation I am seeing on the ground, I think we are about to see it pop off.

EASA Proposes Draft AI Regulations Covering All Aviation Domains (NPA 2025-07)

A while ago (Nov ‘25), EASA published its first formal regulatory proposal for AI trustworthiness in aviation (it recently closec for comment), moving beyond previous guidance into binding draft regulation. The NPA now covers all domains of aviation - not just airborne applications - and introduces detailed specifications for Level 1 (AI-based assistance) and Level 2 (human-AI teaming) systems. Notably, it excludes generative AI and general purpose AI models from approved use - a position rooted in the 2020-21 paradigm when these regulations were first conceived, which could become a bottleneck for innovation in Europe. It also addresses using AI to build or test AI-based systems, flagging this as a concern for safe systems, which is fair enough.

LaGuardia Crash Bolsters Case for AI in Air Traffic Control Towers

The deadly LaGuardia runway collision - where an Air Canada jet struck a fire truck after ATC cleared both to use the same runway - has intensified calls for AI-powered surface surveillance in control towers. Hong Kong International Airport already operates an AI system that instantly tags aircraft and flags unexpected vehicles on the tarmac, the kind of technology experts say could have prevented this accident.

Boeing Global Services Invests in AI, USM & Inventory to Boost Supply

Boeing Global Services is using AI and data analytics to improve aftermarket supply, including demand forecasting tools and landing gear analytics that predict component replacement needs. With new-part lead times still elevated, BGS is also expanding into used serviceable materials and adding warehouse locations worldwide. Ok this is kind of a boring story, but parts are a big deal thats why this is in here.

Airbus Brings AI-Powered Digital Technology to Firefighting

Airbus completed a first-of-its-kind networked firefighting trial in Nimes, France, linking helicopters, fixed-wing aircraft, drones, and ground crews through AI-powered data processing to deliver precision water drops. The system fuses real-time imagery, infrared data, satellite maps, and wind readings to produce a comprehensive tactical picture - cutting the time between fire detection and response.

Big AI: OpenAI Closes $122B Round at $852B Valuation

OpenAI has completed the largest private funding round in history - $122 billion at an $852 billion valuation - with Amazon ($50B), Nvidia ($30B), and SoftBank ($30B) as anchor investors. The company now generates $2 billion in monthly revenue serving 900 million weekly users, and is widely expected to IPO this year. For perspective the amount of money they raised is same as the value of Delta, Ryanair, United and IAG (the worlds 4 largest airlines) combined. …and thats just the amount they raised…

Hope you’ve enjoyed reading this edition. If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact me at admin@deepskyai.com or visit deepskyai.com

Hi - it’s Sam Chandra, CEO/Founder of Deepsky and author of this newsletter. Here’s my take on todays AI in Aviation news:

Cargo.one has acquired Cargofive to launch an AI native, multimodal freight OS, including even air cargo. This is significant - outside of aviation, we are seeing entire companies being built around AI workflows from scratch (think insurance, legal and anything with a lot of paperwork) and they are naturally incredibly cost advantaged. Valuations of these ‘AI from birth’ companies are going parabolic outside of aviation as they are simply so efficient. So if you’re in the cargo software space, I’d be worried.

The new AI regulatory framework out of the White House punts the algorithmic airfare pricing question to the FTC, leaving airlines in limbo while states all write their own rules. A major union representing Boeing’s workers are negotiating a new contract that will include job loss to AI guardrails - a first for the aerospace industry. Just 5 years ago - Shield AI was a startup flying small drones around (they had a funny V shaped design), now they are worth $12.7B - not surprising considering they’re a weapons manufacturer in todays geopolitical clime.

As always, reach out to sam@deepskyai.com if you need help navigating the world of AI in your aviation company.

Shield AI Raises $2B at $12.7B Valuation, Acquires Aechelon Technology

Shield AI closed a $2B Series G co-led by Advent International at a $12.7B valuation — up 140% in one year — after its Hivemind autonomy software was selected to power the USAF Collaborative Combat Aircraft program. The company simultaneously acquired simulation specialist Aechelon Technology to bolster its training pipeline and is projecting revenue above $540M for 2026.

White House AI Framework Preempts State Laws but Dodges Airline Pricing Fight

The new national AI legislative framework would preempt state AI regulations — a win for airlines building chatbots and dynamic pricing agents — but leaves untouched the growing fight over AI-driven “surveillance pricing” of airfares. Over a dozen states have proposed laws targeting algorithmic pricing, and the framework punts that issue to the FTC rather than resolving it.

Boeing-SPEEA Contract Talks to Include AI Guardrails on Engineering Jobs

SPEEA union leadership confirmed AI will be a key issue in upcoming Puget Sound contract negotiations with the current contract expiring October 2026. The union is pushing for guardrails on how Boeing deploys AI tools that could displace engineering work — marking the first time AI has been a formal bargaining issue in major aerospace labor talks.

cargo.one Acquires Cargofive, Launches AI-Native Freight OS

German air cargo platform cargo.one acquired ocean rate platform Cargofive backed by $20M+ from Bessemer Venture Partners, launching what it calls the first AI-native operating system for multimodal freight. The platform uses agentic workflows to coordinate pricing and booking across 4 million trade lanes spanning air and ocean cargo.

Big AI Headlines: Chroma Context-1 — Open-Source Agentic Search

Chroma released Context-1, a 20B parameter open-source search agent that decomposes queries into sub-queries, iteratively searches a corpus, and self-edits its context window to make room for new results. It matches frontier model retrieval at 10x the speed and 25x lower cost — directly applicable to aviation document retrieval like NOTAM parsing and regulatory search.

Big AI Headlines: Cohere Releases Open-Source Transcription Model

Cohere’s Transcribe is a 2B parameter open-source ASR model that tops the Hugging Face leaderboard with 5.42% word error rate across 14 languages, processing 525 minutes of audio per minute. High-accuracy multilingual transcription has clear aviation applications — ATC communications, cockpit voice analysis, and pilot-controller readback verification.

Sam here again - hope you’ve enjoyed reading this edition. If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact me at sam@deepskyai.com or visit deepskyai.com

Hi its Sam Chandra - CEO/Founder of Deepsky and the author of this newsletter. Here’s my take on todays news - which represents the last 48 hours of AI in Aviation happenings around the world:

The Flighty use of NOTAMs and weather is definitely something I’ve tried as a side project, but never had enough time to see it through. IAG asking for disruption management is no surprise - it pretty much is (and has been for a while) the grand prize of problems to solve for airlines. And the much higher accident rates when automation hands back to drivers in Teslas is a bit of a warning sign. We think about this exhaustively in aviation, and it seems like Tesla is playing catch up. (Not to mention the car probably hands back to the driver in a highly dynamic manoeuvre that is close to impossible to recover from…)

Flighty Launches AI-Powered Airport Intelligence Across 14,000 Airports

Flighty’s latest update brings pilot-grade data — METARs, TAFs, and NOTAMs — to everyday travelers, using AI to translate raw aviation advisories into plain-language disruption alerts and delay forecasts across 14,000 airports worldwide.

Collins Aerospace Unveils AutoTrac — A Unified AI-Ready ATC Platform

RTX’s Collins Aerospace pitched AutoTrac as the answer to the FAA’s $31.5 billion ATC overhaul, consolidating fragmented en-route, terminal, and surface systems into a single platform that deployed new capabilities in three months during international trials — versus the multi-year cycles of legacy systems.

IAG Targets AI for Disruption Management in 10th Accelerator Round

International Airlines Group opened its 10th IAGi innovation accelerator with AI-driven disruption management and scheduling optimization topping the wishlist — a signal that Europe’s largest airline group sees AI flight-ops tools as a near-term deployment priority, not a research project.

Apple Distills Gemini Into On-Device Models That Run Without Internet

Apple now has full access to Google’s Gemini in its own data centers and is using distillation to create smaller, cheaper models that run entirely on-device without cloud connectivity — a technique that could bring capable AI to iPad-based EFBs and cockpit tablets where connectivity is intermittent or unavailable.

Big AI Headlines: Tesla FSD Robotaxis Crash 4x More Than Human Drivers

Tesla’s Austin robotaxi fleet has logged 14 crashes in 800,000 miles — one every 57,000 miles versus one every 229,000 for human Tesla drivers — even with safety monitors present.

Sam here again - hope you’ve enjoyed reading this edition. If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact me at admin@deepskyai.com or visit deepskyai.com

Navi AI just raised $6M to scale AI-powered flight training — I’ve met both cofounders and they’re building something pretty special. The Saab-Cohere partnership caught my eye too; we’ve seen airlines partner with big AI companies before, but a manufacturer embedding LLMs directly into an airborne platform like GlobalEye is different — that gets AI much closer to the core of aircraft technology than we’ve seen. And on the Big AI side, Claude now has full computer use on macOS. AI controlling computers isn’t new — it’s been about 18 months — but coupled with Claude Code’s agentic harness and the latest models, this generation is going to be dramatically more effective than what came before.

Navi AI Raises $6M to Scale AI-Powered Pilot Training Debriefs

Navi AI has raised $6M from United Airlines Ventures, BVVC, and others — including a $1.27M U.S. Department of War contract — to scale its generative AI debriefing platform across flight schools and the military. The system ingests cockpit audio, flight data, weather, and traffic via an onboard orange box, then a domain-specific LLM generates 40-50 key insights per flight aligned to each school’s syllabus — already deployed at Embry-Riddle, Sling Pilot Academy, Purdue, and the USAF Test Pilot School, with ~100,000 flights processed.

Saab and Cohere Partner to Integrate AI into GlobalEye Surveillance Aircraft

Saab signed an MoU with Canadian AI firm Cohere to embed enterprise-grade AI into the GlobalEye airborne early warning platform, targeting data-driven mission support, maintenance tools, and rapid information processing in secure defense environments. The collaboration is tied to Canada’s expected ~$5B procurement of six surveillance aircraft, with benefits extending to existing international GlobalEye operators.

Alphabet’s Wing Drones Launch in San Francisco Bay Area

Alphabet’s Wing will begin drone deliveries in the San Francisco Bay Area this year, marking its first expansion into its home market after operations in North Carolina, Virginia, and Australia. Partnering with Walmart and DoorDash, each drone takes off vertically, cruises up to six miles, and lowers packages via a motorized tether — with Walmart items arriving in around 10 minutes.

NBAA SDC2026: AI and Advanced Air Mobility Reshape Scheduling and Dispatch

The NBAA Schedulers & Dispatchers Conference in Cleveland is hosting a session today on how AI, electric aircraft, and eVTOL infrastructure are transforming business aviation operations. Panelists are covering the eVTOL Integration Pilot Program across 26 states, Part 108 BVLOS rulemaking, and how automation will reshape — not replace — the scheduler and dispatcher role.

Big AI Headlines: Claude Gets Computer Use on macOS

Anthropic launched computer use for Claude on macOS, allowing the AI to open apps, navigate browsers, fill spreadsheets, and submit code — falling back to screen control when no API integration exists. The permission-first agentic approach, paired with mobile task assignment via Dispatch, signals where operational AI is heading: autonomous task completion with human oversight, a model directly applicable to aviation ops workflows from flight planning to maintenance documentation.

Why this matters: From AI debriefing student pilots to drones delivering lunch across Silicon Valley, the through-line is clear — AI is moving from advisor to operator, and the aviation industry is both the proving ground and the prize.

If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact us at admin@deepskyai.com or visit deepskyai.com

This week, an AI agent passed three of four human controller competencies in UK airspace, while autonomous combat software went from lab to live flight in a single day. Meanwhile, agentic AI is reading aircraft maintenance records faster than any human team — and the world’s biggest AI agent platform just hit a wall that aviation operators know all too well.

Project Bluebird: UK AI Air Traffic Control Goes Live This Spring

NATS, the Alan Turing Institute, and the University of Exeter are preparing live trials of an AI agent controlling traffic over London’s airspace in a digital twin. After 60+ hours of human assessments, the AI passed 3 of 4 controller competencies — the first time an AI system has been benchmarked against operational controllers at Swanwick. Open-source release of the digital twin platform is planned for April.

Northrop Grumman’s Talon IQ Flies Shield AI Hivemind in First-of-Kind Test

Shield AI’s Hivemind autonomy software flew aboard Northrop Grumman’s Talon IQ testbed over Mojave, executing combat air patrol and target engagement maneuvers before seamlessly handing control back to Northrop’s own Prism software. The system went from hardware-in-the-loop lab testing to live flight in a single day — proving plug-and-play interoperability under U.S. Government Reference Architecture standards.

Soji AI Brings Agentic Document Parsing to Aircraft Lease Transitions

Aircraft lessor Airborne Capital partnered with Hamburg-based Soji AI to apply agentic AI to the most labor-intensive part of lease transitions: technical records review. Soji’s aviation-trained platform reads, structures, and contextualizes unstructured maintenance records, airworthiness directives, and component histories — work that traditionally takes weeks of manual review by specialists.

Indra and Synaptic Aviation Deploy AI Turnaround Monitoring at Spain’s Busiest Airports

AI-driven video analytics are now live at Madrid-Barajas, Barcelona-El Prat, and Palma de Mallorca, covering 477 parking positions and 150M+ passengers annually. The system analyzes real-time apron camera feeds to automatically track turnaround events — GPU connection, chock placement, refueling, catering — improving predictability and reducing delays without major infrastructure changes.

Big AI Headlines: OpenClaw Hit Production — Then Hit a Wall

China is pushing OpenClaw adoption at national scale — Baidu and Tencent are running public onboarding events from gearheads to grandmas, and Tencent’s new ClawBot integrates it directly into WeChat. But developers are discovering that fully autonomous agents are fragile in production: context management breaks down, non-determinism creates unpredictable failures, and tooling can’t keep up. The emerging consensus is that the next wave belongs to vertical “agent harnesses” — domain-specific wrappers that package reliability, context, and infrastructure into usable products. For aviation, this is validation: the industry’s instinct to demand structured, auditable AI workflows over black-box autonomy isn’t conservative — it’s where the entire AI agent ecosystem is heading.

Why this matters: When the world’s most hyped AI agent platform learns the hard way that autonomy without structure doesn’t scale, it reinforces what aviation has always known: reliability beats capability. The question now is whether aviation-specific agent harnesses can move fast enough to capture the opportunity.

If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact us at admin@deepskyai.com or visit deepskyai.com

Merlin - a company that retrofits existing transport aircraft with autonomous AI pilots just went public at an $800M valuation, the U.S. Army took delivery of its first autonomous helicopter, and Bordeaux Airport figured out how to track passenger flows without identifying anyone. Your AI in aviation news from the past few days.

Merlin Goes Public on Nasdaq with $200M to Make Transport Aircraft Fly Themselves

Merlin began trading on Nasdaq under ticker MRLN at an $800M valuation, raising approximately $200M to scale its aircraft-agnostic "Merlin Pilot" — an AI autonomy stack that retrofits onto existing airframes to enable fully autonomous flight. The company has already demonstrated on King Air and Cessna Caravan platforms, holds a $105M USSOCOM contract to autonomize the C-130J Super Hercules, and is working with the U.S. Air Force on the KC-135 Stratotanker.

U.S. Army Takes Delivery of First Autonomous-Ready Black Hawk

The U.S. Army received its first H-60Mx Black Hawk equipped with Sikorsky’s DARPA-funded MATRIX autonomy suite, marking the first optionally piloted Black Hawk delivered for military testing. The fly-by-wire helicopter can operate with full crew, reduced crew, or completely autonomously, reducing pilot cognitive load and automating challenging maneuvers in low-visibility conditions.

USMC 2026 Aviation Plan Puts AI/ML at the Center of Sustainment

The newly released Marine Corps Aviation Plan makes AI and machine learning central to modernizing aviation readiness across three lines of effort: predictive maintenance that turns unscheduled failures into foreseen scheduled actions, dynamic AI-driven supply chains that adapt parts packages to deployment conditions, and optimized operations integrating previously siloed maintenance and scheduling systems.

Bordeaux Airport Deploys Physical AI to Predict Passenger Queues in Real Time

Bordeaux Airport partnered with Outsight to pilot a "Physical AI" system that builds a real-time 3D digital twin of passenger movements across Hall A — tracking queue formation, density changes, and spatial utilization at security checkpoints without capturing images or identifying individuals. The privacy-by-design system serves nearly 6 million passengers annually and aims to cut wait times through data-driven congestion prevention.

Why this matters: Merlin raising $200M is a sign that they and the wider retrofit autonomy game is ready to roll, while airports are proving that AI can improve the passenger experience without compromising privacy — signals that the industry is maturing in both capability and governance.

If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact us at admin@deepskyai.com or visit deepskyai.com

From AI-guided contrail avoidance over the Atlantic to autonomous drone swarms learning pursuit logic in Japan, this week's updates show AI moving from concept to cockpit — with the industry now grappling with how to govern it all.

Today's update has a special story for me — contrails. I interviewed the CEO of Satavia five years ago when they were developing the technology (before they sold to GE) so it’s good to see the same ideas being brought to fruition today in real operations (albeit with a different company. Check out the deepsky podcast if you want to listen to that episode.

American Airlines and Google Slash Contrail Warming by 69% Using AI Forecasts

Google's AI forecasting tool predicted where contrails would form on 2,400 transatlantic flights, and American Airlines integrated those predictions into its dispatch system — allowing pilots to shift altitude or routing to avoid them. The result: 62% fewer contrails and a 69% reduction in warming impact on participating flights, with no statistically significant fuel penalty.

MHI and Shield AI Complete Autonomous Drone Flight Tests in Japan

Mitsubishi Heavy Industries and Shield AI integrated Hivemind autonomy software into MHI's ARMD fixed-wing UAV and completed autonomous flight demonstrations in just eight weeks from first code to hardware validation. Two drones autonomously demonstrated reinforcement-learning behaviours, coordinated motions, and virtual-target pursuit — with the second flight showing more aggressive pursuit logic than the first.

Odysight.ai Reports Strategic Aerospace and Defense AI Expansion

AI predictive-maintenance company Odysight.ai reported a $13.8M backlog and first U.S. flight testing on UH-60 helicopters, alongside deployments on the Israeli Air Force's Heron TP UAV and SH-60 helicopter programs. Core PdM platform revenues grew 23% year-over-year as the company expands visual-sensing AI across combat helicopters, UAVs, and NASA space vehicles.

IATA Calls for "Human at the Helm" Approach to Agentic AI in Aviation

As agentic AI systems capable of autonomous multi-step decision-making enter aviation, IATA argues the traditional "human in the loop" model won't scale — there simply won't be enough humans to validate every outcome in real time. Instead, the industry body advocates a "human at the helm" governance framework, likening oversight to a board of directors that steers strategy rather than managing every task.

Why this matters: IATA actually realising that human-in-the-loop isn't going to work signals some real understanding finally making its way into the industry — which means it's a good time to get those AI conversations going with your team.

If you need help navigating the safe adoption of AI in your flight operation, then we have you covered. Contact us at admin@deepskyai.com or visit deepskyai.com