Skip to content
NewA calmer, faster ZeroHour

Talos finds 700 internet-reachable PentAGI instances as AI attack tools spread

Original title: Researchers Find 700 Exposed PentAGI Instances as AI-Powered Hacking Frameworks Spread

GBHackersby Mayura Kathir ()2 min read1
Part of a story covered by 8 sourcesBarracuda links 44-hour cloud breach to leftover key and ARTEXMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

The count covers reachable deployments, not confirmed abuse; named operators have used the separate ARTEX framework against hundreds of organizations.

Cisco Talos found more than 700 internet-reachable PentAGI instances through a Censys search on the page title, and said visibility alone does not show unauthenticated access or unauthorized testing. The report describes agent frameworks, including Hephaestus and ARTEX, orchestrating familiar security tools. CrowdStrike linked VAPOR PANDA to ARTEX activity against more than 200 organizations in at least 14 countries from October 17, 2026, with post-exploitation confirmed at nine. Microsoft attributed separate ARTEX-driven activity to Storm-0940 after ConnectWise application credentials were compromised.

Unconfirmed: Reachable PentAGI instances are not confirmed malicious, and overlapping tools do not prove one actor controlled all activity.

Key facts
  • Storm-0940 attempted authentication against more than 2,250 downstream tenants.
  • Microsoft said successful authentication did not establish data access or exfiltration.
  • One agent moved from GitHub secrets to Azure Key Vault exfiltration within five minutes.
  • Talos identified more than 2,500 ARTEX-associated command-and-control IP addresses.
  • Overlapping tools included VShell, AdaptixC2, Mimikatz, and Cobalt Strike.
Security assessment · AI, from the article

Impact: Cloud reconnaissance, credential changes, and possible data exfiltration

What to do
  1. 1.Correlate authentication, endpoint, and cloud logs for retries, token use, and privilege changes.
  2. 2.Revoke exposed secrets and limit workload identity privileges.
  3. 3.Preserve evidence while disabling confirmed malicious access and checking for persistence.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

From the original article

2 min read at gbhackers.com

More than 700 internet-reachable PentAGI instances as AI-enabled penetration testing frameworks increasingly appear in real-world offensive operations. The finding highlights the growing availability of platforms that automate reconnaissance, vulnerability testing, and follow-on activity, reducing manual effort and allowing operators to pursue multiple intrusion paths simultaneously.

GBHackers

By Mayura Kathir · 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.