By Hamza Faisal

The FBI has warned that ARTEX, an open-source AI penetration-testing framework, has been weaponised by criminals and state-linked groups to automate intrusions around the world — with some attacks going from first contact to full privileged access in under three hours.

In a 13-page FLASH advisory released on November 13 (FLASH-20261113-001) through its Internet Crime Complaint Center, the bureau said a “variety” of malicious actors have used ARTEX since at least August 2026. The targets span financial services, government, technology, defence, semiconductors, energy, healthcare, manufacturing, telecommunications and legal firms. The advisory, marked TLP:CLEAR for unrestricted sharing, draws on eleven industry and government sources and maps the observed activity to the MITRE ATT&CK framework.

What ARTEX does

ARTEX is an autonomous, multi-agent penetration-testing platform written in Go, driven by large language models. It takes a goal written in plain English and turns it into commands carried out by a team of AI agents — scanning a target, finding vulnerabilities, exploiting them, escalating privileges and stealing data — with very little human involvement after that.

The tool was released on GitHub in July 2026 by a developer using the name “Autumn-27”, and gained an estimated 1,500 to 2,000 stars within ten weeks. The developer pulled the repository on October 8, after reports emerged that attackers were abusing it — but the FBI says forks, localised variants, archived copies and a paid version called ARTEX PRO keep it circulating. Researchers counted more than 1,600 publicly exposed ARTEX consoles on the internet in early November, with the true number of deployments believed to be far higher.

The FBI says Chinese AI models are the most common engines behind attacker-run ARTEX instances, with DeepSeek’s Flash family dominating because it is cheap and — framed as a penetration test — shows no observable guardrails. The result, the bureau says, is that people with minimal technical expertise can now run complex, multi-target operations simultaneously.

The attacks it has powered

The advisory ties ARTEX to several major incidents. Attacks on South Korean financial institutions stole personal information on about 67,000 people. The October supply-chain attack on ConnectWise Cloud Backup, which exposed more than 2,250 downstream customer accounts and opened access to Microsoft 365 customer environments, is attributed by Microsoft to a group it calls Storm-0940, which it describes as China-state aligned.

The FBI’s advisory also names a second criminal crew. Microsoft reported on November 6 that a financially motivated cluster it calls Storm-3358, based in Cambodia, began using ARTEX against financial organisations in October, mostly in Southeast Asia. And in Japan, an investigation led by the country’s Computer Emergency Response Team Coordination Center found about 1,000 servers suspected of serving as bases for AI-driven cyberattacks — most of them hosting AI platforms such as ARTEX, the Straits Times reported on November 14.

What defenders are being told

The FLASH includes indicators of compromise — including 22 IP addresses — a prioritised list of mitigations and a mapping of attacker techniques, so defenders can hunt for ARTEX activity on their own networks. The FBI cautions that some listed addresses are historical and may since have been reassigned to legitimate services, so teams should check them against their own telemetry before blocking.

Separately, security researcher MalwareMustDie released an open-source ARTEX detection signature pack on Friday to help defenders spot attacks involving the framework. The FBI’s bottom line is blunt: agentic offensive frameworks are no longer a research curiosity. They are, in the bureau’s words, a mainstream threat — and the window defenders have to respond is now measured in hours, not days.