SOCRadar Finds 916 Exposed ARTEX AI Pentest Consoles on U.S. Infrastructure
Attack Surface Exposure · 5 sources · Updated 1d ago · First seen Nov 12, 2026

SOCRadar identified 916 internet-exposed ARTEX-family penetration-testing consoles on U.S. infrastructure in a November 11 snapshot, including 896 served over unencrypted HTTP. Shared builds, certificates and deployment patterns suggest large coordinated fleets, but researchers did not attribute them to a named threat actor. ARTEX is a legitimate open-source, LLM-driven security-testing framework; separate research has linked deployments to a South Korean financial-sector intrusion and Storm-0940’s ConnectWise Cloud Backup supply-chain campaign. Those links do not establish that the newly mapped U.S. consoles are malicious.
Trojanized Docker images named tomicdd/artex and congdonglinux/artex contain a pre-authentication backdoor enabling attacker-supplied commands to execute as root. Earlier reporting identified 62 affected consoles worldwide, including 26 in the United States; that count is distinct from the broader exposure inventory. Defenders should inventory externally accessible ARTEX installations, distinguish authorized deployments from suspicious activity, investigate associated network traffic using durable console fingerprints, and treat any installation using either backdoored image as compromised.


