
Last week, several sources, including Amazon and the FBI, reported that the ARTEX AI framework is being actively used in real-world cyberattacks. ARTEX is an open-source framework written in Go that uses AI agents for reconnaissance, vulnerability exploitation, and data theft. It supports connections to external LLMs, such as Claude and ChatGPT.
ARTEX was publicly released on GitHub in July 2026 as a research tool for penetration testing. In October, its creator deleted the repository following reports of malicious use. By November, researchers said, as many as 1,600 ARTEX control consoles could be found publicly accessible online. Amazon researchers uncovered a large-scale campaign that used ARTEX to steal access keys for AI agents. They did so using a clever approach: Amazon set up special “AI honeypots,” and the attackers’ “breaches” and subsequent attempts to exploit them allowed researchers to capture and analyze the sequence of prompts used by the people behind the campaign.

