When I first heard the news this summer that a group of artificial intelligence agents created by OpenAI had hacked into Hugging Face, an A.I. infrastructure company, I filed it in the “Bad but Probably Not Catastrophic A.I. Safety Incidents” subfolder of…