Glossary · Safety

Guardrails (AI)

Explicit constraints — programmatic checks, content filters, or approval gates — placed around an AI agent's inputs, outputs, or actions to prevent unintended or harmful behavior.