Google Reports Attacks That Find and Exploit Software Flaws With No Human Involved

Google’s Threat Intelligence Group (GTIG) says it has seen attackers use AI to find and exploit software flaws without any human stepping in. In its quarterly AI Threat Tracker, published Nov. 16, 2026, the group says attackers used frameworks built on large language models to fully automate the discovery and exploitation of flaws. GTIG says this is the first time it has seen that happen.
What’s new is that no one had to guide the attack step by step. The report describes a framework that watches for patch announcements, compares a program’s code before and after the fix to work out what was repaired, then writes an attack and runs it without human intervention. GTIG says this shortens the time organizations have to install patches.
GTIG says it found that framework, which it calls exploit_agent, on a staging server left open to the internet, along with data taken from victims. The group says records on the server point to an operator interested in French websites in media and entertainment, and in education. The account rests on Google's own monitoring data and that one server.
The report names four open-source frameworks that run several AI agents together and that GTIG says are already in use: ARTEX, CAIRN, CyberStrikeAI and PentAGI. GTIG says it has independently seen two of them in live intrusions.
For outside corroboration, the report points to two flaws in PaperCut print management servers disclosed in late August 2026. Security researchers and the company Huntress generated working exploits within a day of disclosure, and Greynoise reported a campaign four days after disclosure that ran against hundreds of organizations.
Not all of it works. The report says the same lower barrier has produced a wave of attacks that claim to work but do not, leaving defenders to check each one. GTIG says the frameworks trade stealth for speed, generating traffic and code traces that make them easier to catch.
