"The use of agentic AI tooling alongside traditional offensive capabilities highlights the continued evolution observed by CrowdStrike in adversarial tradecraft. This activity demonstrates how AI tooling can enable a financially motivated threat actor to conduct multiple intrusions within a short time span," CrowdStrike wrote in a blog published on October 7.
How ARTEX and Anthropic’s Claude were used in the campaign
CrowdStrike reported that a suspected China-based threat actor deployed ARTEX — a recently released open-source agentic pentesting tool developed in China — together with Anthropic’s Claude AI model during a series of intrusions that ran from late September to early October 2026. According to the firm, the attacker used ARTEX primarily to discover vulnerabilities and compromise specific services inside victim organizations. The actor also asked Claude to help locate Korean Telegram data sales groups as potential marketplaces to sell stolen information.
Infrastructure links and forensic artifacts tying the intrusions together
Investigators linked the campaign to a single IP address that hosted an ARTEX instance and an open directory containing a Claude Code markdown document. That document included a Chinese-language pentesting prompt that instructed a large language model how to perform pentesting activities. The ARTEX instance used DeepSeek v4.1-flash as the primary LLM backend; the actor supplemented that LLM with GLM-5.3 (from Chinese firm Zhipu AI) and Grok 4.6 to run additional Claude Code sessions.
CrowdStrike identified a separate Hong Kong–based IP address that appeared to function as primary attacker-controlled infrastructure. Open directories on that host contained Claude Code session histories, ARTEX configuration files and Claude memory files. In one recovered Claude Code session, the user supplied personal details — including the Telegram username YY520CN and the location Maoming, Guangdong, China — which CrowdStrike assessed were likely linked to the individual who conducted the ARTEX-related activity.
This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildData exfiltration from South Korean banks and public response
The campaign resulted in exfiltrated data from multiple South Korea–based financial firms, CrowdStrike reported. Citing Singapore-based newspaper The Straits Times, the reporting named Shinhan Bank and Yegaram Savings Bank as affected, with breaches impacting 25,000 and 40,000 people, respectively. CrowdStrike said at one breached bank the attacker compromised a loan progress inquiry service used by financial brokers; at another, the actor pierced an employee mobile work–support system.
The firm emphasized that the total number of organizations affected remained unconfirmed at the time of its blog. South Korea’s Financial Services Commission issued a consumer alert on October 6 warning customers of the hacked companies to be vigilant for potential phishing attacks and loan scams. The agency added that affected organizations will continue to investigate the extent of the data breaches and provide updates accordingly.
What this means for technologists, the Financial Services Commission, and affected customers
- Technologists and security teams: The incident shows AI-driven agentic tooling can be integrated into attacker toolchains to accelerate vulnerability discovery and exploitation; teams will need to monitor for novel LLM-backed components such as DeepSeek v4.1-flash, GLM-5.3 and Grok 4.6 appearing in telemetry or on attacker infrastructure.
- The Financial Services Commission and regulators: The Commission’s consumer alert on October 6 illustrates a regulatory focus on rapid public warning and oversight; regulators will be watching investigative outcomes and the scope of breaches as affected organizations report findings.
- Affected customers and enterprise users: Individuals whose data may have been exposed are being asked to watch for phishing and loan scams, per the Commission; affected firms are continuing investigations and are expected to update customers as new information becomes available.
CrowdStrike concluded its analysis by assessing with moderate confidence that the attacker is a Chinese speaker and financially motivated. The firm’s work links the same attacker-controlled infrastructure to multiple intrusions and to explicit use of AI tooling for both technical compromise and post-exfiltration monetization efforts. How many additional organizations beyond the named banks were impacted remains unconfirmed; affected organizations and regulators will be the primary sources for further disclosure as investigations continue.
