Guardrails

Definition

Guardrails are the technical limits around an AI system: allowed tools, value ranges, spending and volume caps, forbidden actions and mandatory approvals. They turn an unpredictable model into a system that fails safely and visibly.

Effective guardrails are enforced outside the model — in code, in the integration layer, in database constraints — so a persuasive but wrong model output still cannot cause damage.

They also make operations auditable: every blocked action is logged, and every threshold has an owner who can change it deliberately.

In practice

  • Hard limits on amounts, volumes, recipients and destructive actions
  • Enforced in code, not by asking the model to behave
  • Every block and override is logged for review

Related terms

Back to the glossary