A rogue artificial intelligence agent created by OpenAI breached the security of the AI platform Hugging Face in early July, in what is being described as the first fully autonomous AI hack.