Last updated 2026-09-25
What we keep, and what we don't
The gateway never stores prompts or responses. Here is everything we do store.
| Data | Stored? | How long |
|---|---|---|
| Prompts and responses through the API, Claude Code or Codex | Never | — |
| Request metadata (time, model, token counts, cost, latency, status, key) | Yes | 90 days |
| Chat app conversations and uploaded files | Yes, encrypted at rest | Until you delete them. Temporary chats are deleted automatically after 30 days |
| Background job inputs | Only while the job runs | Deleted when it finishes |
| Background job results | Yes | 90 days |
| Account, payments, invoices and ledger | Yes | As long as the law requires for tax records |
| API keys | Only as a keyed hash | Until revoked and the account is deleted |
| Card numbers | Never — held by Stripe | — |
Where
Our gateway, accounts and chat history run on Amazon Web Services in the United States (us-east-1). Claude, GPT and DeepSeek models run in Microsoft Azure AI Foundry (Sweden Central) and Gemini runs on Google Cloud Vertex AI (global endpoint); each supplier processes requests under its own data terms and doesn't use API traffic to train models.
Deleting your data
Delete any chat conversation from the chat app, or your whole account from Settings. Deleting your account revokes every key and removes your personal details; we keep payment and ledger records only where tax law requires.
Model suppliers
Prompts are sent to the supplier that serves the model you choose (Microsoft Azure or Google Cloud). They are processed to produce the response and are not stored by Aethene.