Trusence Every claim has a source
Last updated 21 November 2026 Search Türkçe
Security

Criminals adopt AI penetration testing frameworks to speed up intrusions

Researchers have tracked 45 such frameworks, and although they are currently easy to detect, copies of the most prominent one persist after its removal from GitHub.

Since late August, criminals have been using a wave of AI-driven penetration testing frameworks to attack companies, banks and government bodies, with the most prominent, Artex, first posted on GitHub in July under a pseudonymous developer. These harnesses do not invent new exploits; they wire external AI models to existing security tools so an operator can issue a plain-language instruction and have agents plan, run and report an intrusion, compressing work that once took days into hours. Reports link the tools to the theft of personal data from South Korean financial institutions, possible cryptocurrency theft and attacks on Microsoft 365 accounts, with one group gaining privileged access in under three hours. The developer removed Artex from GitHub on October 8, but copies persist, and researchers have tracked 45 such frameworks, though they are currently easy to detect.

Why it matters

The tools lower the effort needed for an intrusion: rather than inventing new exploits, they connect external AI models to existing security tools so an operator can issue a plain-language instruction and have the system plan and run the attack. That makes it easier for less skilled actors to reach companies, banks and government bodies, and it shows how quickly misuse can spread once such a guide is published openly. The developer removed Artex from GitHub, but copies remain, so the pattern is likely to persist as the number of frameworks grows.

Sources · 1