A suspected Chinese threat actor deployed artificial intelligence pentesting tools to break into South Korean financial institutions and steal data from at least 65,000 people, according to a report published October 7 by CrowdStrike. The cybersecurity firm traced the breaches to a financially motivated hacker who used ARTEX, an open-source agentic pentesting tool developed in China, alongside Anthropic's Claude AI model during a campaign that ran from late September through early October 2026. The attacker leveraged AI to identify system weaknesses and break into specific services at victim organizations.
CrowdStrike linked all attacks to a single IP address that hosted an ARTEX instance and an open directory containing a Claude Code markdown document with Chinese-language pentesting instructions. The ARTEX setup ran DeepSeek v4.1-flash as its main large language model backend, supplemented by GLM-5.3 from Chinese firm Zhipu AI and Grok 4.6 for extra Claude Code sessions. A second Hong Kong-based IP address controlled by the attacker held more open directories with Claude Code session histories, ARTEX configuration files, and Claude memory files. In one session, the user included personal information such as the Telegram username YY520CN and a location in Maoming, Guangdong, China—details CrowdStrike believes likely belong to the threat actor behind the ARTEX activity.
The campaign resulted in data theft from multiple South Korean financial firms, including Shinhan Bank and Yegaram Savings Bank, which reported breaches affecting 25,000 and 40,000 people respectively, according to Singapore newspaper The Straits Times. At one targeted bank, the hacker compromised a loan progress inquiry service used by financial brokers; at another, the attacker broke into an employee mobile work-support system. CrowdStrike wrote in its blog that "the use of agentic AI tooling alongside traditional offensive capabilities highlights the continued evolution observed by CrowdStrike in adversarial tradecraft." The firm noted this activity shows how AI tooling can allow a financially motivated threat actor to carry out multiple intrusions within a compressed timeframe. The total count of affected organizations remains unconfirmed, the report states.
CrowdStrike's investigation assessed with moderate confidence that the attacker is a Chinese speaker driven by financial gain. The researchers uncovered evidence that the hacker also consulted Claude for help locating Korean Telegram data sales groups to monetize the stolen records. South Korea's Financial Services Commission issued a consumer alert on October 6, warning customers of breached companies to watch for possible phishing attacks and loan scams. The agency said affected organizations will keep investigating the scope of the data breaches and share updates as they learn more. Organizations facing this new generation of AI-assisted intrusions will need to rethink detection strategies built around human attack speeds and patterns. Security teams that assume adversaries still move at traditional reconnaissance and exploitation paces may miss threats that compress weeks of manual work into hours of automated scanning.