Conversation content
UnsavedAi does not persist prompts, responses, character cards, conversation history or exposed reasoning in application logs, analytics, databases or backups. We do not use that content for training, licensing or sale. This applies equally to Free and Priority.
Temporary processing and caching
Your request must be processed in memory to generate a response. GPU inference caches may temporarily retain computational state for reuse within your account. That cache is not a conversation archive and is not exported as a dataset. It is evicted or cleared as the model runtime manages memory; we do not promise immediate erasure after each response.
Your client may store your chat history independently. Its storage settings are separate from this service.
Account and operational records
No content retention does not mean no account records. We store account identifiers, authentication records, hashed API keys, plan and billing records, and content-free usage and health metrics needed to operate the service. These metrics can include token counts, timings, request status and GPU utilization. We do not include the text of your conversations in those metrics.
Deleting your account revokes its keys and removes its active account record. Operational records and backups have separate lifecycles; account deletion does not promise instant removal from every backup.
Infrastructure
Requests may run on owned or rented third-party GPU infrastructure. Host operators may have technical access to memory while a request runs. Our no-content-storage policy describes our software; it is not a verified guarantee about every third-party operator. We do not currently claim certified end-to-end zero data retention across all providers.
Questions and data requests
Contact support@puppygenes.com. The service is still in preview; operator and general launch terms remain to be finalized.