Why I think AI guardrails designed to stop cyber criminals from abusing LLM to generate things like malware and phishing emails are essentially useless security theater.