3.1.1. Information Protection and Sensitivity Labels
💡 First Principle: Sensitivity labels are persistent metadata tags attached to content that travel with the file or email wherever it goes — and they can enforce protection actions (encryption, access restrictions, content marking) automatically based on the label. The label is the control; the policy defines what the label does.
Think of a sensitivity label like a classification stamp on a physical document — "Confidential" stamped in red on a folder. But unlike a physical stamp, a digital sensitivity label can also lock the folder: encrypt it so only specific people can open it, prevent printing or forwarding, and watermark every page. The label travels with the content even if it's downloaded, emailed, or shared externally.
Label hierarchy example:
Personal
Public
General
Confidential
Confidential \ All Employees
Confidential \ Specific People
Highly Confidential
Highly Confidential \ All Employees
Highly Confidential \ Legal
What a sensitivity label can do:
- Apply encryption (Azure Rights Management: restricts who can open, edit, copy or print the file, and the protection travels with it — anyone opening it, inside or outside the organization, needs rights granted by the label's encryption settings, and people without those rights can't open it, whether the label was applied by a user or automatically)
- Add content marking (headers, footers, watermarks) — applied by Word, Excel, PowerPoint and, for email, Outlook (in the message body)
- Apply to SharePoint sites, Teams and Microsoft 365 Groups (restrict external sharing, prevent guest access) — this needs the Groups & sites label scope, whereas the file and email scopes label individual items. A container label governs the site's privacy, external sharing and unmanaged-device access; the items inside don't inherit it and aren't encrypted by it (file labels do that)
- Control meeting protections in Teams (prevent recording, prevent transcription)
- Trigger auto-labeling based on content inspection (if a file contains credit card numbers, auto-apply "Confidential"). Service-side auto-labeling labels content in SharePoint, OneDrive and Exchange with no user action; its conditions use sensitive information types and can require a minimum instance count before a label is applied. Run it in simulation first, and tune its detection conditions if it labels too much
Label policies control which labels are available to which users, and whether labeling is mandatory (users must label before saving).
Other label-policy settings: a default label automatically applied to new content created by the targeted users (who can still change it), and require justification to remove or lower a label — a user who removes a label or switches to a lower-priority one must give a reason, which is recorded in the audit log. Mandatory labeling only makes sure content has a label; it doesn't stop downgrades.
Sublabels: when a label has sublabels (e.g., Confidential \ All Employees), users pick a sublabel directly; the parent only groups them in the label picker. Roles: creating and managing labels is a Purview permission (e.g., Information Protection Administrator or Compliance Administrator), not something workload admin roles grant. After a label policy is published, allow up to 24 hours for it to reach users' apps.
⚠️ Exam Trap: Creating a sensitivity label and creating a label policy are two separate steps. A label that hasn't been published via a label policy won't appear to users. The label exists in Purview; the policy makes it available.
Reflection Question: An organization creates 10 sensitivity labels but users report they don't see any labels in Word or Outlook. What's the most likely cause?