An artificial-intelligence-controlled X-62A VISTA experimental aircraft successfully intercepted crewed fighter jets during a series of flight tests, the U.S. Air Force has confirmed, marking one of the most operationally significant milestones yet in the service’s autonomous-air-combat program. The tests, first reported by Space Forces Magazine, demonstrate that AI agents can now perform an offensive air-to-air intercept mission against live adversary aircraft — not merely simulated targets — under real flight conditions. The advancement has direct implications for the Air Force’s broader Collaborative Combat Aircraft program, which aims to field uncrewed wingmen capable of operating alongside manned platforms in contested environments. Notably, the Pentagon AI marketplace effort underlines how broadly the Defense Department is pushing autonomous systems into operational roles across multiple domains.

The intercept tests were conducted at Edwards Air Force Base in California, home to the Air Force Test Pilot School, which has managed the X-62A program in partnership with DARPA under the Air Combat Evolution, or ACE, initiative. According to Air and Space Forces Magazine, the AI successfully maneuvered the X-62A to achieve an intercept geometry against the crewed jets, a task that requires split-second spatial reasoning, energy management, and threat response — capabilities previously considered to demand a human pilot in the loop.

ACE Program Reaches an Operational Threshold

The ACE program has been steadily building toward this moment. Earlier milestones included within-visual-range AI-versus-AI engagements and, in a widely publicized 2023 development, AI-versus-human simulated dogfights. Executing a real intercept against a crewed aircraft in actual airspace, however, represents a qualitative step beyond controlled simulations: the AI must respond to an aircraft flown by an experienced human pilot using unpredictable tactics, not a scripted scenario.

Air and Space Forces Magazine noted that a safety pilot remained aboard the X-62A throughout the tests, consistent with the program’s incremental approach to expanding AI authority. That safety pilot can intervene if the autonomous system behaves unexpectedly, preserving a human override during what officials still characterize as a developmental test phase. The Air Force has not released specific details on the number of intercept runs conducted, the aircraft types that served as targets, or the exact engagement parameters involved.

What the Results Mean for Collaborative Combat Aircraft

The broader significance of these tests lies in their relationship to the Air Force’s Collaborative Combat Aircraft, or CCA, program. The service intends to procure large numbers of autonomous uncrewed aircraft — currently represented by the Increment 1 designs from General Atomics and Anduril — that would fly alongside crewed fighters such as the F-35 and F-22, extending their reach and absorbing risk in high-threat environments. Demonstrating that an AI can reliably execute an intercept against a manned aircraft removes one of the central remaining question marks about whether such systems are technically credible for offensive air-to-air missions.

Officials have not confirmed a timeline for translating X-62A test results directly into CCA operational software, and the gap between experimental demonstration and program-of-record fielding remains significant. Still, the ACE program was designed explicitly to generate data and validated autonomy algorithms that could be transferred downstream to production systems. Each intercept run adds to that dataset. The Air Force’s willingness to publicize these results also signals institutional confidence that autonomous air combat is no longer a distant research ambition but an engineering and policy problem actively being solved. How quickly that translates into deployed capability will depend as much on acquisition decisions and U.S. AI development priorities as on the technical progress already achieved.