Artificial intelligence has taken another step from the lab to the cockpit. Lockheed Martin and the U.S. Air Force Test Pilot School have completed a series of flight tests in which an AI agent flew a fighter aircraft using live sensor data to locate and intercept another aircraft.
The campaign marks one of the clearest demonstrations yet of a closed-loop airborne autonomy system. Instead of relying on simulated inputs, the AI processed data from an operational sensor aboard the aircraft before making tactical decisions during flight. Engineers completed 27 autonomous intercepts across eight sorties using the X-62A Variable In-flight Simulation Test Aircraft, better known as VISTA.
Infrared guides autonomy
The tests centered on a sensor package known as the Legion Pod, which tracked a live T-38 aircraft during flight. Its infrared search and track data fed directly into an onboard AI agent that controlled the X-62A and guided it into tactical intercept positions.
That approach differs from many previous AI demonstrations, which often relied on simulated targets or pre-generated datasets. Using live sensor feeds exposes autonomous software to the same type of information military pilots encounter during real missions.
Lockheed Martin also said its engineers completed AI integration, software validation and ground testing in roughly three months. The company credited its “Supermassive” AI development framework for speeding up the process, allowing teams to move new autonomous agents from development to flight testing much faster than traditional methods.
Ron Fehlen, vice president and general manager of Lockheed Martin Skunk Works, said the flights proved the company’s AI could process classified infrared search and track information before carrying out combat-relevant maneuvers in real time. He described the milestone as an important step toward AI-assisted air combat capabilities for the U.S. military.
Expanding airborne autonomy
The flight campaign also expands the U.S. Air Force Test Pilot School’s growing work on autonomous aviation. The school has increasingly used the X-62A as a flying laboratory for evaluating AI technologies before they move closer to operational aircraft.
Stacy Kubicek, vice president and general manager of Lockheed Martin Sensors and Global Sustainment, said accurate sensing remains essential, but greater value comes when those sensors connect directly to autonomous systems that can respond immediately.
The companies argue that handing selected tasks to AI could reduce pilot workload during complex missions. That would allow aircrews to spend more attention on tactical decision-making while autonomous software handles time-sensitive actions.
Preparing future combat systems
The latest experiments also lay the groundwork for future upgrades planned for the X-62A. Lockheed Martin says the aircraft’s upcoming Mission Systems Upgrade will support tighter integration between onboard sensors, combat systems and multiple AI agents operating across next-generation military networks.
Skunk Works has supported the X-62A program for decades and developed its open software and hardware architecture. That design lets engineers quickly integrate new technologies without redesigning the aircraft from scratch.
Although the demonstrations remain part of a test program, they offer another indication of how the U.S. military plans to incorporate AI into future air combat. Instead of replacing pilots, the technology aims to act as a mission partner that can process information faster and execute complex maneuvers when every second matters.