انتقل إلى المحتوى الرئيسي

Protect Microsoft Copilot Studio agents

Implementation Effort: Medium

User Impact: Low

Overview​

As organizations adopt autonomous and interactive AI agents to perform tasks previously handled by humans, administrators often lose the visibility and control they traditionally have over user network activity. In a Zero Trust model, agents must be treated as identities whose traffic is never implicitly trusted and is continuously verified against security policy—just like users and devices.

Global Secure Access (GSA) for agents extends Microsoft Entra network security controls to Microsoft Copilot Studio agents. You forward agent traffic to Global Secure Access's globally distributed proxy service, where each request is evaluated in real time against your configured policies. This lets you apply web content filtering, threat intelligence filtering, and network file filtering to agent traffic—regulating how agents use knowledge, tools, and actions to reach external resources, just as you govern user traffic. Enforcement uses the tenant-level baseline profile; security profiles linked to Conditional Access policies aren't currently supported for agents. Network controls for agents require Microsoft Entra Internet Access (included in the Microsoft Entra Suite or licensed separately) together with the appropriate Microsoft Agent 365 licensing.

Protecting Copilot Studio agents follows these key steps. See the linked documentation for detailed implementation instructions.

  1. Enable network controls — In the Power Platform admin center, enable Global Secure Access for Agents on a per-environment or per-environment-group basis so agent traffic (HTTP node, custom connectors, MCP servers, custom tools, and supported connectors) is forwarded to Global Secure Access. Ensure the tenant is onboarded to Global Secure Access first, and re-save any existing custom connectors so their traffic routes through the service.
  2. Create and apply security policies — In the Microsoft Entra admin center, create web content filtering, threat intelligence, and network file filtering policies to enforce least-privilege access for agents, then link them to the baseline profile to apply them tenant-wide.
  3. Monitor and maintain — Regularly review Global Secure Access traffic logs for blocked or anomalous agent activity, update filtering policies as connectors and requirements evolve, and test policy changes in a development environment before production.

Reference​