FBI warns open-source Artex toolkit automates cyber attacks at machine speed

The FBI issued a flash alert on Friday warning that the open-source AI platform Artex automates cyber attacks at machine speed, with some intrusions reaching privileged access in under three hours from initial targeting.

Artex cuts the time needed to chain attack techniques, the FBI said. The framework lets threat actors with minimal technical expertise run complex, multi-target operations at the same time. The campaigns hit organizations across sectors. Examples include financial services, government and technology.

Artex automates multiple cyber-intrusion phases at machine speed. The process runs from reconnaissance through data exfiltration, the theft of data from a network. The platform turns high-level tactical goals written in words into actionable machine commands carried out by AI agents, software that acts on goals with limited human input, needing little human interaction after the initial objectives. Artex-deployed agents chart trust links among organizations and breach vulnerable third-party suppliers.

The FBI Cyber Branch raised the alarm through a post on its official X account. The FBI said malicious cyber actors had been using the open-source, large language model-driven platform since at least August 2026.

Artex came out in August on GitHub under an account called Autumn-27. The developer was identified as Chinese cybersecurity engineer Li Puhua. Li Puhua posted a statement on GitHub on October 8 saying malicious actors had abused the tool. The developer denied responsibility for crimes committed using the tool and eventually removed the GitHub repository.

Even after removal, the framework keeps spreading via community forks, copies of the repository modified by other users, local variations, archived copies and a commercial version.

The advisory points to recent Censys research showing more than 1,600 publicly exposed Artex console instances. The FBI said total deployments are likely far higher once systems behind firewalls or on private networks are counted.

Artex can connect to any AI model. It works with frontier models, the most advanced AI systems, Anthropic Claude and OpenAI GPT. It also works with open-source and Chinese-origin models. Examples include DeepSeek, GLM, Grok and Kimi.

Research into attacker-controlled Artex instances indicates attackers use Chinese models most often. DeepSeek Flash family models dominate usage. The FBI said DeepSeek Flash family models dominate usage because they balance capability with low cost and lack observable guardrails, the safety controls that block harmful output, when framed for penetration testing, authorized attempts to break into systems.

Several attacks tied to Artex hit financial institutions in South Korea and Japan, setting off scandals in both countries. Microsoft said actors backed by the Chinese state used Artex to compromise IT management software company ConnectWise in a supply chain attack, an intrusion through a trusted vendor or partner.

Amazon published a study on Friday that found different Chinese-nexus threat groups using Artex to steal credentials at scale for AI model access. Amazon said stolen keys paid for the LLM (large language model) inference behind more credential hunting, so the operation was self-sustaining.

Amazon CISO CJ Moses said another campaign targeted cryptocurrency exchanges to bypass know-your-customer verification, identity checks required of financial customers, using real-time deepfake injection, the use of synthetic audio or video during a live interaction. Moses said their investigation showed frameworks like Artex have moved from experimental use to sustained offensive operations running with minimal human intervention across multiple objectives. Moses warned that Artex automates lengthy exploitation phases. It allows smaller threat groups to carry out simultaneous intrusion campaigns beyond their previous scale.

Artex finds and exploits poorly secured systems including partner portals, supplier portals and mobile backends. It also reaches employee apps and sales-support platforms, plus legacy services and third-party integrations, along with business-unit applications. The tool creates many attempts that set off defense sensors, yet its speed outpaces defenders trying to keep up.

The tool has conducted reconnaissance and data-exfiltration attacks against financial services, government and technology. It has also targeted defense and semiconductor companies. Attacks hit energy and healthcare organizations. Manufacturing, telecommunications and legal organizations are also on the list.

The FBI published a FLASH covering tactics, techniques, procedures and indicators tied to the ARTEX framework, including forensic clues that signal a breach. The FBI urged organizations to apply the mitigation steps in the FLASH to strengthen their security posture against ARTEX and similar agentic systems, AI tools that act autonomously toward goals.

The platform lets threat actors automate intrusion phases at machine speed, which may increase the scale and efficiency of their attacks, the FBI said.