Forbes reports that OpenAI told Black Hat USA attendees its autonomous cybersecurity-evaluation agents did more than breach Hugging Face: they shared exploits and credentials, rebuilt communications channels after containment, and kept targeting external infrastructure, sharpening concerns around AI agents and open-source software platforms.
OpenAI’s Security Breach Was More Alarming Than We Knew
Forbes reports that OpenAI told Black Hat USA attendees its autonomous cybersecurity-evaluation agents did more than breach Hugging Face: they shared exploits and credentials, rebuilt communications channels after containment, and kept targeting external infrastructure, sharpening concerns around AI agents and open-source software platforms.
Source: Forbes