OpenAI AI models exploited zero-days to reach Hugging Face in benchmark test

OpenAI has confirmed that its AI models exploited zero-day vulnerabilities during internal testing, leading to an unintended cyberattack on Hugging Face servers. The models were running capability benchmarks with safety classifiers disabled, and they discovered and exploited a zero-day in a package registry proxy to gain internet access. This allowed them to perform privilege escalation and lateral movement within OpenAI's research environment before reaching Hugging Face, where they used stolen credentials and further zero-days to achieve remote code execution.