Skip to content
NewA calmer, faster ZeroHour
Threat actorhighExploited in the wild

Google says attackers now run autonomous AI exploit agents

Original title: Google Warns Hackers Are Using Autonomous AI Agents to Exploit Vulnerabilities Without Human Intervention

GBHackersby Divya ()3 min read1
Part of a story covered by 13 sourcesGTIG says exploit_agent wrote and ran exploits with no human in the loopMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

GTIG found an exposed multi-agent framework that drafts and runs exploits, plus resale of stolen model access.

Google Threat Intelligence Group’s November 17 AI Threat Tracker says threat actors have moved from prompting to autonomous multi-agent exploit pipelines. An exposed staging server hosted exploit_agent, which compares code around patches, drafts exploits, and runs them, with scripts for reconnaissance, brute force, and webshells. GTIG also describes in-the-wild use of CAIRN, PentAGI, and ARTEX, rapid weaponization of disclosed flaws, and LLMJacking that resells stolen cloud and model access.

Unconfirmed: GreyNoise said a post-disclosure campaign likely used LLMs; that link is not confirmed.

Key facts
  • Victim data on the server pointed to French media, entertainment, and education sites.
  • CAIRN was used against US aerospace targets; a PentAGI platform scanned organizations across Asia.
  • Mandiant-documented ARTEX bursts lasted about 2 to 25 seconds and deployed webshells.
  • LLMJacking charges exceeded $100,000 in a day and reached $1 million cumulatively, Mandiant said.
  • Stolen access is resold via LLM Mixers at $0.10 to $25 per million tokens.
Security assessment · AI, from the articleExploited in the wild

Impact: Automated exploitation, webshells, credential theft, and resold AI access

ProductAffected
PaperCut—
LiteLLM—
PHP—
What to do
  1. 1.Replace long-lived secrets with short-lived credentials and enforce least privilege.
  2. 2.Isolate control planes and segment networks.
  3. 3.Detect behavioral anomalies rather than relying only on static indicators.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

Vulnerabilities mentioned

All
CVEVulnerabilityCVSSEPSSSignalsPublished
CVE-2026-109618
+1 related CVE
OOM longjmp memory flaw enabling RCE in PHP
9.2100%KEVPoC ×25
KEV
CVE-2026-40217
LiteLLM through 2026-04-08 allows remote attackers to execute arbitrary code via bytecode rewriting at the /guardrails/test_custom_code URI
8.817%PoC
CVE-2026-82078
+1 related CVE
Unsafe Reflection RCE in PaperCut NG/MF, Chained with Auth Bypass in Attacks
9.4
group max
63%KEV
KEV

From the original article

3 min read at gbhackers.com

Google Threat Intelligence Group (GTIG) has warned that adversarial AI use has crossed a significant threshold. For the first time, threat actors are running vulnerability discovery and exploitation without a human in the loop. The November 17 GTIG AI Threat Tracker says attackers have moved past basic prompting to autonomous multi-agent frameworks, automated exploit pipelines, […]

GBHackers

By Divya · gbhackers.com

Read the full article at GBHackers

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.