Full definition

Set of rules, validations and controls that limit and guide AI model behavior to prevent dangerous or inappropriate outputs.

Example in a business context

System that automatically blocks responses with PII, offensive language or unverified medical information before showing them to the user.