> ## Documentation Index > Fetch the complete documentation index at: https://docs.yingtu.ai/llms.txt > Use this file to discover all available pages before exploring further. # API data handling and privacy > Understand how API content reaches upstream providers, whether YingTu uses it for training, what service metadata is retained, and which retention periods are not promised. The current [YingTu privacy policy](https://api.yingtu.ai/privacy-policy) states that, under normal service configuration, YingTu does **not** use API inputs or outputs for model training, advertising, marketing, or user profiling. Requests still have to be transmitted to the selected upstream AI provider and the infrastructure required to process them. The upstream provider's own terms, abuse-monitoring practices, retention rules, and training controls apply separately. YingTu's policy must not be read as a promise about Google or another upstream provider. ## What the service processes | Data category | Why it may be processed | | ----------------------- | --------------------------------------------------------------------------------------------------- | | Prompt and source image | Route and complete the requested generation or edit | | Generated response | Return a synchronous result or make an asynchronous result retrievable | | Service metadata | Account, key identifier, time, model, operation, usage, status, latency, route, and error diagnosis | | Billing records | Reserve, settle, or return API quota and support account reconciliation | | Support material | Investigate the specific issue a user chooses to report | Diagnostic or security processing may temporarily include limited request parameters or upstream response fragments. Avoid submitting data that the task does not require. ## Synchronous and asynchronous data A synchronous request carries its content while the upstream model is processing it and returns the result on the same connection. An asynchronous request must persist enough task data to execute after the submitting connection ends. After completion, the result is associated with the task so the authenticated account can retrieve it. Move completed media to storage you control instead of treating the task API as permanent object storage. ## Retention periods are not fixed in this documentation The current public policy does not promise one fixed number of days for prompts, source images, generated results, service metadata, diagnostic logs, or asynchronous task records. Retention depends on operational, billing, security, dispute, backup, and legal needs. A task execution timeout is not the same thing as a record or result retention period. If a procurement or compliance decision requires a contractual retention schedule, data residency, deletion deadline, or processor list, contact [API support](/en/resources/technical-support) before sending production data. ## Reduce the data you expose * Remove secrets, credentials, personal data, and confidential text that the generation does not need. * Keep API keys on the server and separate development from production credentials. * Do not write full Base64 content or Authorization headers to ordinary logs. * Store the returned image in your own controlled system and keep only identifiers needed for support. * Review the selected upstream provider's current policies for the model and account type you use. The complete legal terms remain the privacy policy and [user agreement](https://api.yingtu.ai/user-agreement). Keep credentials out of browser-delivered code, logs, and screenshots. Persist your own task ID and copy completed output to controlled storage.