In our previous blog we set the ground why data governance matters, how Purview can provide a more holistic view to your data estate and explained the role of different type of metadata in the process.
In general, metadata management in Fabric has been fragmented, and continues to be that way in the Fabric and Power BI user interfaces. However, the latest Fabric features and AI tools do help with large scale metadata management:
OneLake Catalog Search API (Generally Available)
Fabric now offers a GA level search API for the OneLake Catalog, giving users deeper visibility into datasets, domains, and access patterns. This reduces governance blind spots and enables enterprise-level reporting and oversight. A single search request can locate matching items across your accessible estate based on catalog metadata and the user’s permissions.
Workspace Tags (Generally Available)
Stewards can tag workspaces with domain, owner, or classification metadata—improving catalog capabilities and aligning assets with business domains.
Bulk Import & Export of Item Definitions (Preview)
New APIs enable stewards to manage metadata definitions at scale. This reduces manual curation effort and supports updates to business terms, data assets, and glossary definitions in reports.
Remote MCP Server for AI Agents
Engineers can run AI agents inside Fabric with governed access to metadata—ensuring copilots follow security and governance rules automatically.
For metadata consumers, like report users, Purview Unified Catalog provides a data lineage view, at a more granular level compared to Fabric. Data engineers, however, need a more efficient way to check the impact of their work. For example, answering a question like "Which reports need to be modified if these column values change?", might require a lot of clicks to solve. Purview helps, but enabling Fabric Data agents can take the user experience to the next level. Whether this is reasonable use of AI resources can be questioned, since there are ETL tools available that generate creates the lineage automatically and provides a proper user interface to it. Building data pipelines in Fabric with dbt would solve the problem effectively.
Compliance and Security teams also face the challenge of tracking where the data flows and how it is shared. The risks related to AI apps and agents are surely discussed and mitigated actively in every organization at the moment. We blogged earlier about Microsoft Purview Data Security Posture Management (DSPM), a solution that focuses on data protection in a landscape with traditional and AI applications. The capabilities have now been extended with direct actions against risk behavior and enhanced monitoring of the data usage in Fabric.
Purview Data Loss Prevention (DLP) Policies for Fabric (Preview)
The update extends DLP capabilities to structured data stored in OneLake, enabling organizations to automatically enforce restrictions and mitigate insider risk or accidental data leak.
Quick Policy for Data Theft, Purview Insider Risk Management (General Availability)
A streamlined “quick policy” for data theft detection and prevention is now generally available—supporting rapid response to insider threats without manual rule creation. In addition, compliance and security team can now access more granular risk reporting tied to Fabric usage, supporting compliance requirements and improving audit readiness.
Purview DSPM for AI – Fabric Copilots and Data Agents (Preview)
Fabric integrates with Purview’s Data Security Posture Management (DSPM) for AI, enabling leaders to monitor how AI agents and copilots use data—and to ensure usage complies with policy.
Modern data governance does require automation, but that is not the starting point. What we at Zure work on with our customers, are first the prerequisites: strategies, policies and roles of people and the value it all drives. While AI agents and copilots become embedded in analytics workflows, the governance model must ensure that innovation does not come at the expense of security or compliance.