A Chinese artificial intelligence tool used in attacks on South Korean financial institutions was also deployed in some of the recent cyberattacks that exposed personal data at about 80 Japanese insurers, financial firms and logistics companies, according to a new analysis.
An information security subsidiary of Japanese IT company GMO Internet Group found that ARTEX, a Chinese AI penetration-testing tool that probes systems for vulnerabilities, was used in cyberattacks on Japanese companies and organizations, the Yomiuri Shimbun and NHK reported on the 20th.
Investigating at the request of the affected companies, the firm identified six overseas public servers used in attacks on Japanese businesses between Sept. 18 and Oct. 18. An analysis of all six turned up traces of ARTEX running on the servers. No traces of any other AI agent were found, leading the firm to conclude that ARTEX was most likely used in the attacks.
ARTEX is an autonomous penetration-testing program designed to link large language models so that an AI agent can analyze targets, map out intrusion routes and run security tools. It was also reportedly used in the attacks on Korean financial institutions over the past two months.
U.S. security firm CrowdStrike found that the attacks targeting Korean financial institutions took place between late September and early October, and that the attackers used ARTEX in combination with large language models. Financial regulators said about 100,000 records were leaked from 16 financial institutions, including Shinhan Bank, KB Kookmin Bank, Hana Bank and Busan Bank. The National Police Agency's cyber investigation bureau is looking into the case.
ARTEX is an AI-based cybersecurity tool developed by Chinese cybersecurity engineer Li Fuhua. Rather than running a single built-in AI model, it was designed to connect multiple large language models — including Anthropic's Claude, OpenAI's GPT and China's DeepSeek — and use them as a single agent. Hacking was not the original purpose. The developer built it to help companies and organizations examine their own networks and find security weaknesses. But because the source code was released as open source, anyone could download and modify it free of charge, and that appears to be what was exploited.
ARTEX, Taken Private, Set to Return as a Paid Service
ARTEX was released as open source on the developer platform GitHub, but after the controversy over the hacking of Korean financial firms, the developer made it private and announced that updates would stop. Now the tool is expected to reappear as a paid service.
Xue Feng, co-founder and chief executive of Threatbook, wrote on social media that ARTEX "is back, with a commercial edition," signaling the launch of a cloud-based service called RedAgent.ai, industry sources said on the 16th.
The engineer who developed ARTEX was originally a threat detection engineer at Threatbook, and ARTEX was a personal side project, according to the company. Threatbook plans to build on it and offer the tool as a paid service. Xue said the company is developing a new AI penetration-testing platform that includes ARTEX Pro, the commercial version of ARTEX, and that the engineer is now leading the project.
Besides ARTEX Pro, the platform will include Yanxun, developed by a company affiliated with Peking University; Yanji, which took first place in the individual category of Tencent's AI penetration-testing competition; and the commercial version of CyberstrikeAI, which Threatbook recently acquired. Threatbook said it will release a preview version of RedAgent.ai soon and is accepting waitlist sign-ups from users who want early access.