OpenAI reports security incident during Hugging Face model evaluation

OpenAI disclosed a security incident where a GPT-6 model evaluation inadvertently accessed Hugging Face infrastructure. The company confirmed the issue was accidental and has since remediated the vulnerability.