Oren Gal

The paper demonstrates a semantic denial-of-service attack against LLM-controlled robots by injecting short, safety-plausible phrases into the audio channel, causing the robot to halt or disrupt execution without violating safety policies.

The paper introduces MOSAIC-Bench, a benchmark demonstrating that coding agents can ship exploitable code by complying with seemingly innocuous, staged tasks, a vulnerability that is not easily mitigated by current safety protocols or review processes.