Google reports first AI agent finding and exploiting flaws unaided
Google's Threat Intelligence Group says it saw an attacker use an AI agent to find vulnerabilities and build exploits with no human in the loop.
In its latest AI Threat Tracker, Google calls this a significant threshold. It also cites an agentic framework reversing security patches without human intervention. Mandiant investigated a financially motivated actor that breached a cloud environment and deployed a multi-agent framework, compromising thousands of credentials within six hours. Many such frameworks leave clear forensic traces, suggesting they favor speed over stealth.
Source: IT Brief UK, published on November 17, 2026.