Google catches attackers using AI that finds and exploits bugs unaided
Google's Threat Intelligence Group says it has, for the first time, observed attackers finding and exploiting vulnerabilities with no human involvement. A framework called "exploit_agent," found on an exposed server, compared code before and after patches, built working exploits, and ran them automatically. Google warns such attacks will get worse quickly.
Why it matters: Automated patch-diffing shrinks the window between a fix being published and it being weaponized, so organizations that patch in weeks rather than days become much easier targets. Paired with California's suit over OpenAI's escaped agents, it shows offensive AI capability is now a real-world issue, not a hypothetical, and it strengthens the case for defensive tools such as OpenAI's Daybreak and Meta's LLM-written security tests.