AI Pentesting Tool Used in Attacks on South Korean Financial Firms
Summary
CrowdStrike says attackers used the ARTEX AI penetration-testing tool to steal data from South Korean financial firms. A separate report details an AI-driven credential-stuffing platform that tested more than 12 million credentials.
Key points
- CrowdStrike reported that a campaign active from late September to early October 2026 exfiltrated data from South Korean financial firms, including Shinhan Bank and Yegaram Savings Bank.
- The suspected financially motivated operator used the open-source ARTEX autonomous penetration-testing tool alongside several LLMs; the campaign has not been attributed to a known group.
- Researchers found exposed session histories and configuration files on a Hong Kong-hosted server that helped reveal the campaign's infrastructure and activity.
- South Korea's financial regulators warned consumers to watch for phishing and loan scams that could exploit leaked data.
- ARTEX's developer said the tool was intended for authorized security testing and would no longer be updated or maintained; it is being converted to closed source.
- A separate report described an AI-orchestrated credential-stuffing and account-takeover platform that used stolen credentials and automated login workflows.
- Analysis of that platform's exposed server found 12,277,358 credentials had been tested, with 11,832 classified as valid across 3,968 domains.
Article Details
- Attack Vectors
- ARTEX was used in attacks against South Korean financial firms that resulted in data exfiltration.
- A separate credential-stuffing and account-takeover platform used credentials from infostealer logs and data leaks to automate logins across target domains.
- The credential-stuffing platform used AI agents to discover login pages, fill in login forms, and classify login results.
- Defensive Notes
- South Korea's Financial Services Commission and Financial Supervisory Service urged financial consumers to watch for phishing attacks and loan scams related to leaked data.
MITRE ATT&CK
Threat Actors
Vendors
AnthropicAlso configured are Anthropic's Claude Opus 4.6, Google Gemini 2.5 Flash, and GLM-5.1 models, with Google's gemma-4-26B-A4B set up as a local model served at "127.0.0[.]1:1237."DeepSeekThe ARTEX instance has been found to use DeepSeek v4.1-flash as the main LLM backend, while using Z.ai's GLM-5.3 and SpaceXAI's Grok 4.6 to supplement the model. It's suspected that the threat actor accessed DeepSeekGoogleAlso configured are Anthropic's Claude Opus 4.6, Google Gemini 2.5 Flash, and GLM-5.1 models, with Google's gemma-4-26B-A4B set up as a local model served at "127.0.0[.]1:1237."OpenAIWhile OpenAI's GPT-5.6 is used for dork generation (aka search queries) to find authentication pages of companies in a specific sector, GPT-5.5 is used for target classification and rating. The browser agent utilizesSpaceXAIhas been found to use DeepSeek v4.1-flash as the main LLM backend, while using Z.ai's GLM-5.3 and SpaceXAI's Grok 4.6 to supplement the model. It's suspected that the threat actor accessed DeepSeek likely viaZ.aiThe ARTEX instance has been found to use DeepSeek v4.1-flash as the main LLM backend, while using Z.ai's GLM-5.3 and SpaceXAI's Grok 4.6 to supplement the model. It's suspected that the threat actor accessed DeepSeek
Products
ARTEXCybersecurity researchers have disclosed details of a targeted campaign aimed at South Korean financial organizations that used an artificial intelligence (AI) pen testing tool named ARTEX to carry out the attacks.Claude Codeafter it identified a set of open directories hosted at a Hong Kong-based IP address, exposing Claude Code session histories, Claude memory files, and ARTEX configuration files.Claude Opus 4.6Also configured are Anthropic's Claude Opus 4.6, Google Gemini 2.5 Flash, and GLM-5.1 models, with Google's gemma-4-26B-A4B set up as a local model served at "127.0.0[.]1:1237."DeepSeek v4.1-flashThe ARTEX instance has been found to use DeepSeek v4.1-flash as the main LLM backend, while using Z.ai's GLM-5.3 and SpaceXAI's Grok 4.6 to supplement the model. It's suspected that the threat actor accessed DeepSeekDeepSeek-V4-Flashis used for target classification and rating. The browser agent utilizes DeepSeek-V4-Pro and DeepSeek-V4-Flash.DeepSeek-v4-proa specific sector, GPT-5.5 is used for target classification and rating. The browser agent utilizes DeepSeek-V4-Pro and DeepSeek-V4-Flash.Gemini 2.5 FlashAlso configured are Anthropic's Claude Opus 4.6, Google Gemini 2.5 Flash, and GLM-5.1 models, with Google's gemma-4-26B-A4B set up as a local model served at "127.0.0[.]1:1237."gemma-4-26B-A4BAlso configured are Anthropic's Claude Opus 4.6, Google Gemini 2.5 Flash, and GLM-5.1 models, with Google's gemma-4-26B-A4B set up as a local model served at "127.0.0[.]1:1237."GLM-5.1Also configured are Anthropic's Claude Opus 4.6, Google Gemini 2.5 Flash, and GLM-5.1 models, with Google's gemma-4-26B-A4B set up as a local model served at "127.0.0[.]1:1237."GLM-5.3The ARTEX instance has been found to use DeepSeek v4.1-flash as the main LLM backend, while using Z.ai's GLM-5.3 and SpaceXAI's Grok 4.6 to supplement the model. It's suspected that the threat actor accessed DeepSeekGPT-5.5for dork generation (aka search queries) to find authentication pages of companies in a specific sector, GPT-5.5 is used for target classification and rating. The browser agent utilizes DeepSeek-V4-Pro andGPT-5.6While OpenAI's GPT-5.6 is used for dork generation (aka search queries) to find authentication pages of companies in a specific sector, GPT-5.5 is used for target classification and rating. The browser agent utilizesGrok 4.6been found to use DeepSeek v4.1-flash as the main LLM backend, while using Z.ai's GLM-5.3 and SpaceXAI's Grok 4.6 to supplement the model. It's suspected that the threat actor accessed DeepSeek likely via the LLM APIMozilla Firefox"An LLM agent with 46 automation tools drives an instrumented Firefox that spoofs canvas, WebGL, time zone, language, geolocation and viewport, with outsourced captcha solving. The agent finds the login page, fills in
Tools
Countries
ChinaARTEX, a recently released open-source agentic penetration testing (pentesting) tool developed in China, alongside large language models (LLMs)," the cybersecurity company said.South Koreawas active from late September to early October 2026, and resulted in data exfiltration from various South Korea-based financial firms, including Shinhan Bank and Yegaram Savings Bank.