{
 "license": "CC-BY-4.0",
 "attribution": "Fide AI, Agentic Cyber Explorer",
 "url": "https://agentic-cyber-explorer.pages.dev/events/gtig-mandiant-artex-pentagi-incident-response-2026/",
 "asOf": "2026-11-17",
 "id": "gtig-mandiant-artex-pentagi-incident-response-2026",
 "date": "2026-10-30",
 "datePrecision": "day",
 "title": "Google and Mandiant report ARTEX and PentAGI in intrusions since late August 2026, with fast, noisy agent activity",
 "lane": "attack",
 "kind": "misuse-report",
 "summary": "Google Threat Intelligence Group (GTIG) and Mandiant report that since late August 2026 financially motivated and suspected PRC-nexus actors have used open-source agent frameworks, chiefly ARTEX and PentAGI, to carry out post-compromise stages of intrusions seen in Mandiant incident response. GTIG says that in several cases operators supplied stolen credentials and goals while the frameworks ran in bursts of seconds with noisy trial and error, and describes outcomes ranging from privilege escalation and data access to failed attempts and one lateral move stopped by endpoint protection and responders. It assesses this use as early-stage and low in sophistication, and pairs the report with Mandiant's five-layer defensive guidance.",
 "whyItMatters": "It documents agent framework use in real intrusions, the forensic indicators responders used to infer agent involvement, and cases where attempts failed or defenders intervened.",
 "actors": [
  "google-threat-intelligence-group",
  "mandiant",
  "storm-0940"
 ],
 "topics": [
  "ai-enabled-intrusion",
  "autonomous-pentest",
  "multi-agent-security",
  "threat-intelligence"
 ],
 "atlas": [
  "tools",
  "other-agents",
  "credentials"
 ],
 "artifacts": [
  "artex",
  "pentagi"
 ],
 "sources": [
  {
   "url": "https://cloud.google.com/blog/topics/threat-intelligence/artex-and-friends-enter-agentic-offensive-frameworks/",
   "publisher": "Google Threat Intelligence Group and Mandiant",
   "title": "ARTEX and Friends: Enter Agentic Offensive Frameworks",
   "date": "2026-10-30",
   "type": "primary",
   "accessed": "2026-11-01"
  }
 ],
 "keyFacts": [
  {
   "fact": "GTIG and Mandiant say that since late August 2026 they have seen threat actors operate open-source multi-agent pentesting platforms, including ARTEX, PentAGI and CAIRN, and possibly custom harnesses, in Mandiant incident response engagements, corroborated by GTIG research telemetry. The introduction also names Hexstrike among frameworks seen in engagements; only ARTEX and PentAGI are described in detail.",
   "locator": "Introduction; Observations in the Wild"
  },
  {
   "fact": "GTIG describes ARTEX as a multi-agent, LLM-driven pentesting and offensive command-and-control platform written in Go with a Next.js web console and six specialized sub-agents, originally released on GitHub by an individual developer. It says the developer took the original repository down in October 2026 because of widespread threat actor abuse, and that community forks, including a localized variant called ScopeWeaver and a commercial ARTEX PRO, keep it spreading.",
   "locator": "ARTEX Framework"
  },
  {
   "fact": "GTIG reports ARTEX use since late August 2026 by financially motivated data-theft extortion groups and suspected PRC-nexus espionage clusters, for post-compromise reconnaissance, lateral movement and privilege escalation, affecting US organizations in government, technology, chemicals and materials, and manufacturing. It tracks one ARTEX campaign as UNC7228, which it says overlaps with the activity Microsoft tracks as Storm-0940.",
   "locator": "ARTEX Framework"
  },
  {
   "fact": "Citing Wiz research, GTIG says UNC7228 activity dates to at least August 2026 and typically moved from a first compromised account to deep system access in about four hours, in one case about 15 minutes, against US and European entities in enterprise software, industrial automation, semiconductors and telecommunications. These are Wiz's figures as relayed by GTIG, not Mandiant case data.",
   "locator": "ARTEX Framework"
  },
  {
   "fact": "In several cases operators gave ARTEX compromised credentials and instructed it to carry out subsequent attack stages. From log timestamps, GTIG found extended pauses between operational phases, while individual ARTEX execution sequences ran in bursts of roughly 2 to 25 seconds; one operation logged 27 events in 23 seconds of active burst time while probing cloud permissions.",
   "locator": "ARTEX Framework, differences from a human operator"
  },
  {
   "fact": "GTIG says ARTEX reached code execution in a far noisier way than a competent human operator, relying on rapid trial and error: one attempt generated 240 log events, with errors possibly caused by sub-agents overwriting each other while pursuing the same objective. ARTEX also left recognizable, incrementing markers in names of files, created accounts and email subjects.",
   "locator": "ARTEX Framework, differences from a human operator"
  },
  {
   "fact": "GTIG's lifecycle table lists completed actions, including domain privilege escalation, enumeration of over 110,000 Active Directory objects, database queries and bulk storage reads, and abuse of an application's notification feature to email over 1,300 internal users. It also lists failures: cloud privilege-escalation and security-group attempts that failed, and an attempt to move onto domain controllers that ended when endpoint protection blocked execution and defenders cut off access to the server.",
   "locator": "Observed ARTEX Activity Mapped along the phases of a targeted attack lifecycle"
  },
  {
   "fact": "GTIG says most observed PentAGI activity stayed at asset mapping or vulnerability scanning, though some operations progressed to intrusions and persistence across multiple organizations. It describes a customized platform built on the open-source PentAGI, likely used from at least late August to early October 2026 against organizations in Asia including system integrators and telecommunications firms, and another case that chained PentAGI with proxy services against the aerospace sector.",
   "locator": "PentAGI Offensive Framework"
  },
  {
   "fact": "GTIG assesses adversary use at the time of the report as early-stage, with open-source tools used in low-sophistication ways producing broad, noisy but nonetheless successful campaigns that trade stealth for speed. It assesses with high confidence that cybercrime actors will keep adopting such tools and expects stealth-focused actors to be less likely to use public versions. Its seven-stage model places activity at stage 5, with fewer humans in the loop, and forecasts stages 6 and 7 within 1 to 3 and 3 to 6 months; these are forecasts, not observations.",
   "locator": "Evolution of Autonomous Offensive Frameworks; Outlook"
  },
  {
   "fact": "Mandiant's guidance has five layers: restrict control planes to trusted paths; replace long-lived secrets with short-lived, tightly scoped credentials, including for service identities; harden configurations, segment application tiers and limit error detail shown to outsiders; put rate and recipient limits on high-impact business functions; and baseline identity, control-plane, data-plane and outbound behavior to detect bursts and unusual sequences. GTIG calls these long-standing practices and says static indicators will miss much activity that uses valid credentials and legitimate features.",
   "locator": "Defending Against AI-Enabled Attacks: Five Layers of Defense"
  }
 ],
 "significance": 4,
 "fideQuestions": [
  "FID-087"
 ],
 "methods": [
  "agent-orchestrated-intrusion",
  "credential-overreach"
 ],
 "review": "assistant-drafted",
 "addedOn": "2026-11-01",
 "revisions": [
  {
   "id": "scope-forensic-indicators",
   "on": "2026-11-17",
   "kind": "correction",
   "why": "Correction: described behavioral traits as indicators used in the reported cases. The source does not establish a validated way to distinguish agent activity from human activity."
  }
 ]
}