Future TechnologyFuture Technology
AI

"DARPA's AI Just Flew a Real F-16, and a Human Could Take Over With One Switch"

· By Future Technology

Key takeaways

  • DARPA and the US Air Force flew a modified combat spec F-16 under AI control at Eglin Air Force Base in June 2026
  • The VENOM Autonomy Kit lets a human pilot toggle between AI and manual control mid flight
  • The AI handled flight controls and mission systems simultaneously
  • This follows an earlier AI versus AI dogfight test in a separate aircraft

An AI system flew a fully specced combat F-16, not a stripped down test surrogate, at Eglin Air Force Base in June 2026. The flight was disclosed in July by DARPA and the US Air Force, and it marks the first time AI has autonomously controlled a real combat aircraft rather than a dedicated research jet.

The system behind it is called VENOM, short for Viper Experimentation and Next-gen Operations Model. It is built as an autonomy kit that sits alongside the existing cockpit controls rather than replacing them. A human safety pilot stayed in the cockpit throughout the flight, with a simple toggle switch to move control back and forth between AI and manual.

What the AI actually did

During the test, the AI system handled both flight controls and mission systems at the same time, the kind of multitasking that normally requires a pilot and, in many aircraft, a weapons systems officer working together. This followed an earlier milestone: an AI-versus-AI dogfight test conducted in a separate test aircraft, which proved the system could handle combat manoeuvring before it was ever trusted with a production-spec jet.

Why the toggle switch matters more than the AI

The most interesting design choice here is not the AI itself. It is the switch.

Rather than building a fully autonomous aircraft with no human override, DARPA built a system where a pilot can hand off routine flying to the AI and take it back instantly the moment something looks wrong. That is a deliberate, conservative approach to a genuinely high-stakes problem: nobody wants an armed aircraft making decisions nobody can interrupt.

It also hints at how autonomous military aviation is likely to scale over the next decade. Instead of a single leap to unmanned fighters, the more realistic path looks like this: AI handles routine operations such as navigation, formation flying and basic mission tasks, while humans retain control for anything involving weapons release or high-stakes judgement calls. VENOM is effectively a working prototype of that division of labour, tested in the air rather than in a simulator.

The bigger picture

Autonomous fighter programmes have existed for years in simulation and on dedicated test aircraft. What changed with VENOM is that the AI flew a combat-configured F-16, the actual aircraft type flown operationally by air forces around the world, not a purpose-built research platform. That closes the gap between "AI can fly a plane" and "AI can fly the plane a squadron actually uses."

Whether that capability ends up in frontline service, and how quickly, will depend as much on doctrine and trust as on the technology itself. But the toggle switch on this F-16 is a good preview of where the debate is heading: not human versus machine, but how fast a human can take the controls back.

Source: [DARPA](https://www.darpa.mil/news/2026/darpa-us-air-force-fly-ai-controlled-f-16)

Sources