WASHINGTON — The FBI is warning that cybercriminals are using an open-source, artificial intelligence-powered hacking platform to automate attacks against financial institutions and organizations in numerous other industries, with some intrusions advancing from initial targeting to privileged access in less than three hours.
In a Flash Alert, the bureau said malicious actors have used the ARTEX autonomous penetration testing framework since at least August 2026 to conduct cyber campaigns worldwide. Targets have included financial services, government, technology, defense, semiconductor, energy, healthcare, manufacturing, telecommunications and legal organizations.
The FBI said ARTEX uses large language models to automate multiple stages of a cyber intrusion, from reconnaissance and vulnerability exploitation to the theft of data. The tool allows attackers with limited technical expertise to conduct complex attacks against multiple targets at the same time.

Incidents Cited
Among the incidents cited by the bureau were financially motivated attacks against South Korean financial institutions that resulted in the theft of personal information belonging to approximately 67,000 people. The alert also cited a supply chain compromise affecting ConnectWise Cloud Backup customers that provided access to more than 2,250 downstream tenants.
ARTEX was originally released on GitHub in July 2026 by a developer using the name “Autumn-27” as a tool intended for authorized security testing, according to the FBI. The developer took down the original repository Oct. 8 after reports of criminal misuse, but copies, independently maintained versions and a commercial offering called ARTEX PRO remain available.
More Than 1,600 Incidents Publicly Exposed
Researchers had identified more than 1,600 publicly exposed ARTEX console instances as of early November, the FBI said, adding that actual deployments were likely substantially higher because private installations and systems behind firewalls were not included.
The framework, written in the Go programming language, converts natural-language instructions into commands executed by specialized AI agents. It does not contain its own AI model but can connect to external models, including Anthropic Claude, OpenAI GPT, DeepSeek, GLM, Grok and Kimi, the alert said.
According to the FBI, analysis of attacker-controlled instances found Chinese AI models were the most commonly used, particularly models in the DeepSeek Flash family, which the alert associated with low cost, capabilities and limited observable safeguards when used for penetration testing.
Risks Identified
The FBI identified several major risks associated with the technology:
- Faster, simultaneous attacks: Automated agents can handle time-consuming reconnaissance and exploitation tasks, allowing smaller groups to attack multiple organizations concurrently.
- Neglected systems: Attackers can target partner portals, mobile application backends, employee-facing applications, legacy services and third-party integrations that may receive less security scrutiny.
- Supply chain intrusions: The tool can help attackers identify weaker vendors and trusted connections, potentially using compromised cloud credentials, software development pipelines or suppliers to reach larger organizations.
- Lower technical barriers: The platform enables less-experienced attackers to conduct sophisticated campaigns with relatively little direct human involvement.
The FBI urged organizations, particularly those with internet-facing applications, self-service portals, cloud infrastructure and supplier or partner systems, to follow the mitigation recommendations accompanying its alert to reduce the likelihood and impact of ARTEX-enabled intrusions.





