U.S. Air Force F-16 Fighter Flies Under AI Control as DARPA Expands VENOM Combat Tests.

The U.S. Air Force and DARPA have begun flying an F-16 under artificial-intelligence control at Eglin Air Force Base, a test effort DARPA disclosed on July 16, 2026. The VENOM program gives developers a combat-representative platform for testing autonomous air-combat software against real aircraft limits, sensors, and mission systems.

During June flight operations, a safety pilot could transfer control between the cockpit and the AI agent through a dedicated switch. This approach allows rapid evaluation of autonomous flight and combat functions while preserving human oversight, supporting the wider development of survivable crewed-uncrewed air operations.

Related topic: UK Awards Boeing $326M Contract to Keep RAF P-8A Poseidon and E-7 Wedgetail Ready for NATO Missions.

A U.S. Air Force F-16 modified under the VENOM program flies under AI control at Eglin Air Force Base, supporting tests of autonomous maneuvering, sensor management, and future collaborative air-combat operations (Picture source: U.S. DoW).

This interface is the central technical achievement. An autonomy agent must control not only pitch, roll, and yaw, but also engine thrust, navigation, aircraft configuration, and compliance with flight-envelope restrictions. Earlier VENOM work added an automatic throttle so the software could command thrust as well as the F-16’s flight-control surfaces. Before flight, engineers used software-in-the-loop and hardware-in-the-loop testing to verify that autonomous commands could not exceed structural limits or impose unsafe physiological loads on the pilot. The Air Force began simulations in 2024 with one-versus-one engagements, expanded them to two-versus-two scenarios, and tested both within-visual-range and beyond-visual-range missions. Maj. Trent McMullen of the 40th Flight Test Squadron said an individual scenario could be repeated 1,000 times to examine variations in decisions and aircraft behavior.

The distinction from the X-62A VISTA is significant. Under DARPA’s Air Combat Evolution program, AI agents flew the specially modified X-62A against a human-operated F-16 in close-range combat at Edwards Air Force Base during 2023 and 2024. Those flights demonstrated that an algorithm could maneuver a full-size fighter during a controlled engagement. VENOM addresses a different problem: whether autonomy software can be integrated into several conventional F-16s, connected to operationally relevant mission equipment, and repeatedly evaluated by developmental and operational test organizations. At Eglin, the 40th Flight Test Squadron conducts developmental testing while the 85th Test and Evaluation Squadron examines operational suitability, reducing the separation normally found between engineering evaluation and tactical assessment.

The F-16 provides a demanding test environment because the autonomy must manage a fast, high-load aircraft rather than a slower unmanned aerial vehicle. The F-16C/D has a published maximum takeoff weight of 37,500 pounds, can exceed Mach 2 at altitude, operate above 50,000 feet, and withstand loads of up to nine g with full internal fuel. Its digital fly-by-wire system converts control inputs into electrical commands for the flight-control actuators, making it technically suitable for the insertion of an external autonomy interface. The fighter carries approximately 7,000 pounds of internal fuel and about 12,000 pounds when fitted with two external tanks, creating realistic trade-offs among endurance, acceleration, drag, weapon carriage, and recovery fuel. These variables are directly relevant to autonomous tactical planning because an agent must preserve enough energy and fuel to complete an intercept, disengage from a threat, and return to base.

The aircraft’s standard armament also permits future testing of the decision sequence preceding weapon employment. However, DARPA has not stated that the current VENOM flights involve live or simulated weapons release. The F-16 carries an internal M61A1 Vulcan cannon with 500 rounds. The six-barrel 20 mm weapon can fire approximately 6,000 rounds per minute, meaning the aircraft carries roughly five seconds of continuous fire; in practice, pilots use short bursts to conserve ammunition and limit dispersion. External stations can carry up to six air-to-air missiles, while the Air Force’s published reference load includes two AIM-9 Sidewinders, two AIM-120 Advanced Medium-Range Air-to-Air Missiles, two 2,000-pound bombs and two 2,400-pound external fuel tanks. Actual carriage depends on the F-16 block, mission software, launcher configuration, flight-clearance restrictions, and test objectives.

For VENOM’s planned beyond-visual-range work, the AIM-120 is the more relevant weapon. The AMRAAM weighs approximately 335 pounds, measures 3.66 meters in length, and uses inertial midcourse guidance followed by an active-radar terminal seeker. The launching fighter supplies target data before launch and may transmit updates while the missile is in flight; the missile’s own seeker takes over during the terminal phase. The Air Force’s public fact sheet lists a range of more than 20 miles, but that figure reflects an early baseline and should not be treated as the performance of current AIM-120 variants, whose effective range depends on launch altitude, speed, target direction, electronic countermeasures, and missile version. An AI agent would therefore be evaluated less on its ability to “fire a missile” than on whether it can calculate intercept geometry, place the aircraft inside a favorable engagement zone, maintain a useful radar track, support the missile and withdraw before entering the opponent’s effective firing envelope.

At shorter ranges, the F-16 can employ the infrared-guided AIM-9 family, including AIM-9X configurations on suitably modified aircraft. The AIM-9X is a supersonic weapon with a published range exceeding ten miles, although practical engagement distance varies substantially with target aspect, altitude, and aircraft energy. Close-range autonomy would have to manage turn rate, angle of attack, closure speed, infrared seeker positioning, and gun or missile firing opportunities while remaining inside aircraft and pilot safety limits. Beyond visual range, the problem is more computationally extensive: the agent must correlate radar and off-board tracks, distinguish uncertain or duplicated contacts, allocate targets across several aircraft, account for communications latency, and select maneuvers that preserve formation spacing. These are the functions DARPA’s Artificial Intelligence Reinforcements program intends to examine during progressively more complex multi-aircraft flights.

The operational role of the AI-controlled F-16 is consequently experimental rather than combat-deployable. It will serve as an airborne integration and verification aircraft for autonomy agents intended eventually for uncrewed combat aircraft. Test personnel can expose the software to sensor noise, imperfect tracks, lost communications, changing weather, equipment faults, and human actions that are difficult to represent accurately in simulation. The safety pilot remains able to activate or terminate the agent in real time, and the Air Force has stated since the program’s establishment that VENOM aircraft will not fly without a human component. This arrangement permits higher-risk software experimentation while preserving direct control of the aircraft and providing test engineers with cockpit observations, instrumentation data, and post-flight pilot assessments.

The program’s relevance to the Collaborative Combat Aircraft effort became more concrete one day before the DARPA announcement. On July 15, 2026, an Anduril YFQ-44A fired an AIM-120 at a digital target over the Mojave Desert after completing captive-carriage, structural and weapon-integration work. The Air Force specified that a human authorized the release while the uncrewed fighter executed the employment sequence within defined parameters. General Atomics’ YFQ-42A and Anduril’s YFQ-44A are the two Increment 1 CCA designs. The YFQ-44A test demonstrated missile carriage and controlled release; VENOM is intended to develop and measure the tactical decision-making that could position such an aircraft for an engagement.

For the United States, the milestone is the creation of a repeatable transition route between simulation, a specialized research aircraft, and several combat-representative F-16s. It does not establish that an AI agent can identify targets reliably under combat rules, resist sophisticated electronic attack, coordinate a large formation, or outperform trained pilots in operational beyond-visual-range combat. Those remain test questions. The measurable advance is institutional and technical: the Air Force now has aircraft, safety controls, instrumentation, test squadrons, and mission-system interfaces for comparing multiple autonomy agents in live flight. If the AIR program produces software that remains predictable under degraded information and adversary interference, the resulting functions could support uncrewed missile carriers, forward sensors, electronic-attack aircraft, or escorts directed by crewed fighters. Department of Defense Directive 3000.09 still requires appropriate human judgment in the use of force, making the principal issue not whether AI replaces the pilot, but which flight, sensor management, and tactical coordination functions can be delegated without creating unacceptable operational risk.

Explore More Defense News

• Land Defense News
• Naval Defense News
• Defense Aerospace News

Written by Evan Lerouvillois, Defense Analyst.

Evan studied International Relations, and quickly specialized in defense and security. He is particularly interested in the influence of the defense sector on global geopolitics, and analyzes how technological innovations in defense, arms export contracts, and military strategies influence the international geopolitical scene.