1.4 — Sensitivity Labels and Endorsement
DP-600 Domain 1 · Skill 1.4
What You Need to Know
Governance on DP-600 focuses on semantic models and lakehouses: how to label them so that sensitivity follows the data downstream, and how to endorse them so that people find and trust the right ones.
Sensitivity labels
| Fact | Detail |
|---|---|
| Source | Microsoft Purview Information Protection labels, published to users by label policies. A Fabric admin turns on labeling in the tenant settings |
| Apply | In the service: the sensitivity control in the item header or in the item's settings. In Power BI Desktop: Home > Sensitivity |
| Permission | You need edit permission on the item (plus a Pro or PPU license) and must be allowed to use that label |
| Downstream inheritance | A label on a lakehouse or semantic model can be inherited by the items built from it: model, reports, dashboards and exports |
| Inheritance from data sources | Semantic models can inherit labels from supported labeled sources (such as Azure SQL) when they refresh |
| Export protection | Protected labels travel with exported files (Excel, PowerPoint, PDF), and their encryption keeps protecting the data outside Fabric |
| Policies | A default label for new content, mandatory labeling before publishing, and DLP policies that detect sensitive data in semantic models |
A label classifies and protects content. It doesn't grant anyone access, and it doesn't replace RLS or item permissions.
Endorsement
| Badge | Meaning | Who applies it | Items |
|---|---|---|---|
| Promoted | The owner recommends the item for reuse | Anyone with write permission on the item | Every item except dashboards |
| Certified | It meets the organization's quality standards | Only reviewers a Fabric admin names (can be delegated per domain) | Every item except dashboards |
| Master data | It's the authoritative source for core data (customers, products, cost centers) | Only users a Fabric admin names | Items that contain data: lakehouses, warehouses, semantic models and similar |
- Endorsed items appear first in the OneLake catalog and in "get data" pickers, which steers authors toward shared, trusted models instead of personal copies.
- A typical pattern is to certify the enterprise semantic model, give report authors Build permission on it, and let them create live-connected reports.
Key Concept: Labels say how sensitive something is and follow the data downstream and into exports. Endorsement says how trustworthy it is and helps people find it. Neither one grants access.
Exam Traps
- Expecting a sensitivity label to restrict who can open an item in Fabric
- Trying to certify or promote a dashboard
- Letting a model's owner certify their own model when certification is limited to named reviewers
- Marking a report as master data (only data items qualify)
Check your understanding
Practice Scenario
A semantic model has the protected sensitivity label Highly Confidential. A user with view access exports a table visual from a report on the model to Excel. What happens to the exported file?
Log in to save your progress.
Sources
- Apply sensitivity labels to Fabric items, Microsoft Learn
- Sensitivity label inheritance from data sources, Microsoft Learn
- Sensitivity label downstream inheritance, Microsoft Learn
- Endorsement overview, Microsoft Learn