Threat Intelligence firm identified a financially motivated cyber campaign targeting South Korean financial organizations between late September and early October 2026, resulting in data exfiltration. Reported incidents involved a loan progress inquiry service used by financial brokers and an employee mobile work-support system. Although the total number of affected organizations remains unconfirmed, overlapping IP addresses suggest connections between the incidents. The investigation uncovered threat actor-controlled infrastructure containing Claude Code session histories, ARTEX configuration files, and Claude memory files, providing insight into the attacker’s operations.

The threat actor leveraged ARTEX, an open-source agentic penetration testing tool developed in China, alongside large language models (LLMs) to conduct attacks. Analysis revealed a two-server infrastructure: a Hong Kong-based server served as the primary attacker-controlled system, while IP address 38.244.50[.]120 hosted the ARTEX instance. The operation primarily used DeepSeek v4.1-flash as its LLM backend, supplemented by GLM-5.3 and Grok 4.6 for additional Claude Code sessions. The attacker likely accessed DeepSeek through the API proxy or reseller xcai[.]pro. Investigators also identified nine proxy IP addresses associated with ARTEX-related activity.

Claude Code session histories revealed that the attacker sought information about selling stolen South Korean data and locating Telegram groups involved in data sales. The sessions also contained a request to generate a security researcher résumé highlighting the results of ARTEX-related operations. Personal details in the prompt, including a Chinese university, a location in Guangdong, and the Telegram username, may provide investigative leads. However, researchers could not definitively link these details to the attacker.

Security firm assesses with moderate confidence that the threat actor is likely Chinese-speaking and financially motivated, based on Chinese-language prompts and the use of ARTEX. The campaign demonstrates how agentic AI and traditional offensive capabilities can facilitate multiple intrusions within a short period. Adversaries are likely to continue adopting AI-powered tools to improve operational efficiency, accelerate attacks, and expand their capabilities.