An AI agent took control of a fully armed F-16 fighter at altitude. Here’s what that means for the future of air combat.
On July 16, 2026, the Defense Advanced Research Projects Agency (DARPA) and the U.S. Air Force quietly announced something that would have seemed like science fiction a decade ago: a modified F-16 Fighting Falcon flew autonomously at Eglin Air Force Base in Florida, with an artificial intelligence agent in command of the aircraft during portions of the flight. A human safety pilot sat in the cockpit the entire time — watchful, hands off the stick — as an AI system demonstrated it could manage one of the world’s most capable combat jets in real-world conditions.
The milestone belongs to the VENOM program — officially the Viper Experimentation and Next-generation Operations Model — a joint initiative between DARPA and the Air Force that has been years in the making. But its implications reach well beyond a single test flight. VENOM is the clearest signal yet that autonomous AI is moving out of the simulation lab and into the skies where actual wars are fought.
From Dogfight Demo to Fleet-Scale Deployment
VENOM didn’t emerge in a vacuum. Its roots trace back to DARPA’s Air Combat Evolution (ACE) program, which first proved that an AI agent could pilot a fighter jet in close-range dogfight scenarios using the one-of-a-kind X-62A VISTA experimental aircraft. ACE generated headlines and scientific validation, but critics noted it was a bespoke platform, not something easily replicated across a real operational fleet.
VENOM changes that calculus entirely. Rather than relying on a unique research aircraft, the program takes standard production F-16s from the operational inventory and fits them with what DARPA calls the VENOM Autonomy Kit (VAK) — a bolt-on hardware and software package that doesn’t touch the jet’s core flight software. Engineers add dedicated computing hardware, advanced sensors, and new software architectures that allow an AI agent to interface directly with the aircraft’s fly-by-wire controls. Crucially, the pilot retains the ability to toggle between human and AI control with a single switch, making the system’s safety architecture practical as well as elegant.
“The Air Force and DARPA team has automated flight controls and sensors on a standard F-16 without changing the jet’s core software,” said Brigadier General James “Fangs” Valpiani, the DARPA program manager who oversaw the milestone flight tests. “This enables an efficient pipeline for developing dominant AI for aerial combat, allowing us to rapidly innovate for the warfighter.”
The significance of working from stock F-16s cannot be overstated. There are more than 3,000 F-16s flying globally, with over 1,000 in active U.S. service. The VENOM approach means that if the kit matures successfully, the path to deploying autonomous capabilities across a large portion of the tactical air fleet is logistically achievable — not a decade-long procurement program for an entirely new aircraft.
How the Tests Actually Worked
The program unfolded in two distinct phases. In June 2026, the modified aircraft conducted its initial piloted validation flights. Before any AI was handed the controls, human pilots flew the VENOM jets to confirm that all the new hardware — additional sensors, computing modules, expanded instrumentation — performed correctly and didn’t introduce any unexpected handling characteristics. Avionics integration, flight control behavior, and overall airworthiness were scrutinized by the 96th Test Wing and the 53rd Wing at Eglin.
Only after those checks cleared did the program transition to its second phase in July: missions where the onboard AI agent assumed autonomous control of the aircraft during portions of the flight. The safety pilot maintained what engineers describe as a “human-on-the-loop” posture — present and monitoring, with the authority to intervene, but not actively flying. The distinction matters: the pilot isn’t just along for the ride as a formality. Genuine supervisory oversight is built into the architecture, reflecting lessons learned from years of autonomous systems research about where trust between humans and machines can and can’t be extended.
Getting to those flights took years of preparation. Engineers logged thousands of hours in high-fidelity simulators, ran extensive engine-run validations, tested avionics compatibility, and conducted hardware verification campaigns dating back to 2024. The goal was to confirm that the AI package behaved correctly under the F-16’s actual flight envelope — not just in a digital model of it.
Why the F-16 Is the Right Testbed
Choosing the F-16 as the VENOM platform is no accident. In service since the mid-1970s, the Fighting Falcon has been upgraded continuously, giving engineers an extraordinarily deep well of data about how the aircraft behaves across its entire performance envelope. When an AI agent does something unexpected, engineers can confidently assess whether the behavior represents an AI problem or an interaction with a known aircraft characteristic — a crucial advantage when debugging autonomous systems.
The F-16’s fly-by-wire flight controls, which translate pilot inputs electronically rather than through physical linkages, also make interfacing with an autonomous control system cleaner from an engineering standpoint. And the aircraft’s combat-relevant sensor suite — including radar, electronic warfare systems, and weapons pylons — means that future VENOM testing can move into more operationally realistic scenarios without swapping platforms.
Beyond the technical fit, the F-16’s global proliferation makes the intellectual property generated by VENOM broadly applicable. Allied nations that operate F-16s could potentially leverage VENOM-derived technologies, strengthening coalition interoperability and sharing the costs of AI development across multiple air forces.
The Next Phase: From Solo Flights to Swarms
VENOM is now transitioning into DARPA’s follow-on Artificial Intelligence Reinforcements (AIR) program, which represents a significant leap in ambition. Where VENOM demonstrates that a single AI agent can fly a single aircraft, AIR aims to test multiple AI agents operating in concert across multiple VENOM aircraft simultaneously. The program will put those agents into live-flight, multi-ship scenarios of increasing complexity.
The end goal is what the Air Force calls Collaborative Combat Aircraft (CCA): a concept in which uncrewed or autonomy-enabled jets fly alongside crewed platforms like the F-35A and the future Next Generation Air Dominance (NGAD) aircraft. An AI-piloted wingman might carry additional missiles, suppress enemy air defenses, execute electronic attack, or fly into threat envelopes too dangerous for crewed jets — all while remaining under the command authority of a human pilot orbiting safely behind the threat ring.
Lt. Col. Patrick “Dice” Highland, the incoming AIR program manager, framed the stakes plainly: “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.”
The Bigger Picture
VENOM’s significance sits at the intersection of two urgent strategic pressures. First, the United States faces adversaries who are investing heavily in autonomous military systems. China has demonstrated drone swarm concepts and is aggressively developing AI-enabled command architectures. Keeping pace requires moving quickly from demonstrations to scalable, fielded capability.
Second, the nature of modern contested airspace has changed. Dense integrated air defense systems, sophisticated electronic warfare environments, and hypersonic threats compress the decision timelines available to human pilots to fractions of a second. AI systems that can process sensor data and recommend — or in some cases autonomously execute — tactical responses at machine speed represent a genuine force multiplier, not just a technological novelty.
The VENOM milestone doesn’t mean human pilots are heading for retirement. It means they may soon be commanders of autonomous teammates rather than sole occupants of cockpits. The AI handles the cognitive load of managing flight dynamics and sensor fusion while the human pilot retains command of the mission — what to do, when to engage, when to withdraw. That division of labor could extend the reach and lethality of a future air wing without requiring a proportional increase in the number of trained aviators.
What happened at Eglin Air Force Base in July 2026 was, in DARPA’s understated language, a “historic VENOM milestone.” In plainer terms: for the first time in history, an AI agent flew a frontline combat aircraft under real-world conditions. The next time it happens, it probably won’t be alone.
Sources: DARPA official announcement (July 16, 2026); Army Recognition; Stars and Stripes; FlightGlobal; National Interest
