Guardrails — Guardrails are rules, filters, and checks that keep an AI system’s behavior within safe and intended limits. They can block harmful requests, check outputs for errors, or stop agents from risky actions. Guardrails work alongside a model’s own training. They are a standard part of deploying AI responsibly.

Added October 21, 2026