To protect sensitive data, an organization must implement comprehensive security and governance controls. For AI agents in the Power Platform, this includes configuring Data Loss Prevention (DLP) policies to restrict which data sources and connectors the agents can use. It also involves using role-based access control (RBAC) to manage permissions for who can create, edit, and publish agents, along with enabling auditing and monitoring to track agent activities and potential security incidents. These controls create a secure operational environment.

Open access policies: This would drastically increase the risk of data exposure by removing necessary restrictions.

Disable monitoring: This would prevent the organization from detecting and responding to data leaks or policy violations.

Public APIs only: Relying solely on public APIs is not a security strategy and could increase exposure if not managed correctly.

Microsoft Learn, "Security and governance in Microsoft Copilot Studio". This document outlines the available security features. See the sections: "Data loss prevention (DLP)" and "Authentication".

Microsoft Learn, "Data loss prevention policies". This Power Platform documentation provides details on how to configure DLP to protect organizational data across apps and agents. See the section: "What are DLP policies?".