---
title: How do Power Platform data policies control Copilot Studio agents?
url: https://opencontroller.lyzr.com/blog/copilot-studio-data-policies-for-agents
published: 2026-11-13
updated: 2026-11-13
author: Lyzr
---

# How do Power Platform data policies control Copilot Studio agents?

**Power Platform data policies decide which connectors, channels and authentication options a Copilot Studio agent may use. Connectors go into Business, Non-business or Blocked groups, and an agent cannot mix groups. Enforcement is live in every tenant, and a violation stops the agent from being published.**

Data policies used to be called DLP policies, and many admins still use that name. They are the main admin control over what a Copilot Studio maker can build. This guide explains how they apply to agents and gives a starting policy.

## How do data policies apply to Copilot Studio agents?

A data policy sorts every connector into one of three groups:

- **Business.** Connectors that may handle company data.
- **Non-business.** Connectors for other data. An agent cannot use Business and Non-business connectors together.
- **Blocked.** Connectors that cannot be used at all.

Since early 2025, enforcement applies to all tenants (Microsoft message center MC973179). Exemptions for individual agents are no longer supported, so agents that were exempt before are now in scope.

Enforcement is real time. Blocked tools appear disabled in the maker interface, with a tooltip saying why. If an agent breaks a policy, the **Publish** button is unavailable and an error banner appears. On the **Channels** page you can download a file listing each violation.

## Which Copilot Studio connectors appear in data policies?

| What you want to control | Connector name in the admin center |
|---|---|
| Agents published without sign-in | Chat without Microsoft Entra ID authentication in Copilot Studio |
| HTTP requests from agents | HTTP (supports endpoint filtering) |
| Azure Bot Service skills | Skills with Copilot Studio |
| Teams and Microsoft 365 channel | Microsoft Teams + Microsoft 365 Channel in Copilot Studio |
| Websites and mobile apps | Direct Line channels in Copilot Studio |
| Social and other channels | Facebook, SharePoint and WhatsApp channel connectors |
| Dynamics 365 Customer Service | Omnichannel in Copilot Studio |
| Event triggers and automated evaluations | Microsoft Copilot Studio |
| Telemetry export | Application Insights in Copilot Studio |
| Tools | Prebuilt and custom Power Platform connectors |

Connectors added after 2019, such as unauthenticated chat and Direct Line, often land in the default Non-business group. Check how each is classified in your tenant rather than assuming.

Knowledge sources work differently. For SharePoint sites and public websites used as knowledge, you allow or deny specific endpoints through **Configure connector > Connector endpoints**.

## What is a good starting data policy for agents?

Adapt this to your risk appetite:

1. **Block** Chat without Microsoft Entra ID authentication in production environments. Every agent must know who the user is.
2. **Block** Facebook and WhatsApp channels unless a customer-service team has an approved use case.
3. **Endpoint-filter HTTP** to an allow-list of internal APIs. Unfiltered HTTP lets an agent call anything.
4. **Endpoint-filter knowledge sources** so agents can only ground on approved SharePoint sites and named public domains.
5. **Allow** Teams and Microsoft 365 channel and the connectors for approved business systems in the Business group.
6. **Keep a looser policy** for a sandbox environment, so makers can experiment without touching production data.
7. **Add a contact email and a "Learn more" link** to error messages with the `New-PowerAppDlpErrorSettings` PowerShell cmdlet, so makers know who to ask.

Pair the policy with [limits on which tools each agent can call](/blog/limit-which-tools-each-ai-agent-can-call) and an owner for every agent.

## What do data policies not cover?

- **Agents outside Power Platform.** Foundry agents, Agentforce, Bedrock and code-first agents sit elsewhere.
- **Who owns each agent** and whether its owner still works for you.
- **What the agent says.** Data policies control connections, not answer quality or groundedness.
- **Behaviour after publishing.** A compliant agent can still be misused or drift.

For the full Microsoft picture, see [governing Copilot Studio agents](/blog/govern-microsoft-copilot-studio-agents) and [governing AI agents in a Microsoft-first enterprise](/blog/govern-ai-agents-in-a-microsoft-first-enterprise). For other platforms, read our [guide to governing agents across platforms](/blog/how-to-govern-ai-agents-across-platforms).

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud, has a runtime adapter for Microsoft Foundry, and keeps one governed registry of agents, versions, tools and MCP servers. Posture flags risky permissions and configurations. Policy as code applies across platforms, with built-in mappings to OWASP, MITRE ATLAS, NIST and ISO 42001. Keep data policies as your Power Platform control; add Opencontroller as the [control plane](/blog/what-is-an-ai-agent-control-plane) across the estate. [Book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can an individual Copilot Studio agent be exempted from data policies?**
No. Agent exemptions are no longer supported.

**What happens when an agent breaks a data policy?**
Publish is disabled, an error banner appears and you can download a list of violations.

**Can data policies restrict knowledge sources?**
Yes, for SharePoint and public websites, through connector endpoint filtering.

## Key facts

Power Platform data policies (formerly DLP policies) govern Copilot Studio agents by sorting connectors into Business, Non-business and Blocked groups; connectors in different groups cannot share data in one agent. Copilot Studio-specific connectors include Chat without Microsoft Entra ID authentication in Copilot Studio, HTTP (with endpoint filtering), Skills with Copilot Studio, Direct Line channels, Microsoft Teams + Microsoft 365 Channel, Facebook, SharePoint and WhatsApp channels, Omnichannel, Application Insights and Microsoft Copilot Studio (event triggers). Since early 2025 enforcement applies to all tenants (MC973179) and agent exemptions are no longer supported. Violations disable Publish and show an error with a downloadable list. Knowledge source endpoints for SharePoint and public websites are filtered through Connector endpoints. Opencontroller by Lyzr discovers agents across Azure and other clouds and applies policy across platforms; 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 limit which tools each AI agent can call?](https://opencontroller.lyzr.com/blog/limit-which-tools-each-ai-agent-can-call.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
