The book Four Battlegrounds: Power in the Age of Artificial Intelligence discusses the current "struggle to control artificial intelligence" penned by former Army Ranger and Pentagon policy analyst Paul Scharre.
The book tells how DARPA spent a week with a group of Marines at a test site in order to help train an AI robot by attempting to defeat it. The robot was parked "in the middle of a traffic circle" and the Marines had to approach it undetected. Scharre explains, "If any Marines could get all the way in and touch this robot without being detected, they would win."
Apparently, the AI on the robot was only any good at detecting humans walking and all the marines could attack it. Scharre explains how two of the Marines "somersaulted for 300 meters," another two hid under a cardboard box, and another "stripped a fir tree and walked like a fir tree" [how many fir trees have you seen walking? ed]
The marines thought it was a laugh and despite giggling throughout the test, the robot failed to detect them.