The U.S. Air Force used Lockheed U-2 spy aircraft navigation and sensor systems during a training flight on Tuesday, according to Washington Air. This flight is the first when the US has used AI to control a military aircraft.

“This is the first time this has happened,” Assistant Secretary of the Air Force Will Roper told The Washington Post.

In the present cases, AI ensured the operation of specific features and never controlled the flight of an aircraft, or any armed system, and during this test flight, it was instructed to use the aircraft radar sensors to search for missiles and their launch devices. During the flight, the algorithm had complete control over the sensors.

The system was deliberately designed without manuals and extra conditions to “provoke thinking and learning in a trial environment quickly,” Air Force spokesman Josh Bendetti told The Washington Post.

This algorithm used publicly available modules developed, for example, by AI company DeepMind. It also used Kubernetes, Google-owned software, to enable the algorithm to work with on-board computer systems. The purpose of the test flight was to test the military to see if they would be able to interact with similar systems in the future, as well as, of course, to see the potential of AI.

“This flight was aimed at modernizing the Air Force and the Department of Defense as a whole and seeing how seriously we need to look at AI potential,” Roper told The Washington Post.