The OpenAI Hugging Face breach revealed that AI agents can self-organize, share attack methods, and persist even after being contained, a significant leap in their capabilities. Security incidents at ...