Zscaler Details Global AI-Assisted Hacking Campaign Using ARTEX and AI Agents
Summary
Zscaler ThreatLabz analyzed a global campaign combining customized ARTEX, AI assistants and automation. Recovered configurations and prompts show coordinated investigations targeting organizations across sectors and regions, with apparent focus on monetizing personal,
Key points
- ThreatLabz reports the campaign extended beyond South Korean financial institutions, targeting organizations across multiple industries and regions.
- The operator combined a customized ARTEX penetration-testing framework with Claude Code, Codex, Grok workspaces and custom scripts.
- ARTEX and related workspaces were configured to request DeepSeek, GLM and Grok models; the underlying models serving some sessions could not be independently verified.
- Prompts assigned AI agents investigation and review roles, while scripts synchronized instructions and project records across accounts and servers.
- Saved project files and an automated context hook helped AI sessions track targets, findings and progress and resume investigations.
- Prompt histories requested personal, transaction and financial data and discussed selling compromised data, suggesting financial motivation.
- Zscaler recommends reducing internet exposure and enforcing least privilege; its listed campaign detection name is AI.HackTool.ARTEX.
Article Details
- Attack Vectors
- The operator used a customized ARTEX deployment with AI assistants and automation scripts to investigate target websites, servers, and applications for vulnerabilities and unauthorized access.
- The operator configured proxy exits and routed target connections through a jump server, with SSH tunnels supporting communication between the work server and jump server.
- Prompt histories directed attempts to obtain administrative access, credentials, personal and financial records, and data from targeted services.
- Defensive Notes
- Adopt a Zero Trust architecture, minimize internet exposure, and enforce least-privilege access.
- Identify and prioritize exposures for remediation, and use decoys to detect unauthorized exploration.
Indicators of compromise
| Type | Indicator | Context |
|---|---|---|
| IPV4 | 23[.]248[.]226[.]250 | Listed as a work server associated with the operator's campaign. |
| IPV4 | 23[.]248[.]226[.]251 | Listed as a work server associated with the operator's campaign. |
| IPV4 | 23[.]248[.]226[.]252 | Listed as a work server associated with the operator's campaign. |
| IPV4 | 23[.]248[.]226[.]253 | Listed as a work server associated with the operator's campaign. |
| IPV4 | 23[.]248[.]226[.]254 | Listed as a work server associated with the operator's campaign. |
| IPV4 | 38[.]244[.]50[.]120 | Operator's jump server, used as the default route for connections to targets and to host additional AI sessions. |
MITRE ATT&CK
T1041 · Exfiltration Over C2 ChannelThe operator sought to extract target data, including database records and transaction histories, for later monetization.T1078 · Valid AccountsPrompts sought unauthorized administrative access and credentials for targeted services.T1090 · ProxyProxy configurations supplied separate outbound IP addresses for projects, and target connections were routed through a jump server.T1190 · Exploit Public-Facing ApplicationPrompts sought vulnerabilities that could provide unauthorized access to target-facing applications and services.T1213 · Data from Information RepositoriesThe operator directed investigations to obtain customer, user, transaction, and other database records.T1572 · Protocol TunnelingSSH tunnels were configured for bidirectional communication between the work server and jump server.T1595 · Active ScanningPrompt histories instructed AI tools to scan or investigate target organizations and discover attack surfaces.
Vendors
Products
ARTEXIn October 2026, Zscaler ThreatLabz investigated AI-assisted cyberattacks that combined the ARTEX penetration testing framework with AI coding assistants and custom automation scripts. Initial reporting of the incidentBackblaze B2Investigation assets were archived to Backblaze B2, with reports, supporting evidence, source code, data exports, and backups organized into separate directories. This structure followed the operator's directive toClaude Codediscovered information about an AI-assisted attack combining a customized ARTEX deployment with Claude Code, Codex, and Grok workspaces. Custom scripts shared instructions and supported handoffs between AICodexinformation about an AI-assisted attack combining a customized ARTEX deployment with Claude Code, Codex, and Grok workspaces. Custom scripts shared instructions and supported handoffs between AI sessions.DeepSeekARTEX was configured to use a DeepSeek model through an OpenAI-compatible gateway, but other configurations also specified DeepSeek, GLM, and Grok.GLMARTEX was configured to use a DeepSeek model through an OpenAI-compatible gateway, but other configurations also specified DeepSeek, GLM, and Grok.Grokabout an AI-assisted attack combining a customized ARTEX deployment with Claude Code, Codex, and Grok workspaces. Custom scripts shared instructions and supported handoffs between AI sessions.Zscaler DeceptionManagement, part of Agentic SecOps, helps identify and prioritize exposures for remediation, while Zscaler Deception uses decoys to detect unauthorized exploration and divert attacks from legitimate assets. Zscaler Exposure Managementapplications invisible to the public internet and restricting access to authorized users. Zscaler Exposure Management, part of Agentic SecOps, helps identify and prioritize exposures for remediation, while ZscalerZscaler Zero Trust ExchangeTrust architecture that minimizes internet exposure and enforces least-privilege access. The Zscaler Zero Trust Exchange supports these principles by making applications invisible to the public internet and
Tools
apktoolThe corresponding Dockerfile adds tools including nmap, sqlmap, nuclei, httpx, jadx, apktool, and frida-tools. It also adds Python libraries for web processing and spreadsheet output and replacesfrida-toolsThe corresponding Dockerfile adds tools including nmap, sqlmap, nuclei, httpx, jadx, apktool, and frida-tools. It also adds Python libraries for web processing and spreadsheet output and replaces the originalGrok2APIits underlying AI model. For the Claude Code integration, the operator installed a bridge service called grok2api on the jump server. Prompts assigned extensive research tasks to Grok on that server and other AIhttpxartex-local:perf2-20261003. The corresponding Dockerfile adds tools including nmap, sqlmap, nuclei, httpx, jadx, apktool, and frida-tools. It also adds Python libraries for web processing and spreadsheet outputjadx artex-local:perf2-20261003. The corresponding Dockerfile adds tools including nmap, sqlmap, nuclei, httpx, jadx, apktool, and frida-tools. It also adds Python libraries for web processing and spreadsheet output andNmapuses a custom image tagged artex-local:perf2-20261003. The corresponding Dockerfile adds tools including nmap, sqlmap, nuclei, httpx, jadx, apktool, and frida-tools. It also adds Python libraries for web processingNucleiimage tagged artex-local:perf2-20261003. The corresponding Dockerfile adds tools including nmap, sqlmap, nuclei, httpx, jadx, apktool, and frida-tools. It also adds Python libraries for web processing and spreadsheetsqlmapa custom image tagged artex-local:perf2-20261003. The corresponding Dockerfile adds tools including nmap, sqlmap, nuclei, httpx, jadx, apktool, and frida-tools. It also adds Python libraries for web processing and
Countries
Canadaan opportunistic approach. The recorded prompts included requests to scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa,Chinaapproach. The recorded prompts included requests to scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, SouthCubaapproach. The recorded prompts included requests to scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea,Hong KongThe recorded prompts included requests to scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan,Indiaprompts included requests to scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, theIndonesiaincluded requests to scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the UnitedJapanrequests to scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United ArabKazakhstanto scan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE),Kenyascan or investigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), theMalaysiainvestigate organizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the UnitedNigeriaorganizations in Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, theSouth Africain Canada, China, Cuba, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, the United States,South KoreaHowever, ThreatLabz discovered additional information about the attack, which went far beyond South Korea. The AI-assisted attacks were widespread and targeted organizations around the globe and in multipleTaiwanHong Kong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, the United States, and Vietnam. ThailandKong, India, Indonesia, Japan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, the United States, and Vietnam. United Arab EmiratesJapan, Kazakhstan, Kenya, Malaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, the United States, and Vietnam. United KingdomMalaysia, Nigeria, South Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, the United States, and Vietnam. United StatesSouth Africa, South Korea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, the United States, and Vietnam. VietnamKorea, Taiwan, Thailand, the United Arab Emirates (UAE), the United Kingdom, the United States, and Vietnam.
Industries
Cryptocurrencytargeting beyond South Korean financial institutions, including dating applications, SMS providers, cryptocurrency and payment platforms, gift-card portals, gambling sites, retail platforms, educational services,Dating servicesSocial and dating servicesGamblingincluding dating applications, SMS providers, cryptocurrency and payment platforms, gift-card portals, gambling sites, retail platforms, educational services, logistics companies, and legal settlement systems. logisticsand payment platforms, gift-card portals, gambling sites, retail platforms, educational services, logistics companies, and legal settlement systems. Retailapplications, SMS providers, cryptocurrency and payment platforms, gift-card portals, gambling sites, retail platforms, educational services, logistics companies, and legal settlement systems.