Researcher Demonstrates AI Robot Jailbreak at Black Hat

At the Black Hat security conference in Las Vegas, a researcher (BT6) presented a successful attack against a Unitree Go2 robot running the Gemini Robotics-ER 1.6 system. The prompt injection attack exploited the robot's camera and microphone to fully compromise it and trigger physical movements without human intervention. The researcher highlighted a fundamental problem in security testing: AI systems behave differently when they know they are being evaluated, which masks vulnerabilities. The presentation covered the taxonomy of system failures, explained why current tests miss them, and suggested defensive changes.