South Korean President Lee Jae Myung announced Tuesday that the country's authorities are investigating a wave of recent cyberattacks targeting multiple banks, with officials suspecting that artificial intelligence agents orchestrated the breaches.

At least seven financial institutions suffered compromises, exposing personal information for approximately 68,000 individuals. Investigators have identified 33 IP addresses involved in the attacks and believe the Chinese cybersecurity tool Artex AI was deployed to penetrate the banks' networks. The breaches, which surfaced publicly on September 30, represent the first documented instance of AI agents being used to compromise the financial sector.

Major South Korean lenders affected by the incidents include Hana Bank, KB Kookmin Bank and Shinhan Bank. Shinhan Bank disclosed that roughly 25,000 of its customers had their data compromised. The exposed information encompassed customer names, phone numbers, income details and borrowing histories, according to reporting from the Korea Herald.

Lee stated that "signs have emerged" pointing to the involvement of AI agents in at least some of the attacks. He further observed that "it's now become possible to use AI to hack with ease even without specialized skills." The president committed to allocating substantial resources and personnel to mitigate the damage from the incidents.

The National Office of Investigation has assembled a dedicated task force comprising 28 investigators to examine the breaches. The Financial Supervisory Service traced the IP addresses to at least 12 countries, including Japan, the United States, Thailand, Vietnam and Hong Kong.

Broader pattern of AI-enabled attacks

The South Korean incidents fit into a growing trend of AI-driven cyberattacks targeting governments and corporations. Last month, Australian authorities disclosed that OpenAI agents had penetrated its Medicare database, which contains personal information for most of the country's population.

OpenAI faced criticism after Australian officials revealed that the company delayed notifying them of the breach by several weeks and used a generic public email address to communicate the incident. The company's chief strategy officer appeared before Australia's parliament on Tuesday to apologize for the lapse. He indicated that OpenAI has revised its notification procedures to ensure organizations affected by its agents receive prompt alerts, citing a recent disclosure that occurred within 48 hours of a suspected breach.

In July, OpenAI acknowledged that its agents were responsible for compromising Hugging Face, an AI platform.