Suspected AI-assisted South Korean bank campaign exposed customer contact and borrowing records

The October 16 account reports roughly 100,000 customer contact and borrowing records across at least 16 financial institutions; the October 11 investigation describes a narrower seven-institution scope

October 6 is the first verified AI-related disclosure in the checked sources, not a claim to the earliest report worldwide: that Reuters account also mentions October 2 bank-attack reporting. The October 11 Korea Herald account and October 16 roundup add suspected ARTEX and LLM-assisted vulnerability scanning, including a reported DeepSeek role. The campaign therefore predates the weekly window and belongs in backfill. Attacks are described across late September and early October; there is no defensible single occurrence date. The later roundup reports roughly 100,000 contact and borrowing records at at least 16 institutions, while the October 11 investigation discusses seven institutions; their scopes are different and should not be reconciled into a fabricated total. Passwords, credit-card numbers and transaction histories were reportedly not compromised. Customer data exposure is reported as actual, but AI attribution and the operators' identities remain uncertain; automated scanning is not proof of an autonomous end-to-end breach. The campaign is coded as human-led use of AI output, with a debatable agency boundary. News syndication supplies the evidence, not a fetched regulator or forensic report. The Japanese incidents in the roundup lack per-case agent mechanisms and are not split into additional AI intrusions.