Description

Researchers have identified approximately 45 open-source projects that combine large language models (LLMs) with offensive security utilities to automate reconnaissance, vulnerability discovery, and other stages of penetration testing. The findings indicate that the removal of the ARTEX platform did little to contain interest in this technology. Instead, its code and design patterns have continued to circulate through forks and related projects, raising concerns about unauthorized access, automated attacks, and exposed operator interfaces. An analysis by GitGuardian researchers grouped 32 frameworks into five functional categories, with another 13 projects identified as forks or variations. Network-focused platforms automate activities from reconnaissance to post-exploitation, often using Kali Linux environments and tools such as Nmap, Metasploit, and sqlmap. PentAGI coordinates specialized agents, while RedAmon combines attack-surface mapping with human approval controls. Other projects concentrate on web and API testing, command-and-control integration, or extending general-purpose coding agents with offensive capabilities. Academic prototypes form another part of the ecosystem. The research also examined secret-discovery capabilities and internet exposure. Only four reviewed frameworks included a dedicated secret-scanning component, but most could potentially obtain similar functionality through shell access and external tools. Among 41 fingerprinted frameworks, 11 exposed a web interface by default, while 18 did so under certain configurations. On November 7, 2026, researchers identified 1,428 Shodan services associated with the ARTEX family. PentAGI accounted for another 419 services, highlighting the scale of externally reachable deployments.