Google's GTIG AI Threat Tracker, published November 16, 2026, reports the first observed LLM framework, exploit_agent, fully automating vulnerability discovery and exploitation. Zero-day exploits spiked to 22 in August 2026, and LLMJacking charges exceeded $100,000 in a single day.
What changed in the GTIG AI Threat Tracker
On November 16, 2026, Google Threat Intelligence Group (GTIG) and Mandiant published the latest GTIG AI Threat Tracker, which for the first time documents a multi-agent framework, exploit_agent, that fully automates vulnerability discovery and exploitation.
The framework is wired to both frontier and local LLMs, and it continuously compares popular web application code before and after patch announcements, identifies likely vulnerabilities, drafts exploits and executes them on its own. Its operator targeted French media, entertainment and education websites.
The report also documents the live intrusion use of open-source agentic pentesting frameworks ARTEX, CAIRN and PentAGI, with experimentation in CyberStrikeAI and HexStrike AI.
The tracker shows disclosures and exploits nearly doubled from January to August 2026, and zero-days rose from a 2025 average of 8 per month to 11 per month in 2026 before spiking to 22 in August. In late August, PaperCut NG/MF flaws CVE-2026-81578 and CVE-2026-82078 were exploited as zero-days, with LLM-assisted exploits within a day of disclosure. The campaign hit hundreds of organizations, reaching domain admin within minutes.
In a separate finding, LLMJacking — renting out compromised cloud AI resources — saw single-day charges above $100,000, and up to $1 million cumulatively in the worst cases. Google has released SecOps rulepacks to detect ARTEX, alongside a Mandiant hardening guide.
How to fix it
- Patch internet-facing applications as soon as fixes ship.
- Apply the Mandiant hardening guide and enable the SecOps rulepacks.
- Monitor cloud and LLM bills for LLMJacking, and rotate exposed keys.
- Audit AI coding assistant configuration files (for example, .claude/settings.json).
GTIG AI Threat Tracker by the numbers
- Zero-day exploits, August 2026: 22 (Nov 16, 2026)
- LLMJacking charges, single day: >$100,000 (Nov 16, 2026)
- LLMJacking charges, cumulative worst cases: $1,000,000 (Nov 16, 2026)
Why the GTIG AI Threat Tracker matters for business
The patch window is now days or hours, not weeks: agentic frameworks turn a patch announcement into a working exploit in under a day, so patching internet-facing apps immediately, using the hardening guide and SecOps detections, and auditing bills, keys and agent configuration files is the minimum viable response.
Top comments (0)