The FBI has issued a FLASH cybersecurity advisory on ARTEX, an open-source, LLM-driven penetration testing framework that it says criminals have turned into a tool for automating intrusions "at machine speed, from reconnaissance through data exfiltration." The advisory, FLASH-20261113-001, is dated November 13 and was published through the FBI's Internet Crime Complaint Center; it drew wider attention on November 16.
According to the FBI, a range of malicious actors have used ARTEX since at least August 2026 against organizations in financial services, government, technology, defense, semiconductors, energy, healthcare, manufacturing, telecommunications and the legal sector. Some intrusions went from initial targeting to privileged access in under three hours. The bureau cites attacks on South Korean financial institutions that exposed personal information of about 67,000 people, and a supply-chain compromise of ConnectWise Cloud Backup customers that gave access to more than 2,250 downstream tenants.
ARTEX is written in Go and turns goals described in natural language into commands run by specialized agents. It ships no model of its own and plugs into external LLM providers, including Anthropic Claude, OpenAI GPT, DeepSeek, GLM, Grok and Kimi. The FBI says analysis of attacker-controlled instances shows Chinese models are the most common, with DeepSeek's Flash family dominating because of its cost and lack of observable guardrails when requests are framed as penetration testing.
The tool was published on GitHub by a developer using the name "Autumn-27" in July 2026, and the repository was taken down on October 8 after reports of abuse. The FBI says that has not slowed its spread: forks, archived copies and a commercial edition called ARTEX PRO remain available, and researchers counted more than 1,600 publicly exposed ARTEX consoles as of early November.
The FBI's recommendations focus on peripheral attack surface, which it says ARTEX is particularly good at finding: partner and supplier portals, mobile backends, legacy services and third-party integrations. It advises organizations to inventory every internet-facing application, restrict cloud administration and publishing endpoints such as Kudu/SCM to trusted networks, disable basic publishing authentication where it is not needed, apply least privilege to multi-tenant app permissions, and monitor for new credentials and privilege grants on high-privilege applications. Microsoft 365 administrators who have ever authorized affected ConnectWise applications are told to review those tenants by application ID rather than display name.