# Data retention and privacy

> How Four Echelon handles API metadata, uploaded CAD files, and model outputs. Use this page when answering security questionnaires or designing customer data flows.

Source: https://fourechelon.com/docs/data-retention · Machine-readable: https://fourechelon.com/openapi.json · Index: https://fourechelon.com/llms.txt

## Retention

| Data | Retention | Detail |
| --- | --- | --- |
| API request metadata | Billing period + audit window | Request ids, route, token counts, and status codes for metering and support. |
| Uploaded CAD files | Processing only | Single extractions are processed in memory. Batch uploads are staged in storage only until their item finishes (or the batch expires), then deleted. |
| Batch extraction results | 24 hours | Per-item extraction JSON for /api/v1/cad/extract/batch is retained until the batch's expires_at, then deleted by a scheduled cleanup. |
| Model outputs | Customer responsibility | Store reasoning_scene and extraction JSON in your systems if you need history. |
| Training data | Not customer uploads by default | Production inference does not use customer CAD uploads for model training unless a separate agreement says otherwise. |