Hundreds of Artex AI Hacking Tools Left Open on the Web
Share: https://www.ellixorlabs.com/blog/artex-ai-hacking-tools-exposed-web (copy URL for iMessage or social)
Published (display in IST)
Updated (IST)

Censys found 1,465 exposed Artex AI pentest consoles on the open web, with about 815 distinct deployments. The tool has already been used in Korean finance attacks and the ConnectWise supply-chain breach.
Photo: Sergei Starostin / Pexels, free licence. Source: https://www.pexels.com/photo/network-servers-on-an-enclosure-6466141/
Censys researchers have found 1,465 publicly exposed Artex (ARTEX) AI hacking consoles on the open web as of 4 November 2026, with roughly 815 distinct-looking deployments. The autonomous pentest framework has already been used in real attacks against Korean finance firms and in Storm-0940's ConnectWise supply-chain compromise. For founders, the direct answer is clear: agentic offensive tools are now commodity software, and misconfigured copies are sitting unlocked for anyone to find.
Hundreds of Artex consoles left unlocked
On 5 November 2026, Censys principal security researcher Silas Cutler reported that Censys had observed 1,465 hosts running an exposed ARTEX console. About 650 of those hosts sat in two blocks of identically configured hosting, which left roughly 815 distinct-looking deployments. TechCrunch's Zack Whittaker, writing on 6 November 2026, framed the finding as more than 800 distinct deployments accessible without a password. Of the 1,465 hosts, 155 (10.6%) also ran other Censys-detected threats such as Cobalt Strike, Vshell, Adaptix C2 and other AI attack frameworks. Censys also saw 32 open directories on ARTEX hosts exposing implants, stagers and post-exploitation kits. ARTEX was open-sourced on 26 July 2026 on GitHub as Autumn-27/ARTEX under AGPL-3.0. After CrowdStrike linked it to attacks on South Korean financial institutions, the author said the project would go closed-source; the repository has since been taken down, though copies still circulate. Microsoft has described ARTEX as able to run from reconnaissance through data exfiltration with little human interaction. In a 21 October 2026 Security Blog post, Microsoft said China-state aligned actor Storm-0940 used ARTEX from as early as September 2026, obtained privileged cloud access in under three hours in multiple instances, and during the ConnectWise compromise attempted sign-ins to more than 2,250 tenants, with password spraying peaking at 2,900 failures per minute. Primary reporting: Censys at https://censys.com/blog/artex-agentic-asbestos/ , TechCrunch at https://techcrunch.com/2026/11/06/hundreds-of-artex-ai-hacking-tools-found-publicly-exposed-to-the-web/ , and Microsoft at https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/ .
Why it matters for founders
This is not a distant research curiosity. Artex turns recon, password spraying and OAuth abuse into machine-speed workflows that leave defenders little time to respond. Many of the linked attacks exploited weak or known passwords and a lack of multi-factor authentication, then abused trusted third-party applications. If your team uses cloud apps with broad Graph or admin scopes, or relies on IT vendors with multi-tenant access, you sit in the same blast radius as the ConnectWise victims. For related AI-assisted attack patterns hitting Indian fintech, see our note on CERT-In Crafty Falcon at /blog/cert-in-crafty-falcon-ai-attacks-indian-fintech . For core web stack risk, update now if you still run the WordPress flaw covered at /blog/wordpress-core-flaw-cve-2026-119977-exploited-update-now . Practical steps: enforce MFA everywhere, ban weak passwords, apply least privilege to OAuth apps and vendor integrations, review which suppliers can sign into your tenants, and never expose offensive tooling or open directories to the public internet. If you are building AI agents into your own stack, our AI solutions page at /services/ai-solutions covers security-minded designs for founders.
What to watch next
Watch whether more ARTEX clones and rebrands keep circulating after the original GitHub repo was taken down. Watch for fresh Storm-0940 or CrowdStrike follow-ups naming additional victims. Watch how many of the 1,465 exposed consoles operators actually lock down, and whether open directories continue to leak implants. The wider pattern is that agentic offensive frameworks are spreading faster than teams can inventory them.
Frequently asked questions
What is Artex (ARTEX)? Artex is an open-source, LLM-driven autonomous penetration testing framework published on GitHub as Autumn-27/ARTEX on 26 July 2026 under AGPL-3.0. Microsoft describes it as AI agents that can run from reconnaissance through data exfiltration with little human interaction. How many Artex instances were found exposed? As of 4 November 2026, Censys observed 1,465 hosts with an exposed ARTEX console. About 650 sat in two identical hosting blocks, leaving roughly 815 distinct-looking deployments. TechCrunch framed that as more than 800 distinct deployments accessible without a password. What should founders do about exposed AI hacking tools? Enforce multi-factor authentication, ban weak and reused passwords, apply least privilege to OAuth apps and vendor integrations, review supply-chain software access, and never expose offensive security tooling to the public internet.
If you need help hardening identity, OAuth scopes or AI-assisted workflows before the next agentic campaign, talk to Ellixor Labs at https://www.ellixorlabs.com/contact or explore https://www.ellixorlabs.com/services/ai-solutions .
FAQ
What is Artex (ARTEX)?
Artex is an open-source, LLM-driven autonomous penetration testing framework published on GitHub as Autumn-27/ARTEX on 26 July 2026 under AGPL-3.0. Microsoft describes it as AI agents that can run from reconnaissance through data exfiltration with little human interaction.
How many Artex instances were found exposed?
As of 4 November 2026, Censys observed 1,465 hosts with an exposed ARTEX console. About 650 sat in two identical hosting blocks, leaving roughly 815 distinct-looking deployments. TechCrunch framed that as more than 800 distinct deployments accessible without a password.
What should founders do about exposed AI hacking tools?
Enforce multi-factor authentication, ban weak and reused passwords, apply least privilege to OAuth apps and vendor integrations, review supply-chain software access, and never expose offensive security tooling to the public internet. Treat password-spray and Graph anomalies as high priority.
Talk to Ellixor Labs about this topic.
Source: Censys