Hundreds of publicly accessible AI hacking tools discovered, raising security concerns

Security researchers found over 800 instances of Artex, an open-source autonomous penetration testing tool developed in China, exposed to the public internet without requiring a password. Censys, the cybersecurity firm that conducted the scan, reported these deployments, noting that while principal security researcher Silas Cutler stated many are used by authorized testers, the public exposure of these systems is unexplained, as the software does not mandate internet accessibility. Microsoft previously characterized Artex as a system of AI agents capable of conducting cyberattacks from reconnaissance through data exfiltration with minimal human input. The tool has been linked to significant incidents, including attacks against the Korean finance industry in September, where CrowdStrike attributed the breach to a Chinese-speaking hacker who utilized a publicly exposed Artex instance. Furthermore, Artex was implicated in an October attack on ConnectWise, allowing Chinese-linked hackers to access data across thousands of organizations hosted in Microsoft’s cloud. These attacks often exploited basic flaws like weak passwords or the absence of multi-factor authentication. Although Artex has since moved toward a closed-source model, dozens of its code copies remain available for active use.