South Korea's President Says AI Was Used to Hack Seven Banks
South Korean President Lee Jae Myung told a cabinet meeting on Tuesday that "in some hacking cases, details have emerged of the possible use of artificial intelligence," after customer data was exposed at seven financial companies in a week. Reuters carried it and Hacker News put it on the front page. Shinhan Bank lost records on about 25,000 customers through a system loan brokers use to check application status. Yegaram Savings Bank reported about 40,000. Welcome Savings Bank, KB Kookmin, Hana, BNK Busan and Hyundai Capital reported smaller incidents. Over 67,000 people in total, with names, phone numbers, income, loan data and in some cases resident registration numbers exposed. No funds were taken.
What makes this an agent story is the shape of the attacks. The Shinhan intrusion ran from the early hours of September 29 into September 30 and looked like credential stuffing at scale: leaked passwords from earlier breaches tested against a new target by something that never got tired. None of the seven breaches went through core banking. They came in through systems run by employees, contractors and loan agents, which is exactly the long tail an autonomous scanner finds and a human red team never gets budget for. Korean papers quote Lee saying AI has made it possible "to hack easily without special ability." Investigators reportedly found traces of an open-source Chinese-language autonomous penetration-testing tool on a server tied to the attacks.
The response is the part worth filing. Police stood up a 28-member team. The Financial Services Commission and Financial Supervisory Service ordered hundreds of institutions into emergency security checks, distributed dozens of malicious IP addresses, set up a 24-hour response desk and asked cloud providers to block suspicious overseas activity. That is a national regulator treating agent-driven intrusion as a systemic event rather than seven separate incidents.
Put it next to the week's other items. Mistral is shipping open weights that reproduce exploits at 82% because defenders asked for it. Anthropic's Haiku 5.5 safeguards still block penetration testing. And the Korean banking system just learned that the attacker's cost of probing every third-party portal has dropped to roughly zero. The offense-cost-collapse thread now has a sovereign-level casualty.
Links: reuters.com/world/south-koreas-lee-says-ai-appears-have-been-used-bank-hacks-2026-10-06/, koreaherald.com/article/10891104
← Back to all articles
What makes this an agent story is the shape of the attacks. The Shinhan intrusion ran from the early hours of September 29 into September 30 and looked like credential stuffing at scale: leaked passwords from earlier breaches tested against a new target by something that never got tired. None of the seven breaches went through core banking. They came in through systems run by employees, contractors and loan agents, which is exactly the long tail an autonomous scanner finds and a human red team never gets budget for. Korean papers quote Lee saying AI has made it possible "to hack easily without special ability." Investigators reportedly found traces of an open-source Chinese-language autonomous penetration-testing tool on a server tied to the attacks.
The response is the part worth filing. Police stood up a 28-member team. The Financial Services Commission and Financial Supervisory Service ordered hundreds of institutions into emergency security checks, distributed dozens of malicious IP addresses, set up a 24-hour response desk and asked cloud providers to block suspicious overseas activity. That is a national regulator treating agent-driven intrusion as a systemic event rather than seven separate incidents.
Put it next to the week's other items. Mistral is shipping open weights that reproduce exploits at 82% because defenders asked for it. Anthropic's Haiku 5.5 safeguards still block penetration testing. And the Korean banking system just learned that the attacker's cost of probing every third-party portal has dropped to roughly zero. The offense-cost-collapse thread now has a sovereign-level casualty.
Links: reuters.com/world/south-koreas-lee-says-ai-appears-have-been-used-bank-hacks-2026-10-06/, koreaherald.com/article/10891104
Comments