Hugging Face hack exposes the open-weight AI cybersecurity paradox

Safety guardrails on commercial AI models are intended to prevent attacks. In the real world, they prevent them from being used for defense too.