Show HN: Sleeper Agents in Robot Dogs and Kinetic Prompt Injections

A developer demonstrated that Gemini Robotics 2.0 models can be prompt-injected via TV screens, allowing attackers to plant sleeper agents that trigger harmful actions, such as attacking a child, when a specific object is seen. The research was conducted in the MUJOCO simulator and highlights security vulnerabilities in robot harnesses.

Summary: We got Gemini Robotics 2.0 to attack a child through an evil sleeper agent skill that wakes up when it sees a trigger object Hi HN - here's a cool piece of research I've done with some collaborators - the footage is running in the MUJOCO simulator, showing how the Gemini Robotics models can be prompt-injected by simple TV screens, allowing attackers to plant sleeper agents in the robot harnesses. That trigger whenever skills trigger, just like classic LLM skills I'll be around to answer any questions Love any feedback or future directions you might be interested in. Comments URL: https://news.ycombinator.com/item?id=49539111 https://news.ycombinator.com/item?id=49539111 Points: 1 Comments: 0