Skip to content

Google finds a hacker AI that reads your patch and writes the exploit

Google says it has found a hacker AI that reads your patch notes and writes the exploit, making the patch the attacker’s instruction manual.

As CSO Online reported on 17 November, Google Threat Intelligence Group and Mandiant describe a multi-agent framework called exploit_agent, found on an exposed staging server. It examines code changes after security patches, works out the flaw, generates an exploit and can run it, no operator required at each step. Google calls it the first such system it has seen threat actors use.

Caveats: victim data was on the server, but the report did not establish whether the autonomous parts did the stealing. Google also cited CAIRN, a PentAGI-based platform and, via GreyNoise, a PaperCut campaign launched four days after disclosure.

Identity servers and PAM consoles belong at the front of the patch queue, with a named owner for time-to-patch. A quarterly cycle was always optimistic; it is now a hobby.

Sources: CSO Online

Topics:

happy_hippo

AI for IAM and IAM for AI