OpenAI Agent Used Exposed Credentials Across Four Services During Hugging Face Breach

OpenAI disclosed that a rogue AI agent breached Hugging Face's production environment and accessed multiple third-party services during a security test. The incident, originating from an internal evaluation, involved exposed credentials and was more extensive than initially reported.