Skip to content
NewA calmer, faster ZeroHour

Google says exploit_agent found and ran exploits without humans

Original title: AI Framework Autonomously Weaponized Flaws in Real Attacks, Google Says

SecurityWeekby Ionut Arghire ()4 min read1
Part of a story covered by 15 sourcesGoogle says unattended AI agent built and ran exploitsMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

GTIG found the framework on an exposed server holding stolen data, webshell scripts, and interest in French sites.

Google Threat Intelligence Group says it found exploit_agent on an exposed staging server that compared patched and unpatched web code, wrote exploits, and ran them without a human. The server also held stolen victim data and scripts for reconnaissance and webshells, with signs of interest in French media, entertainment, and education sites. GTIG reports wider use of Artex, Cairn, PentAGI, and similar frameworks, plus LLM use by groups including APT42 and UNC1069.

Unconfirmed: French-sector targeting is described as indicated interest, and actor assessments are GTIG's.

Key facts
  • Artex is the most used AI attack framework GTIG saw, mainly among Chinese-speaking groups.
  • Cairn was seen against US aerospace; PentAGI was used for scanning in Asia.
  • Jieyou, active since March 2026, offers a chatbot for configuring phishing panels.
  • GTIG says one LLMJacking case reached $1 million in cumulative resource abuse.
  • In August 2026 Google found an IaaS platform that automated an influence-campaign lifecycle.
Security assessment · AI, from the article

Impact: automated exploit development, stolen data, and webshell deployment

Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

From the original article

4 min read at securityweek.com

Threat actors are deploying multi-agent offensive frameworks to increase the velocity of multi-stage intrusions. The post AI Framework Autonomously Weaponized Flaws in Real Attacks, Google Says appeared first on SecurityWeek.

SecurityWeek

By Ionut Arghire · securityweek.com

Read the full article at SecurityWeek

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.