Data handling
How we process request content and retain account and usage data.
How inference content is handled
Prompts and responses are processed to answer your requests. They are not archived as conversation history or used for model training or advertising. Short-lived caches may temporarily hold request content and expire automatically.
- The usage dashboard does not contain a prompt or response archive.
- Your application is responsible for any conversation history it needs.
- The inference policy applies independently of whether you use streaming.
Temporary processing and caching
Responding to a request requires temporary processing of its content. Short-lived caches support that processing and expire automatically. This is distinct from keeping a permanent conversation archive.
Send only data that your application is permitted to process. Do not include API keys, passwords or unrelated sensitive information in prompts. Your own SDK, application logs and integrations may have separate retention settings.
Training has a separate policy
Account and usage records
| Record | Purpose |
|---|---|
| Account and sign-in information | Provide access, maintain sessions and secure your account. |
| API-key identifiers and hashes | Authenticate requests without displaying the key secret again. |
| Request ID, model, status and timing | Show request activity, investigate errors and support the service. |
| Token counts, applicable rates and recorded cost | Show usage and reconcile charges. |
| Support and security records | Respond to requests, prevent abuse and meet applicable obligations. |
For access, correction or deletion requests, email support@arnict.com. Some records may need to be retained for security, financial or legal purposes, as described in the Privacy Policy.
- Privacy PolicyRead the account-data and retention policy.
- Trust and securityReview the product’s data-handling commitments.