U.S. DARPA Advances AI Air Combat Autonomy on F-16 Testbeds for Future Unmanned Combat Aircraft
{loadposition bannertop}
{loadposition sidebarpub}
The U.S. Defense Advanced Research Projects Agency is advancing its Artificial Intelligence Reinforcements program to bring AI-enabled autonomy into multi-ship beyond-visual-range air combat, with new details published on February 26, 2026. By testing the system first on manned F-16s before transferring it to an unmanned combat aerial vehicle, DARPA is targeting faster tactical coordination across sensors, electronic warfare, maneuver, and air-to-air weapons in contested airspace.
AIR is designed to prove whether autonomous systems can support real combat decisions when radar tracks are incomplete, datalinks are delayed, electronic attack is active, and missile engagement windows collapse within seconds. The program reflects a broader U.S. push toward human-machine teaming and survivable air combat autonomy, where unmanned aircraft could extend reach, absorb risk, and strengthen allied airpower in future high-end conflicts.
Related topic: U.S. Set to Approve $700M GE F110 Engines for Türkiye KAAN Fighter Testing Despite S-400 Dispute.
DARPA’s AIR program aims to develop AI-enabled autonomy for F-16 test aircraft and future unmanned combat aerial vehicles, improving beyond-visual-range coordination of sensors, electronic warfare, maneuver, and AIM-120-class air-to-air missile employment (Picture source: DARPA).
DARPA’s public description places AIR in the continuation of Air Combat Evolution, the earlier effort that moved AI agents from simulation to the X-62A VISTA, a modified F-16D test aircraft operated at Edwards Air Force Base. Under ACE, AI algorithms flew within-visual-range engagements against a human-piloted F-16 in 2023 and 2024, with the U.S. Air Force reporting 21 test flights, more than 100,000 lines of flight-critical software changes, and close-range test conditions that reached approximately 2,000 feet of separation at 1,200 miles per hour. AIR moves that problem outward from dogfighting to BVR combat, where the decisive calculation is less about turn rate and more about track quality, missile kinematics, emissions control, cooperative targeting, and survivability after launch.
On an F-16 test aircraft, the relevant air-to-air armament is the AIM-120 Advanced Medium-Range Air-to-Air Missile family. Open U.S. Air Force data lists the AIM-120 as an all-weather BVR missile compatible with the F-15, F-16, F-22, F-35 development aircraft, and U.S. Navy F/A-18 variants. The missile uses inertial mid-course guidance and an active radar terminal seeker, allowing the launch aircraft to support the weapon during the early phase of flight and then maneuver defensively once the missile’s own radar can guide it to intercept. Official figures for the AMRAAM family include a 143.9-inch length, 7-inch diameter, 20.7-inch wingspan, 335-pound launch weight, blast-fragmentation warhead, and supersonic speed; the National Museum of the U.S. Air Force separately describes the missile as using a solid-fuel rocket motor, reaching Mach 4, and carrying a 40-pound high-explosive warhead.
Those figures explain why AIR is an autonomy program with direct armament relevance. A BVR missile shot is not defined only by maximum range. It depends on the launch aircraft’s speed and altitude, target aspect, closure rate, radar update quality, enemy jamming, the missile’s remaining energy at intercept, and whether the target can force the weapon outside its effective endgame. In practical terms, AIR is intended to help decide which aircraft should shoot, when it should shoot, which target should be prioritized, whether another aircraft should provide sensor support, and when the formation should break, recommit, or continue supporting the missile.
The F-16 is a logical test aircraft because it is numerous, well understood, and already cleared for the AIM-120. U.S. Air Force data lists the F-16 with an M61A1 20mm multibarrel cannon with 500 rounds and external stations able to carry up to six air-to-air missiles, air-to-surface weapons, fuel tanks, and electronic countermeasure pods. A typical air combat load may include AIM-9 short-range missiles, AIM-120 medium-range missiles, and external tanks, creating the same trade-offs that AIR must manage: fuel versus weapon load, radar use versus emissions control, and missile expenditure versus mission duration.
The program’s technical work is divided around two related problems. The first is modeling: creating fast, accurate representations of aircraft, sensors, electronic warfare effects, weapons, and adversary behavior under uncertainty. The second is execution: developing algorithms that can use those models in real time across a formation rather than as a single-aircraft autopilot. DARPA specifically identifies fully integrated sensors, larger engagement scaling, open-world adaptability, and predictive models that account for deceptive effects as unresolved barriers. That language is important because it acknowledges that BVR combat is not a clean geometry problem; it is a contest between sensor confidence, electronic attack, tactics, and weapons timing.
Industry contracts show how DARPA is approaching the problem. On July 8, 2024, Lockheed Martin announced a $4.6 million AIR contract to develop AI tools for dynamic airborne missions, including surrogate models of aircraft, sensors, electronic warfare, and weapons during an 18-month period of performance. On September 10, 2024, BAE Systems announced a $4 million Phase 1 contract for its FAST Labs organization to use machine learning to build simulation models of sensors, electronic warfare systems, weapons, and the physics of aerial maneuver in operationally representative conditions. These are modest contracts by procurement standards, but they are aimed at the hardest prerequisite for combat autonomy: a test environment realistic enough that tactical behavior learned in simulation has value in flight.
The operational use case is a mixed formation of crewed fighters and unmanned combat aircraft. One aircraft may carry the radar picture, another may remain electronically quiet, another may carry additional AIM-120-class weapons, and a fourth may act as a forward sensor or electronic attack asset. In that arrangement, the human pilot is not expected to manually control every maneuver by every aircraft. The more realistic model is mission command: the pilot or battle manager sets intent and constraints, while autonomy proposes or executes time-sensitive actions within approved rules.
This connects AIR directly to the U.S. Air Force Collaborative Combat Aircraft program. In June 2026, the Air Force awarded Increment 1 air vehicle contracts to General Atomics for FQ-42 and Anduril for FQ-44, while also selecting a pool of mission autonomy software vendors that includes Anduril, General Atomics, Lockheed Martin, Northrop Grumman, RTX Collins Aerospace, and Shield AI. The service also stated its intent to field more than 150 combat-capable CCA by the end of the decade and approximately 1,000 over time. The separation of mission autonomy software from air vehicle production is intended to keep software competition active after aircraft manufacturing begins.
AIR’s tactical value would be most visible in the opening phase of an air campaign against a defended opponent. An unmanned combat aerial vehicle could operate forward of crewed fighters, force enemy aircraft or ground radars to react, and create a more favorable launch position for AIM-120-class weapons without exposing a pilot to the same level of risk. It could also carry missiles as a distributed magazine, allowing a crewed fighter to control or authorize engagements while remaining farther from enemy surface-to-air missile zones. This does not remove the need for human command authority, but it changes where the human sits in the engagement cycle.
The main risk is not whether AI can fly an aircraft; ACE has already demonstrated controlled autonomous maneuvering in a safety-managed test environment. The unresolved question is whether AIR can produce tactical behavior that remains predictable, lawful, and useful when communications degrade, adversary behavior changes, and the aircraft must interpret ambiguous tracks. For Congress, NATO planners, and allied air forces, the program should be viewed less as a futuristic AI fighter effort and more as a practical attempt to preserve missile effectiveness, aircrew survivability, and combat mass as adversaries extend radar, missile, and electronic warfare coverage. In that sense, AIR is not a replacement for the F-16, AIM-120, or future unmanned combat aircraft; it is an effort to make those systems operate as a coordinated force under the compressed timelines of modern air combat.
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.
{loadposition bannertop}
{loadposition sidebarpub}
The U.S. Defense Advanced Research Projects Agency is advancing its Artificial Intelligence Reinforcements program to bring AI-enabled autonomy into multi-ship beyond-visual-range air combat, with new details published on February 26, 2026. By testing the system first on manned F-16s before transferring it to an unmanned combat aerial vehicle, DARPA is targeting faster tactical coordination across sensors, electronic warfare, maneuver, and air-to-air weapons in contested airspace.
AIR is designed to prove whether autonomous systems can support real combat decisions when radar tracks are incomplete, datalinks are delayed, electronic attack is active, and missile engagement windows collapse within seconds. The program reflects a broader U.S. push toward human-machine teaming and survivable air combat autonomy, where unmanned aircraft could extend reach, absorb risk, and strengthen allied airpower in future high-end conflicts.
Related topic: U.S. Set to Approve $700M GE F110 Engines for Türkiye KAAN Fighter Testing Despite S-400 Dispute.
DARPA’s AIR program aims to develop AI-enabled autonomy for F-16 test aircraft and future unmanned combat aerial vehicles, improving beyond-visual-range coordination of sensors, electronic warfare, maneuver, and AIM-120-class air-to-air missile employment (Picture source: DARPA).
DARPA’s public description places AIR in the continuation of Air Combat Evolution, the earlier effort that moved AI agents from simulation to the X-62A VISTA, a modified F-16D test aircraft operated at Edwards Air Force Base. Under ACE, AI algorithms flew within-visual-range engagements against a human-piloted F-16 in 2023 and 2024, with the U.S. Air Force reporting 21 test flights, more than 100,000 lines of flight-critical software changes, and close-range test conditions that reached approximately 2,000 feet of separation at 1,200 miles per hour. AIR moves that problem outward from dogfighting to BVR combat, where the decisive calculation is less about turn rate and more about track quality, missile kinematics, emissions control, cooperative targeting, and survivability after launch.
On an F-16 test aircraft, the relevant air-to-air armament is the AIM-120 Advanced Medium-Range Air-to-Air Missile family. Open U.S. Air Force data lists the AIM-120 as an all-weather BVR missile compatible with the F-15, F-16, F-22, F-35 development aircraft, and U.S. Navy F/A-18 variants. The missile uses inertial mid-course guidance and an active radar terminal seeker, allowing the launch aircraft to support the weapon during the early phase of flight and then maneuver defensively once the missile’s own radar can guide it to intercept. Official figures for the AMRAAM family include a 143.9-inch length, 7-inch diameter, 20.7-inch wingspan, 335-pound launch weight, blast-fragmentation warhead, and supersonic speed; the National Museum of the U.S. Air Force separately describes the missile as using a solid-fuel rocket motor, reaching Mach 4, and carrying a 40-pound high-explosive warhead.
Those figures explain why AIR is an autonomy program with direct armament relevance. A BVR missile shot is not defined only by maximum range. It depends on the launch aircraft’s speed and altitude, target aspect, closure rate, radar update quality, enemy jamming, the missile’s remaining energy at intercept, and whether the target can force the weapon outside its effective endgame. In practical terms, AIR is intended to help decide which aircraft should shoot, when it should shoot, which target should be prioritized, whether another aircraft should provide sensor support, and when the formation should break, recommit, or continue supporting the missile.
The F-16 is a logical test aircraft because it is numerous, well understood, and already cleared for the AIM-120. U.S. Air Force data lists the F-16 with an M61A1 20mm multibarrel cannon with 500 rounds and external stations able to carry up to six air-to-air missiles, air-to-surface weapons, fuel tanks, and electronic countermeasure pods. A typical air combat load may include AIM-9 short-range missiles, AIM-120 medium-range missiles, and external tanks, creating the same trade-offs that AIR must manage: fuel versus weapon load, radar use versus emissions control, and missile expenditure versus mission duration.
The program’s technical work is divided around two related problems. The first is modeling: creating fast, accurate representations of aircraft, sensors, electronic warfare effects, weapons, and adversary behavior under uncertainty. The second is execution: developing algorithms that can use those models in real time across a formation rather than as a single-aircraft autopilot. DARPA specifically identifies fully integrated sensors, larger engagement scaling, open-world adaptability, and predictive models that account for deceptive effects as unresolved barriers. That language is important because it acknowledges that BVR combat is not a clean geometry problem; it is a contest between sensor confidence, electronic attack, tactics, and weapons timing.
Industry contracts show how DARPA is approaching the problem. On July 8, 2024, Lockheed Martin announced a $4.6 million AIR contract to develop AI tools for dynamic airborne missions, including surrogate models of aircraft, sensors, electronic warfare, and weapons during an 18-month period of performance. On September 10, 2024, BAE Systems announced a $4 million Phase 1 contract for its FAST Labs organization to use machine learning to build simulation models of sensors, electronic warfare systems, weapons, and the physics of aerial maneuver in operationally representative conditions. These are modest contracts by procurement standards, but they are aimed at the hardest prerequisite for combat autonomy: a test environment realistic enough that tactical behavior learned in simulation has value in flight.
The operational use case is a mixed formation of crewed fighters and unmanned combat aircraft. One aircraft may carry the radar picture, another may remain electronically quiet, another may carry additional AIM-120-class weapons, and a fourth may act as a forward sensor or electronic attack asset. In that arrangement, the human pilot is not expected to manually control every maneuver by every aircraft. The more realistic model is mission command: the pilot or battle manager sets intent and constraints, while autonomy proposes or executes time-sensitive actions within approved rules.
This connects AIR directly to the U.S. Air Force Collaborative Combat Aircraft program. In June 2026, the Air Force awarded Increment 1 air vehicle contracts to General Atomics for FQ-42 and Anduril for FQ-44, while also selecting a pool of mission autonomy software vendors that includes Anduril, General Atomics, Lockheed Martin, Northrop Grumman, RTX Collins Aerospace, and Shield AI. The service also stated its intent to field more than 150 combat-capable CCA by the end of the decade and approximately 1,000 over time. The separation of mission autonomy software from air vehicle production is intended to keep software competition active after aircraft manufacturing begins.
AIR’s tactical value would be most visible in the opening phase of an air campaign against a defended opponent. An unmanned combat aerial vehicle could operate forward of crewed fighters, force enemy aircraft or ground radars to react, and create a more favorable launch position for AIM-120-class weapons without exposing a pilot to the same level of risk. It could also carry missiles as a distributed magazine, allowing a crewed fighter to control or authorize engagements while remaining farther from enemy surface-to-air missile zones. This does not remove the need for human command authority, but it changes where the human sits in the engagement cycle.
The main risk is not whether AI can fly an aircraft; ACE has already demonstrated controlled autonomous maneuvering in a safety-managed test environment. The unresolved question is whether AIR can produce tactical behavior that remains predictable, lawful, and useful when communications degrade, adversary behavior changes, and the aircraft must interpret ambiguous tracks. For Congress, NATO planners, and allied air forces, the program should be viewed less as a futuristic AI fighter effort and more as a practical attempt to preserve missile effectiveness, aircrew survivability, and combat mass as adversaries extend radar, missile, and electronic warfare coverage. In that sense, AIR is not a replacement for the F-16, AIM-120, or future unmanned combat aircraft; it is an effort to make those systems operate as a coordinated force under the compressed timelines of modern air combat.
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.
