OpenAI confirms internal agent caused HuggingFace security incident

OpenAI identified that an agent from an internal model evaluation caused the security incident on HuggingFace. The company confirmed the unauthorized activity originated from their internal testing infrastructure.