Building Secure, Enterprise Ready AI Agents with Purview SDK and Agent Framework

Microsoft announced the public preview of Purview integration with the Agent Framework SDK. This integration aims to make AI agents secure, compliant, and enterprise-ready from their inception. As AI agents move from demos to production, governance becomes a critical, non-negotiable requirement. Traditional applications have security risks at the UI or API layer, but AI agents process sensitive data, collaborate, and act autonomously, introducing new compliance challenges. The Purview SDK embeds Microsoft's enterprise data security and compliance directly into the agent runtime, ensuring governance travels with the agent. This integration offers inline data protection by evaluating prompts and responses against Data Loss Prevention policies in real time. It also provides built-in governance by sending AI interactions to Purview for audit, eDiscovery, and lifecycle management without custom integration. Developers can now ship enterprise-ready agents from the start, with governance feeling native to the platform through Agent Framework middleware. Enforcement operates by evaluating content against configured policies, returning decisions to allow, redact, or block, with governance signals logged for auditing. The enforcement model supports user-to-agent, agent-to-agent, and multi-agent workflows. Adding the Purview SDK to the Agent Framework is a simple process, allowing prompts and responses to be evaluated against enterprise Purview policies. Sensitive data can be automatically blocked, and interactions logged for governance. The Purview SDK is designed for real agent systems, supporting agent-level, workflow-level, and agent-to-agent governance for multi-agent architectures. Developers can explore this integration through Microsoft Learn documentation and GitHub samples. By integrating Purview SDK into the Agent Framework, Microsoft is making governance a default capability, enabling the secure and confident scaling of AI agents.