I’m evaluating the DeepSeek API for an internal tool that may process customer support messages. Before testing it with real data, I’m trying to understand what happens to prompts and outputs after a request is completed.

Does the API offer usable prompt retention controls, or should teams assume submitted content may be retained under the service’s standard policies? I’d appreciate pointers to the current documentation as well as practical advice from anyone who has had to review this for a workplace deployment.