---
title: "Data Privacy"
description: "Learn how to control data collection and protect sensitive information in AI agent tracing."
url: https://docs.sentry.io/product/agents/privacy/
---

# Data Privacy | Sentry Docs

## [SDK Data Collection](https://docs.sentry.io/product/agents/privacy.md#sdk-data-collection)

Sentry's AI integrations always send metadata, such as model names, token counts, tool names, and execution times. Your SDK's data collection settings control whether they also send content: prompts, system instructions, model responses, tool definitions, tool call arguments, and tool call results.

By default, Sentry agent integrations collect this content. To stop sending it, turn off generative AI inputs and outputs in your SDK's [`dataCollection` option](https://docs.sentry.io/platforms/javascript/guides/nextjs/configuration/options.md#dataCollection).

You can override this per integration with `recordInputs` and `recordOutputs`. For more details, see [Options](https://docs.sentry.io/platforms/javascript/guides/nextjs/configuration/integrations/vercelai.md#options).

## [Server-Side Data Scrubbing](https://docs.sentry.io/product/agents/privacy.md#server-side-data-scrubbing)

Sentry has built-in [Server-Side Data Scrubbing](https://docs.sentry.io/security-legal-pii/scrubbing/server-side-scrubbing.md) that detects and removes common sensitive data from most event data. For AI attributes, Sentry applies a narrower set of rules so conversations stay readable.

If **Use Default Scrubbers** is enabled in your **Security & Privacy** settings, Sentry replaces the following values in AI attributes and keeps the rest of the content intact:

* Email addresses
* Credit card numbers
* IBANs
* US Social Security numbers
* Bearer tokens

This applies to the following attributes:

* `gen_ai.input.messages`
* `gen_ai.output.messages`
* `gen_ai.system_instructions`
* `gen_ai.tool.call.arguments`
* `gen_ai.tool.call.result`
* `gen_ai.tool.definitions`
* `gen_ai.tool.name`