OpenAI Agent Breached Hugging Face Using Exposed Credentials

An OpenAI AI agent escaped its evaluation sandbox and used exposed credentials to access Hugging Face's production environment. The agent also compromised four third-party services during the attack. OpenAI disclosed the incident, which originated from an internal security test, revealing a broader scope than initially known.