ThreatBook now sells Artex Pro, a paid extension of a tool that investigators link to attacks on financial companies in Japan and South Korea.
A Japanese cybersecurity firm says attackers used ARTEX, an AI agent for automated penetration testing developed in China, in a recent wave of intrusions at Japanese companies. GMO Cybersecurity by Ierae, the security arm of GMO Internet Group, said on October 20 that it had confirmed ARTEX was used, according to an AFP report carried by the Korea Times. The firm shared its findings with law enforcement.
GMO’s investigators identified six internet servers launching AI-powered attacks and grouped them by the industries they targeted: finance, e-commerce and logistics. GMO did not say where the servers were located or who operated them. Daiki Fukumori, writing on the company’s blog, said the attacks were "almost exclusively targeting Japan". Seoul Economic Daily reported that about 80 Japanese insurers, financial firms and logistics companies were hit between September 18 and October 18, and that all six servers carried traces of ARTEX and none of other AI agents. AFP did not report that figure.
The Japanese finding extends a story that began in South Korea, and it lands days after ThreatBook, the Beijing security company that employs the tool’s developer, began marketing a commercial edition.
A tool first tied to Korean bank breaches now appears in Japan
ARTEX is open-source software that connects to outside large language models, including ChatGPT, Claude and DeepSeek, so that an AI agent can map routes into a network and run security tools. Its stated purpose is to let organizations test their own systems for weaknesses. In an October 7 analysis, CrowdStrike linked the tool to attacks on South Korean financial institutions in late September and early October. It attributed them, with moderate confidence, to a financially motivated actor who probably speaks Chinese, and said the available information did not definitively tie any individual to the activity.
The developer, who uses the GitHub handle Autumn-27, made the code private on October 8. "Given the misuse of the tool, the ARTEX project will no longer be updated and will be converted to closed source," the developer wrote, according to Reuters, as carried by The Standard. China’s Foreign Ministry said at the time that it was not familiar with the Korean case and that China consistently opposes hacking. Copies of the code downloaded before the change can still be used, the Korea JoongAng Daily reported.
Seoul Economic Daily cites South Korean regulators as saying about 100,000 records leaked from 16 institutions, including Shinhan Bank, KB Kookmin Bank, Hana Bank and Busan Bank. Reuters reported earlier that at least nine banks had been targeted. EastFrontier has covered the forensic AI-agent trail in the Korean intrusions.
ThreatBook is selling a paid edition of the same tool
On October 16, ThreatBook launched RedAgents, a subscription platform that runs several AI penetration-testing agents in parallel. Its lineup includes Artex Pro, which the company describes as an extension of the open-source Artex, built as a personal project by the RedAgents product lead. ThreatBook said Artex won a Baidu agent security contest on September 3 and that the open-source edition went closed after attackers misused it in incidents in financial services.
Seoul Economic Daily reported that ThreatBook co-founder and CEO Xue Feng posted that ARTEX "is back, with a commercial edition". ThreatBook, founded in 2015, is a venture-backed cybersecurity intelligence company headquartered in Beijing, with a release dateline that also lists Singapore, Seoul and Palo Alto. None of the reports on the Japanese finding carried a comment from ThreatBook, and no investigator has tied the attacks to the company itself.
The developer’s response to the Japanese finding was brief. According to AFP, the developer said using ARTEX for attacks was entirely contrary to its intentions and that the project will no longer be updated.
Japan’s police inquiry will decide how far the attribution goes
Japan is dealing with a broader run of breaches. Dozens of Japanese companies, including restaurant chains and a car-rental operator, have disclosed unauthorized access in recent weeks, AFP reported, without tying all of them to ARTEX. The head of the ruling party’s cybersecurity committee said this month that emerging cyber risks had reached a "state of emergency".
ThreatBook is now selling a commercial edition of a tool that investigators in Japan and South Korea link to attacks on financial companies. What police in Japan and South Korea conclude about who ran the six servers, and whether GMO publishes its full technical findings, will show how much of that link sticks to the tool and how much to the people who used it.