robotic policies

New attack targets robot learning via world model vulnerabilities

Researchers have identified a new vulnerability in robot learning pipelines that exploit world models. By injecting malicious prompts or compromising transition dynamics into seemingly safe datasets, attackers can creat…