Attacker's AI agent reportedly built exploits from patches unaided
Google Threat Intelligence Group (GTIG) reports that a suspected Russian-speaking operator used an AI agent to turn security patches into working exploits with no human involvement. GTIG says it is the first such case it has documented, and that the operator later stole data from victim organizations. The finding sits in Google's fourth-quarter 2026 AI threat tracker, which describes an advance over Q3 in how attackers use large language models to find and exploit vulnerabilities on their own.
Researchers found an exposed staging server running a multi-agent framework called exploit_agent. It works with several frontier and local models and uses Russian-language tooling. It was built for continuous differential analysis of popular web application software: compare code before and after a security update, find the flaw the update fixed, build an exploit and run it automatically. The server also held stolen victim data and post-compromise scripts for internal reconnaissance, brute-forcing, installing webshells and removing rival attackers' webshells. Victim telemetry points to French media, entertainment and education sites.
The report also documents other criminal and espionage groups using autonomous multi-agent offensive frameworks, including ARTEX, CAIRN, CyberStrikeAI, HexStrike AI and PentAGI. GTIG saw PentAGI used with proxy services against the aerospace sector.
This is a setback for security. A patch has long been a signal that defenders should act, and now it can also be a blueprint. When exploit building is automated, the time between a fix being published and attacks beginning shrinks, so organizations that update slowly carry more risk. The same techniques can also help defenders find flaws, but this report documents only the offensive use.
Caveats: the operator's Russian-speaking identity and the PRC link some experimenting actors are suspected to have are both unconfirmed, and the claim of a first occurrence is GTIG's own assessment. GTIG believes such frameworks will very likely keep spreading because entry barriers are low. Watch for faster patch deployment and for defenders adopting similar automation.
Source: The Register
Many Minded