The FBI has formally warned that cybercriminals are using ARTEX, an AI-driven penetration-testing framework, to automate break-ins against organizations worldwide. In a FLASH advisory issued on November 13, the bureau says that in some observed attacks the criminals went from choosing a target to holding privileged access in under three hours, leaving defenders very little time to react. The advisory was highlighted by researchers at Malwarebytes in their write-up.
ARTEX has been the subject of a run of vendor reports over the past month, from Google's observation of AI agents running real intrusions to Microsoft's account of a China-aligned group using it after compromising ConnectWise credentials. The FBI FLASH adds an official government warning that the tool is also in criminal hands.
A testing tool that does most of the work
ARTEX is open source and was built for authorized security testing, where a company pays someone to break in and find weaknesses before criminals do. An operator gives it a target and an objective. A planning agent splits the job into tasks, and other AI agents run security tools to carry them out and share what they find. ARTEX has no AI model of its own; it connects to external large language models. Once a task starts, it needs relatively little human input.
According to the FBI, attackers have used it since at least August against organizations in finance, government, healthcare and technology. The danger is less a new technique than speed: the tool chains individual weaknesses, such as a forgotten employee portal or an exposed cloud credential, into a full intrusion, and does it quickly.
Fix the forgotten portals and leaked keys first
The FBI's recommendations are aimed at organizations and target exactly the weaknesses an automated agent finds fastest:
- Patch internet-facing systems promptly and inventory overlooked applications, including old employee and supplier portals.
- Restrict administrative interfaces to approved access paths.
- Scan for exposed passwords and API keys, rotate any compromised secrets, and give applications only the access they need.
- Watch for rapid bursts of requests, unusual data access, and unexpected credential or privilege changes.
- Prepare procedures in advance to revoke access and isolate affected systems quickly.
The last point matters most. An alert at minute 30 of a three-hour intrusion is only useful if the team can act on it before minute 180, and that depends on decisions made long before the attack.
This briefing is provided by IntelFusions for informational and defensive purposes only. It is based on sources assessed to be reliable at the time of writing, and analytic judgments carry the confidence levels indicated. Indicators of compromise are defanged; re-arm them only in controlled environments. IntelFusions is not affiliated with the organizations named and makes no warranty as to completeness or accuracy.