On November 13 the FBI Cyber Branch published FLASH-20261113-001 on ARTEX, an open-source, LLM-driven multi-agent penetration testing platform. Since at least August 2026, various threat actors have used it to automate intrusion phases from reconnaissance to data exfiltration against financial, government, technology, defense, semiconductor, energy, healthcare and other sectors, with some intrusions going from targeting to privileged access in under three hours. The original developer pulled the GitHub repo in October, but forks, mirrors and a commercial ARTEX PRO keep it available, and researchers found more than 1,600 internet-exposed ARTEX consoles in early November.
Key Takeaways
- ✓ARTEX is written in Go and turns high-level goals given in natural language into commands run by specialized agents, needing almost no human input after the objective is set; it ships no model of its own and plugs into external LLMs including Claude, GPT, DeepSeek, GLM, Grok and Kimi
- ✓The FBI says Chinese models are the most common in attacker-controlled ARTEX instances, with DeepSeek Flash family models dominating thanks to their capability-to-cost balance and no observable guardrails when requests are framed as penetration testing
- ✓Known campaigns include attacks on South Korean financial institutions that exfiltrated personal data of about 67,000 people, and a ConnectWise Cloud Backup supply chain compromise that reached more than 2,250 downstream tenants
- ✓Developer Autumn-27 released it on GitHub in July; it reached an estimated 1,500 to 2,000 stars in 10 weeks and the repo was pulled on October 8, but the FBI says forks, localized variants, archived copies and a commercial ARTEX PRO keep spreading
- ✓FBI advice: inventory every internet-facing app including partner portals, mobile backends and legacy services, keep cloud admin, Kudu/SCM publishing endpoints and backup platforms on trusted networks, and scan repos and CI/CD pipelines for embedded secrets and rotate them
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background
AI coding agents can run commands, read files and call APIs on their own, and the same abilities can be turned on victims. ARTEX was open-sourced for "authorized security testing", and CrowdStrike had already reported its use against South Korean banks; now the FBI has issued a formal FLASH alert with tactics, techniques and indicators of compromise (IOCs).
Architecture highlights
ARTEX is a multi-agent platform written in Go: the operator states a goal in natural language and the system splits it into reconnaissance, initial access, foothold, privilege escalation, internal recon and lateral movement steps run by specialized agents. It has no model of its own and is driven by external LLMs. Techniques listed by the FBI include sub-second checks of Azure credential permissions, unsigned SAML assertions, XXE, SSRF, SQL injection, admin passwords hardcoded in SPA JavaScript bundles, web shells, Kerberoasting and DCSync.
Benchmarks and data
This is not a benchmark; the FBI gives incident figures: some intrusions reached privileged access in under three hours, the South Korean finance campaign exposed data on about 67,000 people, the ConnectWise supply chain compromise reached more than 2,250 downstream tenants, and more than 1,600 ARTEX consoles were exposed online in early November, with real deployments likely higher.
Developer quick start
Web and cloud developers should check peripheral systems first: partner portals, mobile backends and small internal apps. Keep cloud admin and publishing endpoints on trusted networks, disable basic publishing auth where not needed, keep secrets out of frontend bundles and CI/CD, and add rate limits and threshold alerts to high-impact APIs. The full FLASH includes IOCs and a MITRE ATT&CK mapping for security teams to work through.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.