Zero Data Retention

Zero retention does not equal zero learning—the fact that data is not stored does not mean that patterns and information extracted from it will not influence the model in some form

Zero Data Retention means that after completing a user's request, the service provider does not retain any input or output data on its servers.

Virtual models can specify which providers to use, configure fallback models, and enforce compliance requirements such as zero data retention (ZDR), so that compliance policies are enforced uniformly at the gateway layer rather than relying on each caller.

Zero data retention (ZDR) means the model provider commits not to persistently store request and response content, but to discard it immediately after completing inference in memory only, and is typically used to meet industry compliance or internal security requirements with strict limits on data retention.

Different providers vary in the scope of ZDR support and how it is enabled; some stipulate it through account-level agreements, while others apply it per request via headers or parameters.

If a virtual model requires Zero Data Retention or disallowPromptTraining, these security policies remain enforced even if the request explicitly sets them to false.

Gateway-level settings on a virtual model such as fallback, Zero Data Retention and tags switch along with the virtual model, but provider-specific options take effect per provider, so provider options need to be rechecked when switching providers.

Token Pass enables Zero Data Retention by default, so request data is not retained.

Zero Data Retention and "data not used for training" are two different things: the latter constrains use, while the former constrains storage duration.

Zero Data Retention means the provider does not persistently store prompts and outputs after completing request processing, keeping only the necessary billing and operational metadata.

Zero Data Retention and No Training are the two most common compliance thresholds when enterprises procure AI services, and are usually provided as per-request switches.

The Decisions API supports Zero Data Retention, is HIPAA eligible, and offers US/EU data residency options.

ZDR requires that data be securely wiped after processing in memory, and ensures that no recoverable data fragments remain in temporary caches, log pipelines, or monitoring systems.

Zero data retention is a data processing policy in which the service provider does not retain any input or output data on its servers after completing a user's request.

Zero data retention is a critical threshold for adopting large model services in highly regulated industries such as finance, healthcare, and law, involving regulatory frameworks such as SEC, PCI-DSS, and HIPAA.

ZDR requires that data be securely erased after processing in memory, with no recoverable data fragments remaining in temporary caches, log pipelines, or monitoring systems.

Zero data retention means that after completing a user's request, the service provider does not retain any input or output data on its servers, nor uses it for subsequent training.

There is a tension between ZDR and model improvement, as large model vendors typically rely on user interaction data for RLHF and model fine-tuning.

OpenAI stated it will continue to offer a Zero Data Retention (ZDR) option for frontier models.