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

# Provider Logging

> How upstream AI providers handle logging and data retention when you use ARouter.

Each AI provider that ARouter routes to has its own data handling policies for logging and retention. This page explains what that means for your data, and how to control it.

## Provider Data Policies

Every upstream provider (OpenAI, Anthropic, Google, DeepSeek, etc.) has its own terms of service governing:

* Whether prompts are used for training
* How long prompts and completions are retained
* Whether responses are logged for abuse detection

ARouter does not alter or override these provider-side policies. When your request reaches a provider, that provider's own policies apply.

### Training on Prompts

By default, most major providers (OpenAI, Anthropic, Google) do **not** train on API traffic under their enterprise or standard API terms. However, policies vary by provider and plan.

You can review each provider's current data policy by checking their official documentation or terms of service page. ARouter surfaces known provider data policies in the [Providers](/en/providers) reference.

## Controlling Provider Selection

If your use case requires strict data handling guarantees, you can control which providers ARouter routes to:

1. **Specify a provider explicitly** — use the `provider/model` format (e.g., `openai/gpt-5.4`) to ensure only that provider handles your request.
2. **Use an ordered candidate list** — set `models` to a prioritized list of provider-specific model IDs, all of which meet your data requirements.

See [Model Routing](/en/model-routing) and [Provider Routing](/en/provider-routing) for configuration examples.

## Data Retention by Provider

Below is a general summary. Always verify with the provider's current terms:

| Provider      | Prompt Retention           | Training Use         |
| ------------- | -------------------------- | -------------------- |
| OpenAI (API)  | 30 days (abuse monitoring) | No (API terms)       |
| Anthropic     | Up to 30 days              | No (API terms)       |
| Google Gemini | Varies by product          | No for Gemini API    |
| DeepSeek      | Varies                     | Check provider terms |

<Note>
  This table reflects general policy summaries and may be outdated. Check each provider's official API terms for the authoritative answer.
</Note>

## Provider Policy Details

### OpenAI

* **Prompt retention**: OpenAI retains API inputs and outputs for up to 30 days for abuse and safety monitoring, then deletes them
* **Training**: OpenAI does not use API data to train models by default
* **Zero Data Retention**: Available for eligible accounts; prompts are not stored at all
* [OpenAI Privacy Policy](https://openai.com/policies/privacy-policy)

### Anthropic

* **Prompt retention**: Anthropic retains prompts and outputs for safety purposes, typically up to 30 days
* **Training**: Anthropic does not train on API data by default
* **Trust & Safety review**: Anthropic may review interactions flagged by automated systems
* [Anthropic Privacy Policy](https://www.anthropic.com/privacy)

### Google Gemini

* **Prompt retention**: Varies by product and API type. Gemini API (AI Studio) does not use prompts for training by default
* **Training**: Not used for model training via the Gemini API by default
* **Data residency**: Google Cloud offers regional data processing for enterprise customers
* [Google AI Privacy](https://ai.google/responsibility/privacy/)

### DeepSeek

* **Policy**: Check [DeepSeek's current privacy policy](https://www.deepseek.com) for up-to-date terms
* ARouter recommends caution with sensitive data when routing to less-established providers

<Note>
  Provider policies change over time. Always verify with the provider's current terms of service and privacy policy for authoritative information before processing sensitive data.
</Note>

## Zero Data Retention (ZDR)

For providers that offer Zero Data Retention, you can enforce ZDR routing at the request level or API key level. See [ZDR](/en/guides/features/zdr) for details.

## Enterprise: In-Region Routing

For enterprise customers with strict data residency requirements, ARouter supports regional routing configurations. Contact [support@arouter.ai](mailto:support@arouter.ai) for details on region-constrained deployments.