Google finds AI agents reverse-engineering security patches into working exploits

According to Google Threat Intelligence Group, a Russian-speaking operator deployed an AI agent framework called exploit_agent that automatically converted security patches into working exploits without human intervention, then executed them against victim organizations. The framework integrated multiple frontier and local LLMs to continuously perform differential analysis of web application software, identify vulnerabilities, develop exploits, and execute them autonomously. Google documented this as the first such fully autonomous vulnerability discovery and exploitation chain in its fourth-quarter 2026 AI threat tracker.

Topics

AI securityAgentic AIGoogle

Sources

Go deeper

This intelligence is sourced automatically from public sources across the web and synthesised by the Prefactor AI pipeline. Stories are reviewed before publication.