An inadvertent error by hackers has unveiled the use of artificial intelligence tools in breaches targeting South Korean financial institutions. The mishap exposed several servers, shedding light on the AI models and scripts that facilitated the cyberattacks.
Details of the Cyberattack
Between late September and early October 2026, data from seven South Korean financial institutions was compromised. The breach affected the personal information of approximately 66,000 to 67,000 individuals and 2,200 businesses. The stolen data included names, job details, income information, and loan records.
CrowdStrike linked the breach to ARTEX, an open-source AI penetration testing framework originating from China. Their analysis suggested that a single Chinese-speaking hacker, driven by financial motives, rather than an organized group, utilized ARTEX to execute the attack.
Investigative Findings
The breach was traced back to a key server at IP address 38.244.50[.]120, and the domain xcai[.]pro was identified as a reseller for accessing AI models. Genians confirmed these findings, leading to the removal of the ARTEX GitHub page and its Docker image on October 11, although copies of these resources remain accessible online.
Further investigations by Hunt.io revealed additional exposed files on a Hong Kong jump server at 23.248.226[.]250. This server contained substantial evidence of the hacker’s activities, including a state file that documented the use of Claude Code, version 2.1.221, initiated over 500 times since August 2026.
Impact and Recommendations
The Hong Kong server also housed various custom ‘skills’ for the AI agent, used for tasks such as reconnaissance, code review, and web testing. Additionally, it contained a credentials file with a warning against copying passwords into sessions. The presence of numerous tools suggested a broader hacking operation.
The incident underscores the need for organizations to enhance their cybersecurity measures. Experts recommend restricting public access to administrative services, resetting SSH passwords, and monitoring for unusual network activities. It’s crucial to secure AI gateways and prevent unauthorized script downloads.
South Korean authorities have arrested a suspect in a related case involving ARTEX, though there is no definitive evidence linking this individual to the same hacking activities.
Indicators of compromise include specific IP addresses and domains, which should be carefully monitored to prevent further breaches. Security Operation Centers (SOCs) are urged to integrate threat intelligence to swiftly address potential incidents.
