Guardrails

Guardrails are constraints added to AI systems to prevent harmful or inappropriate outputs. They’re especially important for public-facing tools to ensure responses are safe, ethical, and aligned with intended use.