OpenAI’s rogue AI agent didn’t stop at hacking Hugging Face

In a new disclosure, OpenAI says its agent used exposed logins to gain access to at least four “publicly available services” in its unhinged quest to solve a test.