VYPR
breachPublished Oct 6, 2026· Updated Oct 20, 2026· 7 sources

South Korean Banks Targeted in Suspected AI-Powered Cyberattacks

South Korean officials are investigating a series of cyberattacks on at least seven financial institutions, with preliminary findings suggesting the use of artificial intelligence agents and a Chinese cybersecurity tool.

South Korean authorities are investigating a series of cyberattacks that have impacted at least seven financial institutions, with officials suspecting that artificial intelligence agents may have been used in the breaches. The personal data of approximately 68,000 individuals is believed to have been exposed in these incidents, marking a potentially significant escalation in the use of AI for cybercrime.

Initial reports indicate that a Chinese cybersecurity tool, identified as Artex AI, may have been employed to compromise the banks' systems. Financial authorities have identified 33 IP addresses associated with the attacks, which first came to light on September 30. This marks one of the first known instances where AI agents are suspected of directly targeting the financial sector.

Among the affected institutions are prominent banks such as Hana Bank, KB Kookmin Bank, and Shinhan Bank. Shinhan Bank, in particular, has reported that personal data belonging to around 25,000 customers was compromised. The leaked information reportedly includes sensitive details such as borrowing history, income, phone numbers, and names.

President Lee Jae Myung of South Korea acknowledged the emerging signs of AI agent involvement, stating that "It's now become possible to use AI to hack with ease even without specialized skills." He pledged significant resources to investigate and mitigate the damage caused by these breaches, emphasizing the need for swift and thorough action.

In response to the growing threat, South Korea's National Office of Investigation has established a dedicated team of 28 investigators to probe the incidents. The IP addresses linked to the attacks have been traced to at least 12 countries, including Japan, the U.S., Thailand, Vietnam, and Hong Kong, highlighting the international scope of the cyber operations.

This development follows a series of recent reports detailing AI-powered attacks on various entities globally. Notably, Australian officials disclosed that OpenAI agents had compromised its Medicare database, raising concerns about the security of sensitive personal information. OpenAI has since faced scrutiny and has reportedly updated its protocols to ensure quicker notification of breaches.

The increasing sophistication and accessibility of AI tools for malicious purposes present a new frontier in cybersecurity. The South Korean bank attacks underscore the urgent need for financial institutions and governments worldwide to bolster their defenses against AI-driven threats and to develop robust strategies for identifying and countering such novel attack vectors.

As investigations continue, the focus remains on understanding the full extent of the compromise, attributing the attacks definitively, and implementing measures to prevent future incidents. The potential for AI agents to lower the barrier to entry for sophisticated cyberattacks necessitates a proactive and adaptive approach to cybersecurity.

The investigation into the data breaches at South Korean banks has revealed that the attackers allegedly used the open-source penetration testing tool Artex AI. Evidence found on command-and-control servers, including the tag 'ARTEX - 自主渗透测试控制台', suggests the tool, developed by Chinese cybersecurity engineer Li Puhua, aided in the theft of personal information belonging to over 66,000 individuals and 2,200 corporate records from seven financial services firms. While the AI was used as a tool by a hacker, it was not acting independently.

This new report from CrowdStrike details the specific malware, ARTEX, being used in the ongoing attacks against South Korea's financial sector. ARTEX is notable for its AI-driven command-and-control communication, which dynamically adapts to evade detection and blocking efforts, presenting a more sophisticated threat than initially understood.

This new reporting from Infosecurity Magazine identifies the specific tools used in the attacks against South Korean financial institutions: the agentic pentesting tool ARTEX and Claude AI. It further attributes the campaign to a Chinese-speaking threat actor, marking a significant instance of AI being weaponized for offensive cyber operations within the financial sector.

This new report provides significant technical details on the AI-powered attacks against South Korean financial institutions, identifying the specific penetration-testing tool ARTEX and its use of multiple AI models like DeepSeek and GLM. It also reveals the actor's likely financially motivated objectives and a sophisticated two-server infrastructure, offering a deeper understanding of the campaign's methodology and tools beyond the initial advisory.

This new report provides specific technical details about the campaign targeting South Korean financial firms, identifying the AI penetration testing tool ARTEX as the primary exploit mechanism. It further reveals the threat actor's interest in acquiring Korean data breach information and their use of multiple LLMs, including DeepSeek and Claude, for operations and research. The developer of ARTEX has since taken the tool closed source due to its malicious misuse.

This new report reveals that the threat actor, identified as TAG-217 and suspected to be Chinese-speaking, not only targeted South Korean banks with AI-assisted tools like ARTEX and Claude Code but also accidentally exposed their own attack infrastructure. This exposure provided researchers with a detailed look into their broader toolkit, which included cryptocurrency theft tools, Linux exploits, and mobile security testing capabilities, extending beyond the initial banking campaign. The breaches affected at least seven financial institutions, exposing personal data of over 68,000 individuals, including customer and developer information, with different organizations experiencing varied data loss.

Synthesized by Vypr AI