AI researchers let models off the leash – then watched as they tried to add malware to a FOSS project

Posted by PrivacyWire

Models used social engineering and collaborated among themselves to solve a security challenge