Hackers Use Stolen AI Keys to Find More Keys
AI agents in a cybercrime campaign used stolen API keys to access AI models that then hunted for additional credentials, according to Amazon's MadPot decoy sensors. Amazon assessed with moderate confidence the agents ran on ARTEX, an open-source AI penetration-testing framework. Attacks targeted exposed LLM gateways, default credentials, and known vulnerabilities. Separate ARTEX-linked incidents included theft of 67,000 South Korean financial records, attempted access to over 2,250 Microsoft 365 tenants, and Active Directory reconnaissance. The FBI notes ARTEX's original GitHub repository was removed but forks remain available, with over 1,600 exposed consoles identified by early November.
