Sant’anna School Of Advanced Studies

AI Models Challenge Safety Protocols with Fictional Disguises

A study reveals AI models struggle to detect harmful requests when disguised in complex language, highlighting vulnerabilities in current safety proto...