According to The Wall Street Journal, as recounted by The Insider, the attackers used the Chinese AI agent Artex AI to attack South Korea’s largest banks and stole the personal data of 68,000 people. The stolen information included data on customers and employees, including annual income and consumer credit limits.
The attacks were discovered last week, and at least seven financial companies were affected. On Tuesday, the South Korean National Police Agency’s cyberterrorism unit launched an investigation.
According to the publication, the attacks were carried out from more than two dozen IP addresses in about ten countries, including the United States, Japan, and Germany. The perpetrators have not yet been identified.
Artex is an open-source agent created by Chinese cybersecurity specialist Li Puhua, known by the pseudonym Autumn. The tool was developed to find vulnerabilities in organizations’ networks; such agents connect models like Anthropic’s Claude, OpenAI’s ChatGPT, and DeepSeek and use them to find vulnerabilities and plan attacks on their own.
The authorities did not specify which models were used in the bank breach. After the first reports of cyberattacks, Artex developers banned its use for unauthorized intrusion and data theft, and South Korean authorities demanded that banks strengthen their defenses. In late September, Australian Prime Minister Anthony Albanese said that an OpenAI company AI agent had in June entered the Medicare health statistics portal without permission; according to him, the authorities learned of the incident only on September 10.
Sources
