Google says attackers have fully automated turning software patches into exploits, with no human in the loop

Google Threat Intelligence Group · Company report

Read the source (opens in a new tab)
What happened
Google Threat Intelligence Group and Mandiant published an AI Threat Tracker update on November 16, 2026 on criminal, espionage and influence-operation use of AI.
Key points
  • GTIG says it identified threat actors using LLM frameworks "to fully automate vulnerability discovery and exploitation without a human in the loop for the first time."
  • An exposed server held a framework, dubbed exploit_agent, that compared code before and after patches, drafted exploits and ran them; victim data suggests a focus on French media and entertainment websites and on the education sector.
  • After two PHP flaws were patched on October 21, hundreds of "claimed but non-functional" exploits, likely made with LLMs, were published, adding verification work for defenders.
  • A suspected North Korean cluster altered Claude Code settings files in victims' internal code repositories as a supply chain route, after luring cryptocurrency staff through AI-made recruiter personas.
Evidence
Threat intelligence report drawing on Mandiant incident response, Google's threat actor tracking and platform defenses. It names no victim organizations and gives no count of intrusions.
Uncertain
GTIG does not attribute the exploit_agent framework to a named actor, and the North Korean link is a suspected attribution.

See it in the latest record