Air Force Tests AI-Controlled Satellite in Orbit
The HAVE SCRAM exercise tested an AI agent managing a satellite's ground mission and proximity tracking of a simulated target.
Updated on Nov. 20, 2026 in Artificial Intelligence

In September 2026, the U.S. Air Force Test Pilot School, Air Force Research Laboratory, and The Aerospace Corporation tested AI-controlled satellite systems in an on-orbit exercise dubbed HAVE SCRAM. The test was reported as successful.
Why it matters
The exercise demonstrates that an AI agent can decide how a satellite divides its attention between competing tasks while operating in orbit. Testing such autonomy on real flight hardware is a step toward satellites that manage their own missions.
The test ran for two weeks on a 6U CubeSat using real flight software, processors, and hardware, with the AI agent keeping a simulated target below a stated uncertainty threshold. The source does not disclose the numeric threshold or how the system compared against non-AI control methods.
The players
U.S. Air Force Research Laboratory
The Air Force Research Laboratory is the service's principal scientific research organization, which designed the AI system and its guardrails for the HAVE SCRAM exercise.
The Aerospace Corporation
The Aerospace Corporation is a federally funded research and development center that arranged the ground testing, simulators, and AI watchdog used in the exercise.
The details
Participants built and tested scenarios on simulators using actual flight software and hardware before uploading them to a 6U CubeSat for live orbital testing. An AI agent trained by AFRL controlled the satellite's orientation and decided when and how to split attention between a primary ground observation mission and proximity tracking of a simulated target satellite.
Timeline
September 2026: HAVE SCRAM two-week exercise and on-orbit test conducted.
The Tech Race
HAVE SCRAM extends the Air Force Research Laboratory's autonomous spacecraft research from ground simulators into live on-orbit testing on a 6U CubeSat. It moves AI-driven satellite control from laboratory validation toward flight demonstration.
The test shows AI can manage satellite task priorities without constant ground control, which could change how future satellites handle multiple missions. Practical deployment for everyday users remains years away, since the source describes only a single successful two-week test.
The takeaway
Autonomous satellites that weigh competing tasks on their own are moving from simulation to orbit. Watch for follow-on testing that reports measured performance against conventional control methods.
Further reading
For more on how AI is being applied in defense and space systems, see Artificial Intelligence.
Source note: This article includes information reported by Aerospace Corporation.






