800+ Artex AI Hacking Tools Exposed Online
Original: Hundreds of Artex AI hacking tools found publicly exposed to the web
Why This Matters
Exposed AI attack infrastructure lowers the barrier for third-party misuse at scale.
Censys found 800+ publicly accessible instances of Artex, a Chinese-made open-source AI penetration testing tool, with no password protection—linked to attacks on Korean banks and ConnectWise in late 2026.
Security firm Censys has identified more than 800 exposed deployments of Artex, an open-source autonomous penetration testing tool developed in China, accessible from the internet without any authentication. Artex, released publicly in July 2026, operates as a system of AI agents capable of conducting attacks 'from reconnaissance through data exfiltration with little needed human interaction,' according to Microsoft. Because it bypasses access restrictions tied to Anthropic and OpenAI's advanced cybersecurity models, it has become popular among threat actors globally.
Censys principal researcher Silas Cutler noted that not every exposed instance is malicious—many serve authorized testers—but there's no technical reason Artex deployments need to be internet-facing at all. The exposed systems also reveal embedded attack tools, including implants for maintaining persistent access to victim networks.
The stakes are real. CrowdStrike linked Artex to September attacks on South Korean financial institutions, tracing the intrusion to a Chinese-speaking operator who left an open directory. Censys data suggests that same instance had been exposed for months and targeted entities well beyond Korean banks, including payment and gaming firms. In October, Chinese-linked hackers used Artex to breach ConnectWise, compromising thousands of organizations' Microsoft cloud environments and exfiltrating data at high speed.