9. Your content and our data handling
We process your prompts to deliver them to the model and return the answer. By default we do not store prompt or completion text; we store request metadata (model, tokens, cost, timing, status, routing) for billing, analytics and support. Storing prompts and completions is an organization-level option you control. We do not use your content to train models. Details, retention periods and your controls are in the Privacy Policy and the privacy & data handling documentation. You grant us the limited licence needed to operate the service as described there.