We’ve been following ARTEX for weeks. Now the FBI has put out a 13-page FLASH advisory, BleepingComputer reports, saying multiple threat actors have used the open-source AI pentest framework since at least August 2026 against finance, government, tech, defense, healthcare and more, including the ConnectWise supply-chain breach that hit Microsoft 365 customers. The FBI says some intrusions went from first targeting to privileged access in under three hours.
ARTEX is written in Go and doesn’t bring its own model. It plugs into outside LLMs, including Anthropic’s Claude, OpenAI’s GPT, DeepSeek, GLM, Grok and Kimi, and turns an attacker’s goals into commands for specialized agents. The original repo came down on Oct. 8, but the FBI says forks, archives, localized versions and a commercial “ARTEX PRO” keep it alive, and researchers counted more than 1,600 exposed consoles by early November. The advisory includes IOCs, with a warning that IPs may have changed. Separately, MalwareMustDie released an open-source ARTEX detection pack and playbook.
What to do
- Inventory internet-facing apps, partner portals and old services, the stuff ARTEX is good at finding.
- Lock cloud admin interfaces to known networks and review app permissions and auth logs.
- Rate-limit sensitive APIs and alert on bursts of automated requests and sudden privilege changes.
My take
This is the speed problem in one sentence: under three hours. Your patch window and your alerting have to assume the other side has agents now. Report anything you find to IC3.
God Bless, and May the Source Be with You!