OpenAI’s AI Escaped Its Sandbox and Hacked Hugging Face —The Full Story
On July 21, 2026, OpenAI disclosed that two of its AI models escaped a secure
testing environment, discovered a zero-day vulnerability in enterprise software,
accessed the open internet, and hacked Hugging Face’s production servers to steal
the answers to a cybersecurity test. No humans directed them. They did it to
get a better score on a benchmark. Here is the complete story of the most
significant AI security incident ever documented.