---
title: How do you govern Microsoft Copilot Studio agents?
url: https://opencontroller.lyzr.com/blog/govern-microsoft-copilot-studio-agents
published: 2026-11-12
updated: 2026-11-12
author: Lyzr
---

# How do you govern Microsoft Copilot Studio agents?

**Govern Copilot Studio agents through Power Platform: route makers into their own managed environments, lock settings with environment groups, block risky connectors and unauthenticated chat with data policies, cap sharing, and audit prompts and responses in Microsoft Purview. Every new agent also gets an Entra Agent ID. These controls cover Power Platform agents only, so agents on other platforms need a separate inventory.**

Copilot Studio makes it easy for anyone with a licence to build an agent. That is the governance problem. The controls exist, but they live across the Power Platform admin center, Entra and Purview, and some depend on Managed Environments. Here is the order to set them up.

## Which controls govern Copilot Studio agents?

| Control | Where | What it does |
|---|---|---|
| Environment routing | Power Platform admin center | Sends new makers to a personal developer environment instead of the default one; needs Managed Environments |
| Environment groups | Power Platform admin center | Locks rules such as sharing limits and generative AI settings across every managed environment in the group |
| Data policies (DLP) | Power Platform admin center | Blocks connectors, channels, knowledge sources and HTTP endpoints; a violation disables Publish |
| Authentication for agents | Power Platform admin center | Requires authentication per environment or group; preview |
| Sharing limits | Managed Environments | Limits viewers, blocks security-group sharing, blocks sharing with editors |
| Entra Agent ID | Microsoft Entra | Created automatically for new agents |
| Agent inventory | Admin center, API, Azure Resource Graph | Lists Copilot Studio and Agent Builder agents, drafts included |
| Audit | Microsoft Purview | Prompts and responses recorded as CopilotInteraction events in the unified audit log |

## In what order should you set the controls up?

1. **Turn on Managed Environments** for every environment where agents are built. Routing, environment groups and sharing limits depend on it.
2. **Turn on environment routing** so new makers land in their own developer environment instead of the shared default.
3. **Create environment groups** for development, test and production, and publish rules to them. Rules lock the setting in every member environment, and per-environment exceptions are not supported, so design the groups first.
4. **Write data policies.** Since early 2025 these are enforced for all tenants and exemptions are no longer supported. Block "Chat without Microsoft Entra ID authentication" in any environment that touches internal data. Use endpoint filtering on HTTP, SharePoint and public-website knowledge sources. Blocking a connector also blocks tools from connected MCP servers.
5. **Require authentication.** The "Authentication for agents" admin control is preview. When it is on, published agents that do not comply stop responding to users.
6. **Set sharing limits.** Allow individual viewers only, cap viewer counts, and block sharing with editors. Limits apply to agents that require authentication and do not remove access people already have.
7. **Check identities.** New agents get an Entra Agent ID automatically. Agents created before May 2026 keep their app registrations for now and will be migrated later, so list them separately.
8. **Confirm audit.** Copilot Studio interactions are included in Purview Audit (Standard). Check that your retention period covers your [evidence needs](/blog/how-long-to-retain-ai-agent-logs).

## What do data policies actually block?

Data policies treat Copilot Studio features as connectors. The governable list includes unauthenticated chat, HTTP requests, Direct Line, Teams and Microsoft 365, SharePoint, Facebook, WhatsApp, Omnichannel, Application Insights and Skills. If an agent uses a blocked connector, the maker cannot publish it. That makes data policies the strongest preventive control in the stack. See the [Copilot Studio data policy docs](https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-data-loss-prevention) for the current list.

## How does Entra Agent ID change governance?

An agent identity gives each agent a directory object you can own, review and restrict. Microsoft notes that no one in the tenant, admins included, can mint tokens with it. One limit matters today: Conditional Access on the agent identity currently applies only when the agent runs in Teams. Do not assume a Conditional Access policy covers web or other channels until you have tested it. For the delegation side, see [on-behalf-of identity for agents](/blog/ai-agent-identity-on-behalf-of-delegation).

## What do these controls not cover?

Inventory and data policies cover Power Platform agents only. Classic (V1) bots are excluded from inventory, and several inventory fields are still preview. Microsoft Agent 365, generally available since 1 May 2026, extends the registry, but its sync with AWS and Google Cloud registries is in public preview and lifecycle actions are listed as coming soon.

Most enterprises also run agents on Salesforce, SAP, Databricks, AWS and Google Cloud, plus coding agents on developer machines. Finding those is the [shadow agent problem](/blog/how-to-find-shadow-ai-agents), and putting them under one policy set is what an [AI agent control plane](/blog/what-is-an-ai-agent-control-plane) is for. If you want to see how a cross-platform registry would sit beside your Power Platform controls, [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can you stop Copilot Studio agents from chatting without authentication?**
Yes. Block the "Chat without Microsoft Entra ID authentication" connector in a data policy. Makers then cannot publish agents that use it in that environment.

**Do Copilot Studio agents get their own Entra identity?**
New agents get an Entra Agent ID automatically. Agents created before May 2026 keep their app registrations until Microsoft migrates them.

**Where are Copilot Studio prompts and responses logged?**
In the Microsoft Purview unified audit log as CopilotInteraction records. Copilot Studio is included in Audit (Standard).

## Key facts

Copilot Studio governance runs through Power Platform: environment routing and environment groups (need Managed Environments), data policies enforced for all tenants since early 2025 that disable Publish on violation, sharing limits, Entra Agent ID for new agents, and Purview audit (CopilotInteraction records). Conditional Access on agent identity currently applies only in Teams. Microsoft Agent 365 has been GA since 1 May 2026. These controls cover Power Platform agents only.

## Related

- [What is an AI agent control plane?](https://opencontroller.lyzr.com/blog/what-is-an-ai-agent-control-plane.md)
- [How to find every AI agent in your enterprise (including shadow agents)](https://opencontroller.lyzr.com/blog/how-to-find-shadow-ai-agents.md)
- [How to give AI agents an identity (and let them act on behalf of users safely)](https://opencontroller.lyzr.com/blog/ai-agent-identity-on-behalf-of-delegation.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
