Why anthropomorphizing AI weakens security: human-like language inflates trust, obscures threat models, and leads to dangerous design decisions.