Guardrails & Moderation Filter

Ship an AI feature to real users and two things happen fast. Someone pastes something harmful and asks the model to help. And the model itself, now and then, says something you never wanted said. Guardrails are the screening layer wrapped around your feature t

8 lessons, each with runnable code in the browser.

1. What Are Guardrails?
2. The Smallest Filter: A Blocklist
3. Input Moderation with a Classifier
4. Checking the Output Too
5. Defending Against Prompt Injection
6. One Policy, Many Signals
7. Costs, False Positives, and Failing Safe
8. Ship the Safety Layer

Compilearn home