Safety
Guardrails
Checks and rules around an AI system that block unsafe inputs or outputs and keep it within its intended scope.
Guardrails can be moderation models, validation of structured output, allow-lists of actions, or human approval before an agent does something irreversible.
Frequently asked questions
What does Guardrails mean?
Checks and rules around an AI system that block unsafe inputs or outputs and keep it within its intended scope. Guardrails can be moderation models, validation of structured output, allow-lists of actions, or human approval before an agent does something irreversible.