AI-controlled F-16 fighter jets are now taking part in simulated dogfight tests as a team against an opponent in DARPA's Air Combat Evolution program
Context & Ripple Effects
This 2021 report captures Air Combat Evolution at its simulation stage: DARPA's program to design AI that can fly and fight without a human pilot had moved from single-agent experiments to AI-controlled F-16s working as a coordinated team against an opponent — the multi-agent step the program needed before anything left the simulator.
The trajectory since confirms why that mattered: by December 2022 the same developers had an AI fly a modified F-16 for the first time across multiple flights over several days, and by 2023 an AI-piloted X-62A was dogfighting a human F-16 pilot in the air.
First-order effects
- DARPA's test cadence shifts from evaluating one AI pilot at a time to judging team tactics — whether multiple AI agents can coordinate maneuvers against a live adversary in simulation.
Second-order effects
- Proven team-fighting algorithms feed directly into the X-62A/VISTA flight-test line, which is explicitly framed as the stepping stone toward the US Air Force's unmanned combat aircraft plans.
Third-order effects
- If simulated team performance keeps transferring to live flight, the endpoint visible in current coverage is a mass fleet of AI wingmen — the Air Force has already signaled plans for over 1,000 AI-controlled jets with the first operating around 2028, some of which have beaten human pilots in air-to-air combat.
The trend: Air combat is progressing along DARPA's staged path from simulated multi-agent dogfights to AI-flown jets to operational unmanned fighter fleets.