KOSUNI CRYPTO
Korea's Crypto Pulse, in English
$BTC $ETH $XRP $SOL $DOGE
News
Neutral

Korean Bank Hacker Inquires About Selling Stolen Data on Claude

Published October 8, 2026 8:23 PM · 4 views
Korean Bank Hacker Inquires About Selling Stolen Data on Claude

CrowdStrike has identified that a suspect in the recent South Korean bank data leak inquired about where to sell leaked data through an AI coding tool. The company discovered this request in session logs stored in an open directory on the attacker's server.

Multiple South Korean banks publicly disclosed customer information breaches last week. According to CrowdStrike's October 7 report, this attack campaign utilized AI penetration testing tools developed in China and various language models.

The attacks occurred between late September and early October across several financial institutions. Shinhan Bank confirmed the breach on September 30 and reported that approximately 25,000 customers were affected the next day. Attackers bypassed identity verification procedures in a mobile service tracking loan applications.

Stolen information included names, phone numbers, annual income, calculated loan limits, and 66 South Korean resident registration numbers. KB Kookmin Bank announced that data for 119 customers was leaked through an internal mobile system on October 2. Hana Bank reported leaks affecting 89 people, while BNK disclosed the exposure of 11 external contractor records.

President Lee Jae-myung raised questions about AI during a subsequent cabinet meeting. Police have launched a full investigation.

CrowdStrike's October 7 report revealed that session logs and configuration files for Anthropic's AI coding assistant Claude Code were exposed on the attacker's open server. The attackers used ARTEX, an open-source penetration testing tool developed in China, according to the report.

The primary infrastructure of the attack was hosted on a server located in Hong Kong. CrowdStrike identified this IP address as the origin point for the Korean attacks. The company noted that ARTEX instances utilized DeepSeek v4.1-flash as their main AI model, while other Claude Code sessions used Zhipu AI's GLM-5.3 and xAI's Grok 4.6.

DeepSeek previously appeared in a Taiwan security firm's August report linking it to Chinese hackers. The Taiwanese company stated that after implementing DeepSeek and open-source AI tools, attacks by China-linked hacking groups doubled.

The attacker also asked Claude about where stolen data from the South Korean breach is typically sold on Telegram markets. The same user requested help finding a Korean Telegram group where such data sales occur.

CrowdStrike has not disclosed the perpetrator behind this campaign but assessed that Chinese-speaking attackers with financial motives are moderately likely to be responsible, based on ARTEX usage and Chinese prompts.

In another session, the attacker asked Claude to draft a security researcher resume using ARTEX results. The prompt included a Telegram ID, age 26, and address in Maoming, Guangdong Province. CrowdStrike noted these details might belong to the attacker but could not confirm the connection. The same Telegram ID appeared in another session exploring vulnerabilities in a Telegram-based NFT gift marketplace.

CrowdStrike warned that AI tools can help financially motivated attackers quickly execute multiple intrusions. Anthropic also acknowledged that AI assists less skilled hackers with advanced attacks, according to their statement.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: 한국 은행 해커, Claude에 탈취 데이터 판매처 문의…CrowdStrike 밝혔다