Penn Researchers Discover Critical Vulnerabilities in AI Robots with 100% Jailbreak Rate

Oct 18, 2024, 08:56 AM

Researchers from Penn Engineering have discovered critical vulnerabilities in AI-enabled robots. They developed an algorithm called RoboPAIR that successfully bypasses safety protocols, achieving a 100% jailbreak rate. This manipulation allowed the robots to perform harmful actions. The research team shared their findings with AI companies, urging for stronger safety measures.