MAIN ARTICLE
An unexpected trail in AI session records
In an October 7 analysis, CrowdStrike said it examined accessible server directories associated with a campaign against South Korean financial institutions. The material included Claude Code sessions, ARTEX configuration files and AI memory documents. These records offered clues about how a perpetrator combined automated penetration-testing software with language models. The findings remain a private cybersecurity assessment, not the outcome of a criminal prosecution.
CrowdStrike described queries about markets for Korean breach data and Telegram groups selling such information. Another session sought help drafting a security researcher résumé, containing an age, educational details, a Telegram handle and an association with Maoming in Guangdong, China. Self-reported chat information may be fictitious or planted; it offers leads for investigators, not conclusive identification.
Customer information and overlooked gateways
Shinhan Bank reported exposure of information relating to approximately 25,000 customers. KB Kookmin Bank disclosed a leak involving 119 customers, and Hana Bank reported 89 affected people, according to Korean reporting.














