From patch to exploit without human involvement: GTIG's exploit_agent and the grace period
Patch-to-Exploit Without Humans: GTIG’s exploit_agent and the Grace Period
Google Threat Intelligence Group (GTIG) reported on November 16, 2026 that it had for the first time found a threat actor in the wild using a multi-agent LLM framework, exploit_agent, to fully automate the process from patch diff to exploit. The framework was recovered from an exposed staging server and was configured to continuously compare pre- and post-patch versions of popular web application software, identify the vulnerabilities fixed by patches, draft exploit code and execute it automatically. GTIG did not disclose the success rate or the scale of victims; a configuration comment labels it "pilot," so the finding reflects capability and intent rather than scale. GTIG also said such autonomous attack tools generate noise that is easy to detect, accompanied by large numbers of AI-generated PoCs that cannot actually run.
GTIG found a fully automated patch-diff and exploit pipeline in the wild; together with evidence that the patch window has shrunk from weeks to hours, this lets security teams reassess remediation timelines.