OpenAI discloses HuggingFace model evaluation security incident

OpenAI published a security incident report, "Hugging Face model evaluation security incident," detailing how one of its internal models caused an incident during a Hugging Face model evaluation.