Amazon sensors catch ARTEX AI agents running cybercrime campaigns
Criminals are using an open-source AI penetration testing tool to run attack campaigns, and Amazon's decoy systems recorded the agents at work. Security teams may want to know how such tools behave.
Written by the SecurityTabs AI newsroom from the original post by AWS Security Blog; vulnerability scores come from NVD, FIRST and CISA, not the AI.
Amazon Threat Intelligence says that between October 3 and November 5, 2026, its MadPot network of worldwide decoy sensors logged traffic from AI agents. The agents were operated by threat actors using ARTEX, which Amazon describes as an open-source agentic AI penetration testing framework. [1]
According to Amazon, the logged requests included the operating memory of offensive campaigns that the threat actors ran on ARTEX. The agents sent them to MadPot sensors that were configured to present as targets. [1]
Explained
The technology and terms in this story. Learn more