Security researchers have discovered more than 800 publicly accessible instances of a Chinese-developed artificial intelligence tool designed for automated cyberattacks. The software, known as Artex, is an open-source autonomous penetration testing application that probes computer systems for security vulnerabilities without requiring continuous human oversight. Because it was released as free software in July, Artex quickly gained popularity among actors who lack access to top-tier commercial cybersecurity AI models.
Technology firm Microsoft previously described Artex as an integrated network of AI agents capable of managing complete attack lifecycles, executing everything from preliminary system reconnaissance to final data exfiltration. However, a recent investigation by cybersecurity firm Censys revealed that hundreds of active deployments were left connected to the public internet without basic password authentication.
Silas Cutler, principal security researcher at Censys, noted in a company blog post that while not all exposed deployments are operated with malicious intent—as many serve authorized security testers—there is no technical requirement for Artex to be accessible via the web. The misconfigurations leave critical operational components visible to anyone on the internet, including hidden implants used to establish persistent access within victim networks.
Security analysts have tied the misuse of Artex to multiple high-profile cyberattacks. In September, threat intelligence firm CrowdStrike attributed an attack on the South Korean financial industry to a Chinese-speaking threat actor who left an open directory hosting the software. Censys noted that data retrieved from that specific deployment revealed compromised targets extending well beyond Korean banks to include online payment providers and enterprise entities in the gaming sector.
Additionally, hackers leveraged Artex during an October breach of IT service management provider ConnectWise. That incident enabled Chinese-linked operators to spy on thousands of organizations whose files were hosted in cloud environments. Microsoft reported that the automated speed of Artex allowed attackers to compromise cloud setups and steal corporate data before security teams had adequate time to detect or mitigate the intrusion.
Experts emphasize that many of these breaches stem from elementary security failures, such as weak default passwords and the absence of multi-factor authentication. While the developers behind Artex have announced plans to transition the project to a closed-source model reserved for paying customers, dozens of code repositories remain hosted on public sharing platforms, leaving the tool readily accessible for future exploitation.
