[ Robotics & Physical AI
](https://www.unite.ai/series/robotics/)
[Add Unite.AI to your preferred sources on Google](https://www.google.com/preferences/source?q=unite.ai)
Lockheed Martin’s (LMT ) Skunk Works and the U.S. Air Force Test Pilot School have flown an AI agent through 27 live, sensor-guided fighter intercepts, the companies’ first flight-test series in which onboard targeting data, rather than simulated tracks, drove the autonomous maneuvers. Announced August 4, 2026, the campaign at Edwards Air Force Base ran across eight flights of the X-62 VISTA, the Air Force’s modified F-16 autonomy testbed, against a live T-38 target aircraft.
In each engagement, a Legion Pod infrared sensor mounted on the X-62 tracked the T-38 and fed targeting data to an AI agent, which then flew the fighter into a tactical intercept position on its own. Earlier X-62 autonomy work, including the program’s widely covered AI dogfights, ran against simulated target data; this series is the first time the aircraft’s autonomy closed what Lockheed calls the sensor-to-action loop using real-time streams from an operational sensor.
“Our autonomous agents consumed classified infrared search and track feeds and executed combat-critical maneuvers in real time,” said Ron Fehlen, vice president and general manager of Lockheed Martin Skunk Works, in the announcement.
The X-62’s three years of AI flight tests #
The intercept series extends a test lineage that has made the X-62 the Air Force’s central venue for flying AI on a tactical aircraft. On February 13, 2023, Lockheed disclosed that an AI agent had flown the VISTA for more than 17 hours, the first time AI engaged on a tactical aircraft. The jet, an F-16D Block 30 upgraded with Block 40 avionics, was built by Skunk Works with Calspan for the Test Pilot School and can mimic the flight characteristics of other aircraft through its variable-stability system.
In April 2024, DARPA confirmed that its Air Combat Evolution program had flown the X-62 in the first-ever in-air tests of AI algorithms autonomously flying an F-16 against a human-piloted F-16 in within-visual-range combat, with engagements at Edwards beginning in 2023. Those dogfights pitted the autonomy against manned F-16s, but the targeting picture came through test infrastructure rather than a combat sensor the way an operational jet would receive it. The August series is the first to hand the AI an operational sensor feed and let it own the intercept.
Lockheed’s autonomy work spans domains well beyond the X-62. The company took a $50 million stake in Saildrone in October 2025 to equip its autonomous surface vessels with defense payloads for the U.S. Navy, part of a broader push to move AI agents from test platforms toward operational fleets.
How the Legion Pod closed the loop #
The mechanism that changed in this series is the data source. In previous campaigns, the X-62’s AI agents maneuvered against target tracks generated or injected for the test, which validates the decision-making software but not the messy front end of combat sensing. Here, the Legion Pod, a Lockheed-built infrared search and track sensor carried in a pod under the jet, detected and tracked the T-38 in flight, and its secure feed went directly to the autonomy stack. The AI agent then flew the intercept itself, completing a chain from live sensor detection to autonomous control that Lockheed says mirrors the data environment pilots will face in future engagements.
Lockheed says the speed of integration is part of the result. Its “Supermassive” agent-generation pipeline took the AI agents from development to full integration and ground test with the X-62 in three months, a cycle the company positions as evidence that autonomy software can be refreshed at software pace rather than aircraft-modification pace. The company has made similar compression claims across its AI portfolio; Google’s robotics group made a parallel argument this summer when it shipped Gemini Robotics ER 2 with coordination skills pushed as a software update rather than new hardware.
What the upgrade enables next #
The series doubles as a shakedown for the X-62’s Mission Systems Upgrade, an overhaul funded by the Test Resource Management Center that Edwards detailed on December 18, 2025. The upgrade integrates advanced radar and sensors so the aircraft can run combat-system integration and airborne-agent collaboration in what the school describes as operationally relevant environments, and Lockheed’s announcement says the proven framework from this experiment will carry into mesh-networked demonstrations linking combat systems, sensors, and airborne AI agents.
The aircraft remains a Test Pilot School asset throughout, flying curriculum and research sorties while hosting industry and government test programs. With sensor-fed intercepts now flown, the next scheduled step is the upgraded jet itself: the Mission Systems Upgrade aircraft is expected to demonstrate combat-system and AI-agent integration within a next-generation mesh network, extending the X-62 from a single-ship autonomy testbed into a node for testing how autonomous fighters share a tactical picture.