LIVE

Google Reports First Instance of AI Agent Autonomously Turning Security Patches Into Exploits

🏢 Big Tech 2 min read 👁 7

Google's Threat Intelligence Group revealed that a suspected Russian-speaking operator used an autonomous AI framework to convert software patches into active exploits without human intervention.

Google Reports First Instance of AI Agent Autonomously Turning Security Patches Into Exploits🏢Big Tech

Photo: theregister.com

Key points

  • Google's GTIG documented the first known case of an AI agent turning patches into active exploits autonomously.
  • The Russian-language framework 'exploit_agent' handled everything from code comparison to execution without human help.
  • Victim telemetry pointed to targeted campaigns against French media, entertainment, and education websites.

A suspected Russian-speaking threat actor successfully used an autonomous AI agent to transform security updates into working cyberattacks without any human involvement, according to the Google Threat Intelligence Group (GTIG). Documented in Google's fourth-quarter AI threat tracker, the incident marks a watershed moment in how malicious actors deploy large language models for end-to-end vulnerability discovery and exploitation.

Investigators discovered an exposed staging server housing a multi-agent framework dubbed exploit_agent. Utilizing Russian-language tooling, the system continuously compared application code before and after security patches, identified flaws, built functional exploits, and executed them automatically. The server also held stolen data from victims and various post-compromise scripts designed for internal network reconnaissance, brute-forcing, and deploying webshells.

Telemetry data indicated the campaign specifically targeted French organizations in the media, entertainment, and education sectors. Google researchers noted that this attack represents a broader shift away from manual prompt-based LLM abuse toward fully autonomous offensive security frameworks, such as ARTEX, CAIRN, and PentAGI, which have also seen experimentation by groups linked to the People's Republic of China.

What to watch next: Monitor enterprise security updates and the adoption of defensive AI measures to counter automated, multi-stage agentic cyberattacks.

Source theregister.com — theregister.com/research/2026/11/16/google-ai-agent-turned-a-securi… ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.