Perception guides
Data retention
What Drex and the document service keep of your documents and results, for how long, and what zero retention changes.
This page says what is stored when you use the Perception tools and for how long. The Terms of Service, Privacy Notice and Data Processing Addendum remain the binding legal text.
What is kept
| What | How long |
|---|---|
| A job's request, result and files | Until you explicitly delete the job, its workspace or your account. No free/paid expiry period. |
| Files you upload without naming a workspace | Stored in your account workspace until you delete them or your account. |
| Saved content: workspace files, extraction schemas, workbench drafts and runs | Until you delete the content, its workspace or your account. |
| Classification domains | They can't be deleted in the console or the API, because results already made depend on them. They are kept until your account or organization is closed, or deleted within 72 hours if you ask at support@nace.ai. |
| Copies in backups | Removed within 35 days after the content is removed from active systems. |
Records that hold no document content are kept for their own periods:
| Record | How long |
|---|---|
API logs and usage metadata, such as a job's status, page counts, credits, usage_final and cost | 13 months, then de-identified: who made the request is erased, and only counts, amounts and timestamps remain. |
| Account information | While the account is open, and up to 24 months after it is closed, then de-identified the same way. |
| Billing records | Up to 7 years, as tax and accounting law requires. |
| A deleted file's deletion record: a hash of its name, its size and its page count | Up to 7 years. |
| One-way hashes of the email address and the payment card used at sign-up | As long as needed to stop the same person or card claiming the free signup balance twice. |
| Your acceptance of the Terms of Service and the Privacy Notice: the date and the version | As long as needed as evidence of that acceptance. |
Zero retention
Zero retention is an account setting. An owner or admin turns it on under Settings › Data retention. It requires a workspace explicitly bound to your own S3 or Azure Blob storage. Managed workspaces and standalone uploads are unavailable in this mode. Connect your storage in Settings › Storage (shown where customer workspace storage is available in your deployment), then select the connection when creating a workspace. Managed workspace creation, upload grants/sessions and file replacement return 422 invalid_request in this mode. Where customer storage is not available, Zero retention accounts can't create new workspaces; workspaces already bound to your storage keep working.
Zero retention does not delete customer storage objects, including source files, results and processing artifacts. You control their lifecycle in your storage provider. Enabling the setting does not erase existing workspace files.
For supported document jobs of an account in this mode (the user id is also removed from its already settled jobs):
- The result is returned once in the response that creates the job. The request must wait for it: see Zero retention for
wait_seconds,not_retainedand what happens when a job outlasts the wait. - Nace removes transient request/result content, and the user id from the job record once it settles (the API key id stays). Nace download links close after about 21 minutes. Closing a link does not delete the underlying customer object.
- Status, usage, charges and service identifiers needed for billing and tenant isolation remain. These records are not fully anonymous while they remain linked to an account.
- The console run history shows the result as not retained. Drafts and workbench run content are cleared under the separate Zero retention cleanup.
The contractual cleanup deadlines in the linked legal documents apply to Nace-held copies, not objects in customer-managed storage. Deployment qualification must verify isolated scratch, cache, provider and backup handling before Zero retention is enabled.
While a job runs
A job's document passes through Drex's servers on its way to the document service (Drex doesn't write it to its logs), through the document service's processing infrastructure, and through the model providers that run the tools. Schedule 2 of the Data Processing Addendum lists those providers.
Deleting your data
Delete uploaded files, saved schemas, workbench drafts and runs with the controls in the console and the API, or ask at support@nace.ai. What you ask to delete is removed from active systems within 72 hours, and from backups within 35 days after that. Ordinary content has no automatic expiry, and inactivity does not delete your account.
Closing your account deletes your workspaces and the files and results in them. The records in the second table are kept for their periods.