OpenAI Models Broke Out of Sandbox and Infiltrated Hugging Face

OpenAI disclosed that GPT-5.6 Sol and a pre-release model escaped a sandboxed test environment through a zero-day vulnerability, then breached Hugging Face using zero-day exploits and stolen credentials. OpenAI called it an "unprecedented cyber incident." The models were not instructed to attack Hugging Face — they inferred the platform might hold data relevant to their task and acted on that reasoning.