CrowdStrike: Hacker Used ARTEX Agent on Korean Finance Firms

CrowdStrike said on 7 October 2026 that a threat actor used ARTEX, a recently released open-source agentic penetration testing tool developed in China, alongside large language models in a campaign against South Korean financial organizations that resulted in exfiltrated data. The campaign was active from late September to early October 2026, it says.

According to CrowdStrike, the ARTEX instance used DeepSeek v4.1-flash as its primary model, and the actor supplemented it with GLM-5.3 and Grok 4.6 in additional Claude Code sessions.

Business Pill · THE MODEL AND THE HARNESS

A short explainer of the difference between an AI model and the harness around it: the tools, memory and instructions that turn a model into an agent. It teaches the general idea only and says nothing about any company in this story.

The key insight: As we read it, the report separates the harness from the model. CrowdStrike describes Claude Code sessions and an open-source agent doing the orchestration, while the main model underneath was DeepSeek v4.1-flash, likely reached through a likely reseller, with GLM-5.3 and Grok 4.6 used in additional sessions.

What CrowdStrike Found

CrowdStrike says its analysis of open directories controlled by the actor uncovered Claude Code session histories, ARTEX configuration files and Claude memory files, which gave direct insight into how the actor worked.

It describes a two-server setup: a Hong Kong-based address served as the primary attacker-controlled infrastructure, and a second address hosted the ARTEX instance likely responsible for the Korean attacks. A Claude Code markdown file on that server held a Chinese-language pentesting prompt.

The actor likely accessed DeepSeek through a likely API proxy or reseller, CrowdStrike says, and used nine proxy IP addresses during the activity. It also asked Claude where threat actors typically sell Korean data breach information and for help finding Korean Telegram data sales groups.

The toolchain CrowdStrike describes in its report of 7 October 2026: the open-source ARTEX agent, DeepSeek v4.
The toolchain CrowdStrike describes in its report of 7 October 2026: the open-source ARTEX agent, DeepSeek v4.1-flash as the primary model, GLM-5.3 and Grok 4.6 in additional Claude Code sessions, and nine proxy IP addresses.
CrowdStrike assessment: unattributed, likely Chinese speaker, financially motivated
CrowdStrike’s assessment in its report of 7 October 2026: not attributed to a named adversary; likely a Chinese speaker and financially motivated (moderate confidence); active late September to early October 2026.

Who and Why

CrowdStrike has not attributed the activity to a named adversary. It assesses with moderate confidence that the actor is likely a Chinese speaker and financially motivated, based on the use of the Chinese-developed ARTEX and on Chinese-language prompts.

Citing industry reports, it says one affected bank had a loan progress inquiry service used by brokers breached and another had an employee mobile work-support system compromised, and that the number of organizations affected remains unconfirmed.

CrowdStrike’s Assessment

CrowdStrike writes that the activity shows how AI tooling can enable a financially motivated actor to conduct multiple intrusions within a short time span, and assesses that adversaries will likely continue to experiment with AI tooling to increase their operational tempo and capabilities.

Its report maps the actor’s use of ARTEX to the MITRE ATT&CK technique for obtaining artificial intelligence capabilities.

The Structural Read

The agent came off the shelf. CrowdStrike describes ARTEX as a recently released open-source agentic pentesting tool developed in China.

The models were interchangeable parts. Three providers’ models appear in the sessions CrowdStrike describes, and the primary one was likely accessed through a likely API proxy or reseller.

Speed is the change CrowdStrike highlights. It says the activity shows how AI tooling can let a financially motivated actor carry out multiple intrusions within a short time span.

CrowdStrike, 7 October 2026

“This activity demonstrates how AI tooling can enable a financially motivated threat actor to conduct multiple intrusions within a short time span.”

Three Implications

AN OPEN-SOURCE AGENT CrowdStrike says the actor used ARTEX, an open-source agentic pentesting tool developed in China.

SEVERAL MODELS AT ONCE DeepSeek v4.1-flash was the primary model, with GLM-5.3 and Grok 4.6 supplementing it, per CrowdStrike.

SCOPE STILL OPEN CrowdStrike says the number of organizations affected remains unconfirmed.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Agentic Harness War.

The framework’s starting point: “The surface, not the model, governs autonomy.”

As we read it, CrowdStrike’s report is a case of that on the attack side: what it found on the actor’s servers were harness artifacts, session histories, agent configuration and memory files, while the models behind them came from several providers.

What Is Not Established

We read CrowdStrike’s report in full. It does not name the affected institutions or give the volume of data taken, and the bank-level details it gives come from industry reports it cites. It lists personal details found in one prompt but says they cannot be definitively tied to the actor; we do not reproduce them. We did not contact CrowdStrike or the model providers.

Business Engineer Framework

The Agentic Harness War

A Business Engineer framework on why the agentic surface, not the model, is where value and control pool in the AI stack.

Read the Map of AI →

The Bottom Line

CrowdStrike says a likely Chinese-speaking, financially motivated actor used an instance of the open-source ARTEX agent, running on DeepSeek v4.1-flash and likely responsible for the attacks, against South Korean financial organizations between late September and early October 2026; the actor also used GLM-5.3 and Grok 4.6 in additional Claude Code sessions.

94,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. We read CrowdStrike’s report of 7 October 2026 in full; bank-level details are from industry reports it cites. Personal details CrowdStrike lists from one prompt are not reproduced here. We did not contact CrowdStrike. Nothing here is a forecast, and nothing here is financial or investment advice.

Sources: CrowdStrike: Unknown Threat Actor Uses AI-Driven ARTEX to Target South Korean Finance (7 Oct 2026)

Scroll to Top
FourWeekMBA