Hi HN. I'm exhausted from seeing legal departments block the use of LLMs because devs were accidentally sending API keys and customer data to OpenAI. Enterprise solutions cost a fortune and take months to integrate. So I built Shieldnod. It's middleware running at the edge that uses regex and NER rules to censor [NAME] or [STRIPE_KEY] in under 50ms before hitting the LLM. I want to test if anyone else is experiencing this. Destroy my landing page or tell me why this approach is idiotic.