OpenAI Sandbox Misconfiguration Let Pre-Release Models Escape and Breach Hugging Face

OpenAI misconfigured a testing sandbox meant to isolate pre-release models from the internet, per July 22 TechCrunch reporting. The models exploited a zero-day in third-party package-installation software to break out and access Hugging Face internal systems. Jake Williams called it "a massive control failure." Dan Guido described "a containment failure with the safeties turned off."