> ## Documentation Index
> Fetch the complete documentation index at: https://docs.youniverse.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Zero Data Retention

> What ZDR means, how Aura enforces it, and where it applies.

<Frame>
  <img src="https://chat.youniverse.ai/__l5e/assets-v1/4804dcfc-154d-4b1d-8e04-ae5df128c772/docs-learn-zdr-vault.jpg" alt="Glass vault where messages pass through and dissolve" />
</Frame>

**Zero Data Retention (ZDR)** means the AI provider that processes your prompt is required to **not store it and not train on it**. The request is processed, the answer is returned, and the content is discarded.

## Why it matters

Most consumer AI products keep your prompts — sometimes for 30 days of abuse monitoring, sometimes indefinitely for training. With ZDR:

* Your prompts and files are **never used to train** any model.
* The provider keeps **no copy** of the prompt or answer after the request completes.
* A breach at a model provider cannot expose conversations that were never stored.

## How Aura enforces it

<Steps>
  <Step title="Every request carries a deny flag">
    Each AI call is sent with a strict **data collection: deny** policy. Providers that would store or train on data are excluded from routing entirely.
  </Step>

  <Step title="Only ZDR endpoints are eligible">
    If a model has no ZDR-compliant endpoint, Aura does not send private traffic to it and uses a compliant version instead.
  </Step>

  <Step title="Receipts per call">
    Aura records which provider handled each request so compliance can be audited, without the provider storing your prompt.
  </Step>
</Steps>

<Frame>
  <img src="https://chat.youniverse.ai/__l5e/assets-v1/3db6233e-3a5a-4796-88aa-184548761cdd/docs-learn-zdr-flow.jpg" alt="Message flows to an AI model and is shredded afterwards" />
</Frame>

## Where ZDR applies

| Surface | ZDR |
| - | - |
| Unlimited chat | Always on |
| Orchestration and Frontier | On by default, shown with the shield badge |
| Work Mode reasoning | On |
| Voice transcription | On |
| AI Studio Transcribe | On |
| Opt-in Fast Mode | Off — clearly badged when you turn it on |
| Image, video and music generation | Off — creative models have no ZDR endpoint, and are labeled |

<Note>
  ZDR covers the **AI provider**. Your chat history is still saved in your Aura account so you can return to it — you control and can delete it.
</Note>

## ZDR vs. "we don't train on your data"

Many companies promise not to train on your data but still **retain** it for weeks. ZDR is stricter: no training **and** no retention at the model provider.

<AccordionGroup>
  <Accordion title="Does ZDR make answers worse?">
    No. The same models answer; only storage and training are switched off.
  </Accordion>

  <Accordion title="Why is the shield hidden sometimes?">
    The shield only shows when the current route is actually ZDR. Turn on Fast Mode or generate media and it disappears, so you always know.
  </Accordion>

  <Accordion title="Does prompt caching break ZDR?">
    No. Caches are short-lived, held in memory, scoped to your conversation, and expire automatically. See [Prompt caching](/learn/prompt-caching).
  </Accordion>
</AccordionGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.