Hackers are exploiting ARTEX, PentAGI, and CAIRN, AI penetration testing tools, to automate substantial portions of cyberattacks This article explores hackers exploiting artex. . Google Threat Intelligence Group and Mandiant have detected these frameworks in real-world intrusions since late August 2026, including internal reconnaissance, database exploitation, lateral movement, and data theft preparations.
The findings, published on October 30, reveal attackers moving beyond merely asking chatbots to write malicious code.
They now use coordinated AI agents that interpret objectives, run tools, review errors, and adjust their next actions with minimal human involvement. The creator of ARTEX, who removed the original GitHub repository in October due to widespread abuse, has allowed community forks and commercial variants to continue circulating, enabling attackers to maintain access to the framework. ARTEX exploited application APIs and MySQL logging features to create JSP webshells.
Other observed actions included domain credential harvesting, creation of administrator accounts, enumeration of over 110,000 Active Directory objects, and sending command output through rotating webhook.site endpoints. It uses familiar tools such as Nmap, Metasploit, and SQLmap within isolated Kali Linux environments and supports multiple cloud and local language models. Mandiant recommends restricting administrative interfaces, replacing persistent secrets with temporary credentials, enforcing least privilege, and separating exposed applications from sensitive systems.
Defenders should keep a close eye on unusual authentication methods, rapid API requests, bulk storage access, and unexpected outbound connections.
