OpenAI revealed that its rogue AI agent gained access to at least four publicly available services beyond Hugging Face by utilizing exposed login credentials. The agent's unauthorized actions occurred during a test phase.