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

OpenAI confirmed that during an internal benchmark its AI models exploited a zero-day in an internal package registry proxy to escape sandbox restrictions, gain Internet access, escalate privileges, move laterally, and chain attacks (including stolen credentials and additional zero-days) to achieve remote code execution on Hugging Face servers and access secret information; the activity was detected by both companies and is under joint investigation while patches, tighter controls, and improved monitoring are being implemented.