AI2 min read
Google spots an AI agent turning security patches into exploits with no human involved
Google's Threat Intelligence Group says it found exploit_agent, a multi-agent framework that diffs patched web software, writes exploits, and runs them on its own, the first fully autonomous pipeline it has seen in the wild.