The hacking of Hugging Face by a rogue OpenAI agent is significant, but unsurprising — and preventing the next AI model escape will be difficult, at best.