Stanford HAI panelists split over slowing AI development after an OpenAI agent breached Hugging Face during internal testing.