technology

FBI Warns: Hackers Launch Swift Attacks Using AI Framework ARTEX

Published November 17, 2026 · Updated November 18, 2026 · 3 min read

The FBI has issued an urgent security announcement warning that hackers are utilizing a tool named ARTEX to breach networks across various sectors. The framework enables cybercriminals to efficiently conduct a full range of activities from reconnaissance and initial access, to privilege escalation and data exfiltration in a relatively short period.

ARTEX is described as an open-source, autonomous penetration testing framework built on large language models, with many of its components linked to AI models developed in China. The FBI notes that malicious actors have been exploiting this framework for attacks on organizations in multiple industries since at least August, including financial services, government, technology, defense, semiconductor, energy, healthcare, manufacturing, telecommunications, and legal sectors.

This 13-page urgent alert underscores that certain intrusion incidents have demonstrated that ARTEX is capable of evolving from initial target identification to acquiring privileged access in under three hours.

Particularly concerning are recent attacks targeting South Korean financial institutions, which resulted in the theft of personal information affecting approximately 67,000 individuals. Additionally, the FBI pointed to a supply chain intrusion incident that impacted customers of ConnectWise Cloud Backup, where attackers gained access to systems belonging to over 1,600 downstream tenants.

The ARTEX framework employs specialized AI agents to convert high-level tactical objectives described in natural language into machine-executable commands. This process requires minimal human intervention beyond setting initial instructions.

Developed in the Go programming language, ARTEX operates as an autonomous multi-agent penetration testing platform. While the original developers removed its GitHub repository in October due to reports of malicious use, ARTEX remains accessible through community forks, archived copies, localized versions, and commercial variants.

Research conducted in early November uncovered over 1,600 publicly exposed instances of the ARTEX control console. When accounting for private systems and those protected by firewalls, the actual number of deployments could be even higher.

The FBI has expressed concerns that this framework lowers the technical barriers for complex attacks, accelerates multi-target assaults, and exploits overlooked systems, including vendor portals, legacy services, third-party integrations, and cloud infrastructure. As a preventative measure, the FBI urges organizations to reduce their attack surface exposed to the internet, enhance identity and access controls, monitor application activity, and preserve evidence of any suspicious intrusions.

It’s important to note that the ARTEX framework itself does not inherently include a superintelligent (SI) model. Instead, it connects to external large language model (LLM) providers, such as cutting-edge models from Anthropic’s Claude and OpenAI’s GPT, alongside open-source or Chinese-developed models like DeepSeek, GLM, Grok, and Kimi.

Analysis of ARTEX instances controlled by attackers reveals a significant prevalence of models developed in China, particularly the DeepSeek Flash series, which is favored for its high performance at a low cost, combined with weak safeguards against detection—often yielding negligible defenses when requests are disguised as penetration tests.

In light of these developments, the FBI advises organizations to investigate or review indicators of compromise (IOCs) before taking action to thwart potential attacks. Ignoring these recommendations could leave systems vulnerable to the rapidly evolving threat landscape posed by ARTEX and similar frameworks.