AI Stack Alert High TD-2026-0104 For CISO, SOC / IR, AppSec, Red Team Unconfirmed AI-assisted
Google: Russian-speaking actors ran AI-driven 'exploit_agent' for automated patch-diff exploitation
Google's Threat Intelligence Group and Mandiant report an intrusion operation they believe was run end-to-end by AI with no human in the loop. The system, called "exploit_agent", chains AI agents and language models to compare code before and after vendor patches, infer the fix, and draft and run an exploit. Victim data pointed to French media, entertainment and education websites.
Why it matters
Automated patch diffing and exploitation shortens the time between a vendor patch and working attacks, so unpatched systems face exposure sooner. Organizations that delay patching of widely used software are most at risk.
Do this
- Shorten patch timelines for internet-facing and widely used software, prioritizing recently patched vulnerabilities.
- Hunt for unauthorized webshells on web servers and review file integrity monitoring on web roots.
- Monitor for automated password guessing and internal reconnaissance following any web-server compromise.
- Treat public exploits appearing right after a patch with caution, as many are nonfunctional and likely AI-generated, but verify exposure anyway.
- Review the Google Threat Intelligence Group report and map its findings to detection coverage.
Signals
- NoCISA KEV
- 34/100Priority
- 70%Confidence
Evidence
“It chains together multiple AI agents with frontier and locally hosted language models to continuously compare code in widely used software before and after vendors publish patches.”
“Victim data on the staging server pointed to targeting of French websites in the media, entertainment and education sectors.”
“Google counted 22 exploited zero-days, flaws unknown to vendors, in August alone, up from 11 per month earlier in 2026 and an average of eight in 2025.”
Sources
Affected / related
exploit_agent, Google Threat Intelligence Group, Mandiant, CAIRN, PentAGI, AI-driven patch diffing and automated exploitation
Published 2026-11-16 22:51 UTC · updated 2026-11-16 23:05 UTC · AI-assisted summary, checked against quoted sources · Unconfirmed: single non-official source · How we score