A model, separate from the assistant, that screens input or output for policy violations before a response ships.
Continue to AI University →