Artex AI pen-test framework weaponized: open-source Go agentic pentest tool abused for credential attacks, cloud breaches at machine speed
Severity: high · Patch status: no-fix-needed · Since 2026-07-01
Artex, an open-source multiagent AI penetration-testing framework written in Go published to GitHub in July 2026, has been widely abused by threat actors to automate credential attacks, database dumping, API abuse, and cloud breaches at machine speed. China-linked Storm-0940 used it in October against 300+ accounts at eight critical organizations (151,000 auth requests, ~2,900/min) and pivoted to steal OAuth credentials from ConnectWise, compromising 2,250+ downstream Microsoft 365/Azure environments in three hours. Cambodia-based Storm-3358 used it against financial organizations in Southeast Asia; CrowdStrike linked it to espionage actor Vapor Panda; Censys found 1,400+ exposed consoles. The creator deleted the repo in late October, but forks and 45 similar agentic frameworks remain active threats.
What to do
- Hunt for authentication bursts, password-spraying patterns (~thousands of attempts/min without lockout) and anomalous OAuth app consent/credential use
- Audit and rotate OAuth application credentials, especially third-party backup apps; enforce MFA and conditional access on cloud identities
- Scan for exposed Artex consoles and pentest-tool artifacts, web shells, and open-source RATs on internet-facing systems
- Remove secrets from public repositories and rotate any exposed GitHub/Azure Key Vault credentials
- Restrict and monitor agentic/AI tooling and service principals with privileged cloud permissions
Sources
All live incidents on Threat Watch
