Output validation

Output validation is the checking of an AI model's response before it is used, for example confirming it matches a required schema, contains no unsafe content and stays within allowed values.

Output validation sits in the Evaluation & safety part of the Agentik {OS} glossary, which defines the words used to build and run AI agent systems.