AI Takes the Stick: DARPA and Air Force Fly Autonomous F-16

DARPA and the U.S. Air Force have achieved a significant milestone: an AI agent autonomously flew an F-16 fighter jet. The flight took place at Eglin Air Force Base in June 2026, as part of the Viper Experimentation and Next-generation Operations Model (VENOM) program. This is not a one-off experiment; it's the beginning of a pipeline to develop and field trusted autonomous combat AI.

The VENOM Autonomy Kit (VAK)

The key technical achievement is the VENOM Autonomy Kit (VAK), a hardware/software interface that connects to the F-16's flight controls and mission systems without modifying the jet's core software. This is a critical design decision: it means any standard F-16 can be converted into an autonomous testbed without requiring a complete avionics overhaul. The kit allows a pilot to toggle between human and AI control with a single switch, maintaining a human-on-the-loop safety model.

This builds on earlier work with the X-62A VISTA, a one-of-a-kind modified F-16 that proved an AI agent could dogfight. VENOM takes that proof and scales it to fleet aircraft. Brig. Gen. James "Fangs" Valpiani, DARPA program manager, stated: "The Air Force and DARPA team has automated flight controls and sensors on a standard F-16 without changing the jet’s core software. This enables an efficient pipeline for developing dominant AI for aerial combat."

From ACE to AIR: The Roadmap

VENOM is the operational testbed for the next phase: the Artificial Intelligence Reinforcements (AIR) program. AIR will use the VENOM fleet to test multiple AI agents in live, multi-ship scenarios. The goal is to enable human pilots to command teams of autonomous uncrewed aircraft. This directly feeds into the Collaborative Combat Aircraft (CCA) programs, which aim to field loyal wingman drones.

Lt. Col. Patrick "Dice" Highland, the incoming AIR program manager, said: "We now have the opportunity to create dominant autonomy for beyond-visual-range, multi-ship combat. These flights give us an early glimpse of how AI agents may begin actively transforming air warfare."

Technical Details from the Source

The source article provides specific metrics and quotes:

  • The VENOM aircraft are modified F-16s from the standard operational fleet.
  • The VAK was designed by multiple performers under the DARPA ACE program.
  • The interface uses a novel approach to integrate with existing flight controls.
  • Testing is taking place at Eglin Air Force Base, with the 96th Test Wing.
  • Terry Wilson, Ph.D., from AFRL's ACT3 team, serves as deputy program manager ensuring continuity.

Why This Matters for Developers

This is not just military hardware news. The underlying patterns are directly applicable to building reliable AI systems:

  1. Human-on-the-loop architecture: The VAK design allows a human to override at any time. This is analogous to a "kill switch" in production AI systems. Developers building autonomous systems (e.g., self-driving cars, robotics) should study this pattern.

  2. Retrofitting legacy systems: The kit interfaces with existing flight controls without core software changes. This is a model for adding AI capabilities to existing codebases without a full rewrite. The interface pattern (toggle between AI and human control) is a clean abstraction.

  3. Scaling from simulation to reality: The AIR program explicitly moves "capabilities out of simulation and into the sky." This mirrors the challenge of deploying ML models from training to production. The VENOM/AIR pipeline—simulation, then single-agent flight, then multi-agent scenarios—is a solid MLOps approach.

What's Next

The AIR program will continue testing through late 2026 and beyond. The immediate next steps involve testing multiple AI agents in beyond-visual-range combat scenarios. For developers, the key takeaway is that the DoD is investing heavily in trusted autonomy. Expect open-source tools and frameworks to emerge from these programs, similar to how DARPA's earlier work led to technologies like the internet and GPS.

If you're working on autonomous systems, watch for publications from AFRL's ACT3 team. They will likely share patterns for safe AI integration and human-on-the-loop control. The code may be classified, but the architectural lessons are public.