Require zero data retention for model inference

Organization admins can turn on Zero data retention in Settings > Models. Language-model and embedding calls then use eligible Vercel AI Gateway routes with its zero-data-retention requirement. Unsupported routes are refused rather than used as a fallback.

Your saved model configurations stay in place. Models that cannot satisfy the policy become unavailable until you choose a supported route or turn the policy off.

This setting governs model-provider inference, not Runtype's storage of records, conversations, or execution logs. Stop or restart existing executions when changing the policy so they load the new setting.