Microsoft Power Platform admin center

Capabilities

• Agent Data Policy Controls

  Govern Agents Data

  Lets admins govern what a Copilot Studio agent can use through Power Platform data policies — connector classification plus allow or block controls for knowledge sources, actions and skills, outbound HTTP, publication channels, and autonomous triggers. Reduces data-exfiltration, oversharing, excessive-permission, and ungoverned-exposure risk.

• Block Unmanaged Customizations

  GovernSecure Agents Infrastructure

  Blocks unmanaged solution imports and direct edits in an environment so agent changes arrive only through approved application lifecycle management. Reduces the risk of untracked tampering and configuration drift in production agents.

• Customer Lockbox

  Govern Data

  Requires an administrator to review and approve or reject the rare requests where a Microsoft engineer needs access to customer data during support, with no access granted until approval. Reduces the risk of unapproved vendor access to the agent's data.

• Customer Managed Key

  Secure Data

  Encrypts Dataverse data at rest with your own key held in Azure Key Vault instead of the default Microsoft-managed key, letting you rotate it or revoke Microsoft's access on demand. Reduces the risk of unauthorized access to agent data at rest and supports encryption sovereignty.

• Data Masking

  Secure Data

  Returns masked values for sensitive Dataverse columns so apps and agents see obfuscated data unless a field security profile explicitly permits the real value. Reduces the risk of sensitive-field exposure in the agent's data store. Currently in preview.

• Dataverse Auditing

  Observe Data Agents

  Logs data changes and user access at the environment, table, and column level, storing the audit trail in Dataverse; production activity logs also flow to Microsoft Purview. Reduces the risk of being unable to reconstruct who changed or accessed the agent's data.

• Dataverse Network and Session Protection

  Secure Data Infrastructure

  Restricts Dataverse access to allowed IP ranges (IP firewall) and binds session cookies to their originating IP address (IP cookie binding), blocking requests and stolen-token replay from unauthorized locations in real time. Reduces the risk of data exfiltration and session hijacking against the agent's data store.

• Environment Groups and Rules

  Govern Agents Infrastructure

  Organizes managed environments into groups and enforces uniform rules across all of them at once — security and sharing, AI features, data retention, and application lifecycle. Reduces configuration drift and inconsistent guardrails as the agent estate grows.

• Environment Routing

  Govern Agents Infrastructure

  Automatically directs new or existing makers into their own managed personal developer environments instead of the shared default environment. Reduces default-environment sprawl and the ungoverned shadow agents that accumulate there.

• Managed Environment Sharing Limits

  Govern Agents Users

  Limits how broadly makers can share canvas apps, flows, and agents — excluding sharing with security groups or capping the number of individuals. Reduces the blast radius of an over-shared agent.

• Managed Environments

  Govern Agents Infrastructure

  Turns on the premium governance boundary that Power Platform environment controls — sharing limits, solution checker enforcement, IP restrictions, and data policies — depend on. Reduces the risk of agents being built in ungoverned environments with no guardrails.

• Power Platform Pipelines

  GovernSecure Agents Infrastructure

  Automates and governs promotion of agent solutions through development, test, and production, enforcing the solution checker, resolving per-environment connection references, and gating each stage with deployment approvals. Reduces the risk of untested changes, development credentials, and unapproved releases reaching production.

• Solution Checker Security Rules

  GovernSecure Agents

  Runs static analysis against the solution checker Security ruleset when a solution is imported, and in block mode stops components with critical security flaws from deploying. Reduces the risk of shipping injection-prone or insecure agent components to production.