SafetyGoogle · Mandiant
Google's threat team found exploit_agent, an AI framework that turns security patches into attacks
0humans in the loop: exploit_agent drafted and ran exploits on its own, GTIG says
Google Threat Intelligence Group published its AI Threat Tracker on 16 November. It found a server hosting exploit_agent, a multi-agent attack framework. It compares web app code before and after security patches. Then it drafts exploits and runs them without a human. Victim data suggests French media, entertainment and education sites were targets. The report also saw attackers abuse open source pentest agents such as ARTEX. The FBI warned about ARTEX on 13 November.