# Zero Data Retention (ZDR)

> Zero data retention (ZDR) is an arrangement with an AI model provider under which prompts and outputs are not stored after a request is processed, apart from narrow exceptions the agreement defines. It is usually available to eligible API or enterprise customers by agreement, may not cover every feature, and is separate from a provider's policy on training models with customer data.

## Why it matters for your business

Training and retention are different questions. OpenAI and Anthropic, for example, do not train on API data by default, but a provider may still keep inputs and outputs for a period under its standard terms. For sensitive legal, financial or health data, ZDR shortens the list of places your data lives. Because eligibility and covered features vary, confirm in writing which endpoints, features and models are included. Remember that your own application logs, databases and observability tools still keep whatever you choose to store.

## Example

A law firm builds a contract review assistant on a model provider's API. Because the documents are privileged, the firm requests a zero data retention agreement covering the endpoints it uses and confirms that the features it relies on are included. The firm's own application stores review results in its document management system under its existing retention policy.

Related terms: https://www.lumeor.studio/glossary/data-residency/, https://www.lumeor.studio/glossary/baa-business-associate-agreement/, https://www.lumeor.studio/glossary/ai-governance/, https://www.lumeor.studio/glossary/self-hosting/, https://www.lumeor.studio/glossary/observability/

---

Source: https://www.lumeor.studio/glossary/zero-data-retention/