Google reports AI framework turning security patches into exploits unaided
Google says attackers used an AI framework with no human in the loop to turn security patches into exploits for the first time.
Named exploit_agent, the tool was found on an exposed staging server, according to the AI Threat Tracker published November 16, 2026. It compares popular web application code before and after each patch and automatically runs the exploit it drafts. The previous edition, about two months earlier, had not seen fully autonomous attack pipelines against real targets. The report was written by Google's Threat Intelligence Group.
Source: SiliconANGLE, published on November 16, 2026.