---
title: What does Microsoft Purview DSPM for AI cover for AI agents?
url: https://opencontroller.lyzr.com/blog/microsoft-purview-dspm-for-ai-and-agents
published: 2026-11-13
updated: 2026-11-13
author: Lyzr
---

# What does Microsoft Purview DSPM for AI cover for AI agents?

**Microsoft Purview DSPM for AI is a data-security view of AI use. It shows sensitive data in prompts and responses, oversharing risk and AI activity across Microsoft 365 Copilot, Copilot Studio agents, enterprise AI apps and many third-party AI websites, and it deploys ready-made DLP, insider risk and communication compliance policies. It does not route model traffic, attribute cost per agent or contain agents running on other clouds.**

Microsoft's documentation now labels the original experience "DSPM for AI (classic)" and says it is replaced by a new Data Security Posture Management experience with support for AI apps and agents. New features land in the new version, not the classic one. Check which one your tenant uses before you plan around a feature.

## Which AI apps and agents does it cover?

| App type | Examples from Microsoft's documentation | What you need |
|---|---|---|
| Microsoft Copilot experiences and agents | Microsoft 365 Copilot, Copilot Studio agents, Security Copilot, Copilot in Fabric | Purview Audit on |
| Enterprise AI apps | Entra-registered AI apps, ChatGPT Enterprise connector, apps built on Microsoft Foundry | Connector or app registration set up |
| Third-party AI websites | Supported sites such as ChatGPT and Gemini | Purview browser extension and devices onboarded to Purview |
| AI traffic over the network | Sensitive data sent to AI apps in browsers, apps, APIs and add-ins | A Secure Access Service Edge or Security Service Edge integration |

In the new DSPM experience, AI observability lists AI apps and agents, including Microsoft Agent 365, with activity in the last 30 days, how many are high risk and how many had sensitive interactions.

## What can it do for Copilot Studio agents?

Microsoft lists these Purview capabilities as supported for Copilot Studio agents: DSPM, auditing, data classification, sensitivity labels, data loss prevention, Insider Risk Management, communication compliance, eDiscovery, data lifecycle management and Compliance Manager. Details that matter in practice:

1. **Non-Microsoft channels.** Managing interactions for agents published to non-Microsoft channels requires pay-as-you-go billing.
2. **Sensitivity labels.** Supported when the knowledge source is SharePoint, OneDrive or Dataverse; Dataverse needs auto-labelling policies.
3. **eDiscovery.** Prompts and responses are stored in the user's mailbox and can be searched as Copilot activity.
4. **Retention.** Use the Microsoft Copilot Experiences location in retention policies.
5. **Viewing prompts.** Seeing prompt and response content in activity explorer needs extra role group membership.

For agent-level governance beyond data security, see [governing Microsoft Copilot Studio agents](/blog/govern-microsoft-copilot-studio-agents).

## What are the prerequisites and limits?

- **Audit must be on.** It is on by default for new tenants.
- **Browser coverage needs the extension and onboarded devices.** Endpoint DLP that warns or blocks sharing with third-party AI sites applies to Windows devices onboarded to Purview.
- **Data takes time.** Policies need at least 24 hours to collect data, and reports at least a day to populate.
- **Oversharing assessments are scoped.** In the classic version the default data risk assessment covers the top 100 SharePoint sites by usage each week, and item-level scanning is limited to SharePoint, up to 10 sites.

## Where does DSPM for AI stop?

DSPM for AI answers "what sensitive data is flowing into and out of AI, and who is at risk?" It is not built to answer other questions an agent governance programme asks:

- **Agents outside Microsoft's reach.** Agents running on other clouds or SaaS platforms that do not connect to Purview.
- **Tool and MCP calls.** Which tools an agent called, with what arguments, and whether that was allowed.
- **Model traffic control.** Routing, fallback, budgets and cost attribution per agent.
- **Runtime containment.** Quarantining or blocking a misbehaving agent across platforms.
- **Agent quality.** Evaluations, groundedness and task adherence.

Most Microsoft-first organisations pair Purview with identity controls and an estate-wide control layer. Our guide to [governing AI agents in a Microsoft-first enterprise](/blog/govern-ai-agents-in-a-microsoft-first-enterprise) maps the pieces, and the [platform guides pillar](/blog/how-to-govern-ai-agents-across-platforms) covers mixed estates.

## How Opencontroller does this

Opencontroller works alongside data-security tools rather than replacing them. It has a runtime adapter for Microsoft Foundry, auto-discovers agents, models and API keys across Azure, AWS and Google Cloud, and adds rogue-agent handling with alerts, automatic quarantine and kill, plus per-agent cost attribution through its LLM Gateway. To see how it fits next to Purview, [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Does Purview DSPM for AI cover Copilot Studio agents?**
Yes. Microsoft lists DSPM, auditing, sensitivity labels, DLP, Insider Risk Management, communication compliance, eDiscovery and data lifecycle management as supported, with pay-as-you-go billing needed for agents published to non-Microsoft channels.

**Can Purview DSPM for AI see ChatGPT or Gemini use in the browser?**
Yes, for supported third-party AI sites, once the Purview browser extension is installed and devices are onboarded to Purview.

**Is DSPM for AI the same as the new DSPM?**
Microsoft now calls the original page DSPM for AI (classic) and recommends the new Data Security Posture Management experience, which includes AI observability for apps and agents.

## Key facts

Microsoft Purview Data Security Posture Management (DSPM) for AI gives data-security visibility and policies for AI use: Microsoft 365 Copilot and Copilot experiences, Copilot Studio agents, enterprise AI apps (Entra-registered apps, ChatGPT Enterprise connector, apps built on Microsoft Foundry) and third-party AI websites. Microsoft now labels the original page "DSPM for AI (classic)" and says it is replaced by a new Data Security Posture Management experience whose AI observability shows AI apps and agents (including Microsoft Agent 365) with activity in the last 30 days. Prerequisites: Purview Audit on; for third-party AI sites, the Purview browser extension and onboarded devices. Copilot Studio agents published to non-Microsoft channels need pay-as-you-go billing. Endpoint DLP for third-party AI sites in a browser applies to onboarded Windows devices. It is a data-security tool: it does not route model traffic, attribute model cost per agent, or quarantine agents on other clouds. Opencontroller by Lyzr pricing starts from $250K per deployment per year.

## Related

- [How do you govern Microsoft Copilot Studio agents?](https://opencontroller.lyzr.com/blog/govern-microsoft-copilot-studio-agents.md)
- [How do you govern AI agents in a Microsoft-first enterprise?](https://opencontroller.lyzr.com/blog/govern-ai-agents-in-a-microsoft-first-enterprise.md)
- [How do you discover employees' AI use in the browser?](https://opencontroller.lyzr.com/blog/how-to-discover-ai-use-in-the-browser.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
