USAF Demonstrates AI-Controlled Fighter Intercepts

Lockheed Martin’s Skunk Works division has executed multiple AI-driven air intercepts against a crewed jet, using what it terms “sensor-driven autonomy” aboard the U.S. Air Force’s X-62A Variable In-flight Simulation Test Aircraft (VISTA).

Announced on August 4, the campaign saw Skunk Works, the U.S. Air Force Test Pilot School, and industry partners collaborate to engage a piloted Northrop T-38 Talon above Edwards Air Force Base, California. Across eight flights, an artificial intelligence agent aboard the X-62A used targeting information from an operational sensor to fly the airplane into what Lockheed called “a tactical intercept position.” In all, the aircraft completed 27 intercepts.

According to Lockheed Martin Skunk Works v-p and general manager Ron Fehlen, the test campaign demonstrated how AI can “effectively and reliably close the sensor-to-action loop aboard an operational combat aircraft.” In practical terms, Lockheed’s externally mounted Legion Pod multi-function sensor system made this possible, feeding “classified infrared search and track feeds” to the onboard AI agents in real time. Skunk Works generated those agents with an in-house capability it calls “Supermassive” and said it completed “full integration and ground test of the agents with the X-62” in just three months.

“This project demonstrates how sensing and autonomous AI can come together as a force multiplier to make faster, more informed action in complex environments,” said Lockheed Martin Sensors and Global Sustainment v-p and general manager Stacy Kubicek. By delegating complex tasks to AI, Lockheed argues, pilots gain bandwidth to focus on the tactical picture—with crucial implications for the effectiveness and survivability of future AI-enabled missions.

The Air Force identified the campaign as “Have Heat”, one of two test programs the Test Pilot School’s research division flew on the X-62A in April. “Have Heat represents a meaningful expansion of avionics capability toward integrated, AI-driven control of multiple sensors and air vehicles for mission autonomy,” said Test Pilot School dean of faculty for research Lt. Col. Joshua Strafaccia. Its companion program, “Have Holidays,” tested modular technologies on the X-62A, among them a third-party autonomous agent and enhanced autonomy safety rules.

A one-of-a-kind airplane, the X-62A VISTA is derived from the F-16. “Built on our open systems architecture, VISTA is modified to test artificial intelligence and autonomy capabilities that [allow] it to mimic the performance characteristics of other aircraft,” Lockheed said. Skunk Works developed it with Calspan for the Test Pilot School, and the Air Force redesignated it from NF-16D to X-62A in June 2021.

The recent intercepts build on what DARPA termed a world first when, in 2023 and 2024, AI algorithms controlled the X-62A against a human-piloted F-16 in a within-visual-range combat scenario.