Domain 1 of 3 · Chapter 2 of 3

Implement and manage sensitivity labels in Microsoft Purview

How a sensitivity label reaches and protects content

A spreadsheet named Q3-forecast.xlsx shows Confidential on the Excel status bar. That word is a sensitivity label: a classification your organization defines, stored in clear text in the file's metadata, so it stays with the file wherever it is saved and other apps and services can read it (Learn about sensitivity labels[1]). The label identification, unique to your organization, is what apps use to apply and enforce the settings you configured on that label (Learn about sensitivity labels[1]).

Under Implement and manage sensitivity labels in Microsoft Purview, the Microsoft SC-401 study guide skills measured[2] runs from 'Implement roles and permissions for administering sensitivity labels' to 'Apply sensitivity labels by using Microsoft Defender for Cloud Apps'. Automatic labeling can detect content with sensitive information types (SITs) or trainable classifiers (Automatically apply a sensitivity label[3]), and the Microsoft SC-401 study guide skills measured lists 'Create and manage trainable classifiers' under Implement and manage data classification and 'Plan and implement the Microsoft Purview Information Protection client' under Implement information protection for Windows, file shares, and Exchange.

The four steps from definition to enforcement

Microsoft describes the deployment as a flow: administrators create and publish labels, users apply them to emails and documents, and Office or other apps and services enforce protection based on the applied labels (Get started with sensitivity labels[4]). A label can also be applied without the user choosing it, as a default label or through automatic labeling (Learn about sensitivity labels[1]). Until a label is added to a publishing policy, which Microsoft Learn calls a label policy, it isn't available to select in apps or for services (Create and publish sensitivity labels[5]). The figure below traces the four steps that the rest of this topic follows: define, publish, apply, enforce.

One label per item, and two label families

Each item that supports sensitivity labels can have a single sensitivity label from your organization, and documents and emails can carry a retention label at the same time (Learn about sensitivity labels[1]). The two families share the word label but not their purpose or their publishing model: retention labels are published to locations such as all Exchange mailboxes, while sensitivity labels are published to users or groups (Learn about sensitivity labels[1]). The Microsoft SC-401 study guide skills measured[2] lists 'Plan for information retention and disposition by using retention labels' under Implement and manage retention.

A sensitivity label protects one of two kinds of target. An item label protects documents, emails and meeting invites, for example by encrypting or marking them; a container label protects a site or group through settings such as privacy and external sharing, and doesn't label the items stored inside (Learn about sensitivity labels[1]).

The takeaway: a label is a portable, readable identity on content. Publishing makes it selectable, applying puts it on content, and apps and services enforce what it configures.

1. Define the labelscopes and settings2. Publish the labelin a publishing policy3. Apply the labeluser, default or automatic4. Enforce its settingsapps and services
Sensitivity label deployment flow: define, publish, apply, enforce (after Microsoft's sensitivity label flow in Get started with sensitivity labels)

Roles and administrative units for label administration

Delegating label work starts from one distinction: permission to configure labels and permission to apply them are different things. Microsoft recommends roles with the fewest permissions and minimizing the number of users with the Global Administrator role (Get started with sensitivity labels[4]).

Role groups and roles for label administration

Members of the compliance team who create and manage labels need permissions to the Microsoft Purview portal, and Microsoft lists these role groups: Information Protection, Information Protection Admins, Information Protection Analysts, Information Protection Investigators and Information Protection Readers (Get started with sensitivity labels[4]). As an alternative to the default role groups, you can create a role group and add the Sensitivity Label Administrator role to it, or use Sensitivity Label Reader for read-only access (Get started with sensitivity labels[4]). Adding users to the Compliance Data Administrator, Compliance Administrator or Security Administrator role group also works (Get started with sensitivity labels[4]).

These permissions are required only to create and configure labels and their publishing policies, not to apply the labels in apps or services (Get started with sensitivity labels[4]). A user who picks Confidential in Word needs the label published to them, not an administrative role.

Content-preview permissions and label-administration permissions are separate. Viewing file or email contents in the source view of an auto-labeling simulation (an auto-labeling policy applies labels to content instead of offering them to people; its simulation is described in Simulate, turn on, and monitor auto-labeling policies below) needs the Data Classification Content Viewer role, which is included in the Content Explorer Content Viewer role group and in the Information Protection and Information Protection Investigators role groups; Global Administrators don't have it by default (Automatically apply a sensitivity label[3]). Seeing content is not configuration authority: when auto-labeling policies are missing from the portal, Microsoft's check is that your role includes Compliance Administrator or Compliance Data Administrator (Automatically apply a sensitivity label[3]).

Administrative units bound administrators and policies

Administrative units are configured in Microsoft Entra ID, need Microsoft Entra ID P1 or P2 licensing, and in Microsoft Purview can also have SharePoint sites associated with them (Administrative units in Microsoft Purview[6]). Sensitivity labels use them in two ways (Get started with sensitivity labels[4]):

  • Restricting administrators. Edit a role group used for Information Protection, select individual members, and use Assign admin units; those administrators are then restricted to managing just the users in the selected units.
  • Scoping policies. When you create or edit a publishing policy or an auto-labeling policy and select administrative units, only the users in those units are eligible for Exchange and OneDrive, and only the sites in those units are eligible for SharePoint.

If your account has been assigned administrative units, you must select one or more of them when you create a publishing policy; otherwise you can keep the default of Full directory (Create and publish sensitivity labels[5]). For SharePoint, administrators who select administrative units can't include or exclude individual sites, and the policy applies to all sites associated with the unit (Administrative units in Microsoft Purview[6]).

The accuracy of unit membership is a Microsoft Entra ID dependency, and it is also what makes units convenient (Get started with sensitivity labels[4]). In Microsoft's example, an administrator assigned the France, Germany and Spain units selects France and keeps the default of all users and groups, which scopes the policy to all users in France and covers new French users without any policy edit (Get started with sensitivity labels[4]).

In short, role groups decide what an administrator may configure, and administrative units narrow whom they manage and whom a policy targets. A label's encryption recipient list, taught in Configuring a label's encryption below, governs who opens content, never which administrator manages which division.

Scopes, label groups, and label priority

Three design decisions shape every label before anyone uses it: which scopes it has, where it sits in a two-tier menu, and its position in the label list. These decisions shape later configuration choices.

Label scopes

A label's scope determines which label settings you can configure and whether apps and services make the label available for users to see and select (Learn about sensitivity labels[1]). Microsoft documents four scopes (Learn about sensitivity labels[1]):

Scope What it covers
Files & other data assets Always selected for a new label; files for Office, Loop and Power BI, items from Microsoft Fabric, and Data Map assets
Emails Email messages; usually selected together with files, because many labeling features need both
Meetings Calendar events, Teams meeting options and Teams chat; also requires Files & other data assets and Emails
Groups & sites Teams sites, SharePoint sites, Viva Engage communities and Loop workspaces; available after you enable labels for containers, and labels the container, not its items

If you don't select a scope, you see the first page of its settings but can't configure them (Learn about sensitivity labels[1]). Scope also gates other features: a label without the Emails scope can't be selected for an auto-labeling policy that includes Exchange, and a label without the Files & other data assets scope can't be a default label for a SharePoint document library (Manage sensitivity labels in Office apps[7]). A label scoped only to Groups & sites isn't displayed in Office apps and can't be applied to files and emails (Use sensitivity labels with Teams, groups, and sites[8]). Adding such a label to another publishing policy never makes it appear in Outlook, because publishing changes who sees a label, not which scopes it has.

Label groups and sublabels

To show one sensitivity level with variants, you can group labels in a two-tier hierarchy: in Microsoft's example from the default sensitivity labels, Confidential holds Anyone (unrestricted), All Employees and Trusted People; users first see Confidential, then choose a label such as All Employees, and the applied label is Confidential \ All Employees (Learn about sensitivity labels[1]). A label group arranges labels in the menu; it is unrelated to the Microsoft 365 groups that a Groups & sites label protects and to the user groups a publishing policy targets (Learn about sensitivity labels[1]). The original design used parent labels with sublabels; label groups are replacing parent labels, look the same to users, and have only the settings they support: name, descriptions, color and priority (Learn about sensitivity labels[1]). In either scheme a second-tier label inherits no settings from the first tier except the label color, so encryption configured on a Confidential parent never protects its sublabels; configure each label that needs protection on its own (Learn about sensitivity labels[1]).

The two schemes differ in what can be published and applied:

  • Label groups (modern scheme). A label group can't be published itself, only the labels inside it (Learn about sensitivity labels[1]). Display names must be unique within a group and among standalone labels, so moving a label into a group that already has a label with the same display name fails with a LocalizedDisplayNameConflictException error (Create and publish sensitivity labels[5]).
  • Parent labels (classic scheme). A parent label with sublabels can't be applied to content, so don't choose it as a default label or configure it for automatic or recommended labeling (Learn about sensitivity labels[1]). When you select a sublabel for a publishing policy, also select its parent label (Create and publish sensitivity labels[5]).

Keep the menu short: Microsoft reports that effectiveness is noticeably reduced when users have more than five main labels or more than five sublabels per main label (Learn about sensitivity labels[1]).

Label priority and the replacement rule

The order of labels on the Labels page sets their priority: put the least restrictive label, such as Public, at the top and the most restrictive, such as Highly Confidential, at the bottom (Learn about sensitivity labels[1]). The bottom label has the highest order number and the highest priority; in Microsoft's example, Public has order number 0 and Highly Confidential order number 4 (Automatically apply a sensitivity label[3]). The figure below shows that order with the Confidential label group from Microsoft's example in the middle.

Priority settles what happens when an automatic method meets a label that is already on the content. Microsoft's default behavior forms one replacement rule (Automatically apply a sensitivity label[3]):

  • A manually applied label isn't replaced by automatic labeling.
  • A lower-priority label that was applied automatically or as a default can be replaced by a higher-priority label.
  • A higher-priority label isn't replaced by a lower-priority one.

Later sections state each mechanism as a delta from this rule: several follow it exactly, a publishing-policy default never replaces any label, and a few mechanisms deliberately override manual labels. Publishing policies have a separate order of their own, explained in Publishing policies: availability, defaults, and precedence below; a policy's position never ranks a label's sensitivity.

What to remember: scopes decide where a label can appear, label groups organize the menu without passing on protection, and the label order (bottom is highest) guides automatic selection and replacement, subject to each method's exceptions.

PublicConfidential (label group)Anyone (unrestricted)All EmployeesTrusted PeopleHighly ConfidentialLowerpriorityHigherpriority
Label priority in the Labels list: the top label is lowest priority and the bottom label highest (labels from Microsoft's default sensitivity labels)

Scripting labels with Security & Compliance PowerShell

Security & Compliance PowerShell can create and configure all the settings of labels and publishing policies: New-Label and Set-Label manage labels, New-LabelPolicy and Set-LabelPolicy manage publishing policies, and Remove-Label and Remove-LabelPolicy script deletion (Create and publish sensitivity labels[5]). The label and the policy that distributes it are separate objects, so translating a label or changing its advanced settings is a Set-Label job, not a Set-LabelPolicy job.

Microsoft recommends identifying a label by its GUID (globally unique identifier): the label name is unique in the tenant, but the display name isn't, so targeting a display name could configure the wrong label (Create and publish sensitivity labels[5]). The listing below applies that tip to Microsoft's localization example, where the -LocaleSettings parameter lets users in multinational deployments see the label name and tooltip in their own language (Create and publish sensitivity labels[5]).

Listing: find a label's GUID and scope, then set localized display names and tooltips

# Name is unique in the tenant; DisplayName is not. ContentType confirms the label's scope.
Get-Label | Format-Table -Property DisplayName, Name, Guid, ContentType

# ... build $DisplayNameLocaleSettings and $TooltipLocaleSettings as in Microsoft's example
#     (LocaleKey DisplayName or Tooltip, with one language key and value per language) ...
Set-Label -Identity <label-GUID> -LocaleSettings (ConvertTo-Json $DisplayNameLocaleSettings -Depth 3 -Compress),(ConvertTo-Json $TooltipLocaleSettings -Depth 3 -Compress)

In the first command, the Guid column is the value to pass to -Identity, and Microsoft uses the ContentType column to confirm the label's scope (Create and publish sensitivity labels[5]). Container labels are the exception to localization: they display only the original language for the label name and description (Use sensitivity labels with Teams, groups, and sites[8]).

In short, a label and the publishing policy that distributes it are separate objects with separate cmdlets, and scripts should target labels by GUID.

Configuring a label's encryption

Encryption is the label setting that decides who can open content and what they can do with it, and it travels with the item. For sensitivity labels to apply encryption, the Azure Rights Management service from Microsoft Purview Information Protection must be activated for the tenant; newer tenants have it on by default, but you might need to activate it manually (Restrict access to content by using encryption[9]).

Two permission modes

A label's encryption uses one of two modes (Restrict access to content by using encryption[9]):

  • Assign permissions now (administrator-defined permissions): you decide exactly which users get which permissions for every item that has the label.
  • Let users assign permissions (user-defined permissions): users decide when they apply the label. In Outlook they get restrictions equivalent to the Do Not Forward option or Encrypt-Only for their chosen recipients; in Word, PowerPoint and Excel they can be prompted to specify users and permissions.

Choose administrator-defined permissions when every item in a category should have the same readers and rights. Choose user-defined permissions when collaborators differ from item to item, instead of creating a new label for each recipient combination. The two Outlook options behave differently: Do Not Forward means recipients can't forward, print or copy the email, while Encrypt-Only gives recipients all usage rights except Save As, Export and Full Control, so they can't remove the protection but can otherwise copy, print and forward (Restrict access to content by using encryption[9]).

User-defined permissions also depend on scope and client: Do Not Forward and Encrypt-Only need the Emails scope, the Office prompt needs the Files & other data assets scope, and not all labeling clients support every option (Restrict access to content by using encryption[9]). A label that lets users assign permissions can be recommended, but it can be applied automatically only with the Do Not Forward and Encrypt-Only options (Restrict access to content by using encryption[9]).

Keep the encrypting catalog small: a tenant can have 1,000+ sensitivity labels, but at most 500 labels whose encryption specifies the users and permissions (Learn about sensitivity labels[1]).

Who the encryption admits

With administrator-defined permissions you choose the audience from these options (Restrict access to content by using encryption[9]):

  • specific users or groups;
  • Everyone in your organization, meaning all tenant members, which excludes guest accounts;
  • Any authenticated users, which still encrypts the content and restricts how it can be used, but doesn't restrict who can open it, as long as the opening app supports the authentication used;
  • any email address or domain, where entering a domain of an organization that uses Microsoft Entra ID automatically includes all verified domains of the tenant that owns it, so you can't restrict access to that one domain.

This encryption audience is independent of the publishing audience. The publishing policy decides who can see and apply the label, while Azure Rights Management decides who can open the protected content, and Azure Rights Management doesn't support non-email-enabled security groups or exclusions (Learn about sensitivity labels[1]). Excluding a guest from a publishing policy therefore doesn't take away access that the encryption already granted.

Usage rights: editing is not removing protection

Usage rights are the individual permissions that the encryption grants. Four of them are easy to confuse (Rights Management usage rights[10]):

Usage right (encoding in policy) What it allows
Edit Content, Edit (DOCEDIT) Modify, rearrange, format or sort the content in the app; doesn't grant the right to save the edited copy
Save (EDIT) Save the item; in Office apps, saving under a new name or location is limited to formats that support Rights Management, so the original encryption can't be removed
Save As, Export (EXPORT) Save the content under a different file name and use other export options
Full Control (OWNER) All rights to the item, including removing encryption and re-encrypting

Removing existing encryption, for example by relabeling, takes the Export or Full Control usage right, the role of Rights Management issuer or Rights Management owner, or super user status (Restrict access to content by using encryption[9]). The Rights Management issuer is the account that encrypted the content and, by default, is also its Rights Management owner; the Full Control usage right isn't the same thing as the Rights Management owner role (Rights Management usage rights[10]). The super user feature is the recovery path: it ensures that authorized people and services can always read and inspect data encrypted for your organization, and then remove or change the encryption (Restrict access to content by using encryption[9]). A label-administration role only lets someone create and configure labels and policies (Get started with sensitivity labels[4]); opening another person's encrypted document still depends on usage rights or super user status.

Double Key Encryption and key choices

Double Key Encryption (DKE) protects data with two keys: one key in your control through your own Double Key Encryption service, and a second key stored in Microsoft Azure; viewing the data requires access to both keys (Double Key Encryption[11]). Select the DKE label option only after the Double Key Encryption service is configured, because after the label is configured and saved you can't edit it (Restrict access to content by using encryption[9]). The trade-off is service processing: SharePoint and OneDrive can store DKE-encrypted files but can't process them, so Office for the web, coauthoring, eDiscovery, data loss prevention and search don't work for those files (Double Key Encryption[11]). A DKE label also can't be a SharePoint library default and can't protect calendar items, Teams meetings and chat, and DKE content is used through Microsoft 365 Apps for enterprise on Windows (Double Key Encryption[11]).

Microsoft positions DKE for mission-critical data only and recommends Microsoft-managed keys or bring your own key (BYOK) for most data, because those options keep the Microsoft 365 services that can't work with DKE-encrypted content (Double Key Encryption[11]). BYOK is therefore not a substitute when the requirement is a second key held in your own service: that second key is what defines DKE (Double Key Encryption[11]).

The pieces fit together this way: the permission mode decides who chooses the readers, the audience and usage rights decide who opens and what they can do, and the key option decides whether Microsoft 365 services can still process the content.

Encrypted content over time: expiry, changes, revocation

Encryption also has a clock. Three behaviors decide what happens after content is labeled: how expiry and offline access limit later openings, how edits to a published label reach existing items, and how a protected document can be tracked and revoked.

Expiry and offline access are different limits

With administrator-defined permissions, User access to content expires ends access on a specific date or a number of days after the label is applied, while Allow offline access (never, always, or a number of days) controls how long users can open the content without reauthenticating (Restrict access to content by using encryption[9]). The mechanism behind offline access is the use license: when a user opens encrypted content, Azure Rights Management grants a use license, a certificate holding the user's usage rights and the content's encryption key; while it is valid the user isn't reauthenticated or reauthorized, and when it expires the next opening requires both (Restrict access to content by using encryption[9]). If offline access is restricted to never or a number of days, users must reauthenticate when that threshold is reached and their access is logged (Restrict access to content by using encryption[9]). An expiry date alone therefore sets an end date, not how often users must reauthenticate.

Only labels configured to assign permissions now support different offline-access values; labels that let users assign permissions, such as Do Not Forward, Encrypt-Only and the Office prompt, use the tenant's Rights Management use license validity period, which defaults to 30 days (Restrict access to content by using encryption[9]). Microsoft recommends leaving expiry at Never unless the content has a time-bound requirement, offline access of 7 days for sensitive business data, and Never for labels with dynamic watermarks (described in Content markings: headers, footers, and watermarks below) and for very sensitive data (Restrict access to content by using encryption[9]). The Rights Management issuer is the exception to both limits: they can still open and edit after an expiry date and can always access the item offline (Rights Management usage rights[10]).

Editing a published label's encryption

A common deployment starts with labels that don't encrypt and later edits some of them to apply encryption, and the effect on existing items depends on where they are (Restrict access to content by using encryption[9]):

  • Newly labeled items get the encryption once the label change reaches apps.
  • Files in SharePoint and OneDrive with file-label integration change their encryption status the next time they are accessed, for example opened in Office for the web or downloaded, with no need to relabel. File-label integration is the tenant setting that lets these services process labeled, encrypted files; it is taught in Labeled files in SharePoint and OneDrive below.
  • Other previously labeled documents and emails keep their previous encryption status until the label is removed and reapplied, and this holds in both directions, adding or removing encryption.
  • Changed users or permissions on an administrator-defined label apply to existing items when users authenticate with the encryption service; a user who already opened the item gets the new settings only after their use license expires.
  • Switching permission modes, for example from assign permissions now to let users assign permissions, or from Do Not Forward to Encrypt-Only, applies only to newly labeled or relabeled items.

Tracking and revoking protected documents

Document tracking tells administrators when a protected document was accessed, and administrators and users can revoke access to tracked documents (Track and revoke document access[12]). A document must be registered first: supported Word, Excel and PowerPoint versions register label-protected local Office documents the next time they are opened (Track and revoke document access[12]). Track and revoke supports Office file types only and doesn't support password-protected documents (Track and revoke document access[12]). Registration gives the document a ContentID, the identifier used for subsequent tracking and revocation operations, as the listing below shows. The cmdlets come from the AIPService PowerShell module, and Connect-AipService must run first to connect to your tenant (Track and revoke document access[12]).

Listing: find, track, revoke and restore a registered document with the AIPService module

Connect-AipService
# Find the ContentID by file name and the user who applied protection
Get-AipServiceDocumentLog -ContentName "test.docx" -Owner "alice@contoso.com" -FromTime "12/01/2020 00:00:00" -ToTime "12/31/2020 23:59:59"
# Return successful and denied access attempts for that ContentID
Get-AipServiceTrackingLog -ContentId <ContentID>
# Revoke access; undo a mistaken revocation with the same ContentID and IssuerName
Set-AipServiceDocumentRevoked -ContentId <ContentID> -IssuerName <issuer>
Clear-AipServiceDocumentRevoked -ContentId <ContentID> -IssuerName <issuer>

Each command follows Microsoft's documented procedure (Track and revoke document access[12]). Three limits shape the result (Track and revoke document access[12]):

  • If offline access is allowed, users keep opening a revoked document until the offline period expires.
  • Documents protected with admin-defined permissions lose their ContentID when uploaded to SharePoint or OneDrive.
  • A copy of a file protected with admin-defined permissions downloaded from SharePoint or OneDrive gets a new ContentID, so revoking the original ContentID doesn't revoke downloaded copies.

Revoking encrypted email is a different workflow that belongs to Message Encryption, covered in Implement information protection for Windows, file shares, and Exchange.

Put together: expiry ends access on a date, offline access sets how stale a cached permission may get, label edits reach existing items only by the paths listed above, and revocation works per registered ContentID, never per every copy.

Content markings: headers, footers, and watermarks

Encryption controls access without being visible; content marking makes the classification visible on the content itself. A label can add watermarks, headers or footers when Office apps apply it (Learn about sensitivity labels[1]).

Static markings

Watermarks apply to documents and to Loop components and pages, but not to email or meeting invites, which support headers and footers (Learn about sensitivity labels[1]). Markings support variables, for example inserting the label name or the document name into a header, footer or watermark (Learn about sensitivity labels[1]). Test documents that templates or workflows depend on, because of string limits: watermarks are limited to 255 characters, and headers and footers to 1,024 characters except in Excel, which has a total limit of 255 characters for headers and footers that includes invisible formatting codes; a string over that Excel limit isn't displayed (Learn about sensitivity labels[1]).

Two boundaries matter for requirements. Users can change or remove static content markings after they are applied (Manage sensitivity labels in Office apps[7]). And auto-labeling policies don't apply a label's headers, footers or watermarks to documents (Automatically apply a sensitivity label[3]). A requirement for visible handling text on printed documents needs a marking; encryption alone restricts access but adds no text.

Dynamic watermarks

A dynamic watermark inserts the viewing user's UPN (their sign-in name, typically the same as their email address) on each page of the file by default, and optional custom text can include the date and time (Restrict access to content by using encryption[9]). It is supported only for labels that apply encryption, and Microsoft recommends setting Allow offline access to Never for these labels (Restrict access to content by using encryption[9]). Because it names the person who opened the document, it works as a deterrent against screen captures in a way a static Confidential watermark can't (Restrict access to content by using encryption[9]). Once applied, a dynamic watermark can't be changed or removed without removing the label or exporting the content, where that is allowed (Manage sensitivity labels in Office apps[7]).

Client support decides access, not only appearance (Restrict access to content by using encryption[9]):

  • In two documented cases, file preview for attachments in Outlook and PowerPoint Live in Microsoft Teams, the document opens without the watermark.
  • Otherwise, the user must be the Rights Management owner, open the document in an app that understands dynamic watermarks, or open it in Office on the web.
  • Office apps that don't support dynamic watermarks deny access rather than open the document without it.

Verify client versions before deploying such a label. The overall rule: static markings label the page for everyone and can be edited by users, while dynamic watermarks personalize the page for each viewer and only work on encrypted labels with supporting clients.

Publishing policies: availability, defaults, and precedence

A publishing policy is the object that turns a configured label into a choice in someone's Sensitivity menu: it holds a set of labels, the users and groups who get them, and policy settings such as default labels (Learn about sensitivity labels[1]). It is distinct from an auto-labeling policy, which applies labels to content instead of offering them to people.

Who gets the labels

Labels can be published to include or exclude specific users and supported groups: security groups (email-enabled or not), distribution groups and Microsoft 365 groups, which can have dynamic membership (Learn about sensitivity labels[1]). Remember from Configuring a label's encryption above that Azure Rights Management ignores those exclusions and non-email-enabled security groups for access to encrypted content. A label is reusable, so you can pilot it in a policy for a few users and later publish it to all users in another policy (Get started with sensitivity labels[4]). Editing a label that is already published needs no new policy, but allow up to 24 hours for the change to replicate to all apps and services (Create and publish sensitivity labels[5]). Microsoft advises having as few publishing policies as possible, and one policy for the whole organization is common (Create and publish sensitivity labels[5]). Menu availability comes only from publishing; an auto-labeling policy never adds a label to anyone's menu.

All users in the organization can see the name of a label applied to content, even if that label isn't published to them (Learn about sensitivity labels[1]).

Settings a publishing policy carries

  • Default label. Separate defaults are available for unlabeled documents (and Loop components and pages), emails and meeting invites, new containers, and Power BI content, and users can change an applied default (Learn about sensitivity labels[1]). Microsoft advises against an encrypting default label for documents, because many organizations share with external users who might not have apps that support the encryption (Learn about sensitivity labels[1]). Meetings and calendar events have their own default, covered in Meetings, calendar items, and Teams channel meetings below.
  • Mandatory labeling. It requires a label before users save files, send emails or meeting invites, create groups or sites, or use unlabeled Power BI content, and a label assigned manually, automatically or by default satisfies it (Learn about sensitivity labels[1]). It guarantees that a label exists, not that it is the right one, and without a matching default label it can frustrate users with frequent prompts (Learn about sensitivity labels[1]). Because of a timing issue, client-side auto-labeling (the label setting described in Client-side auto-labeling in Office apps below) might not trigger when mandatory labeling is also configured (Automatically apply a sensitivity label[3]).
  • Justification for changing a label. For files, emails and meetings, but not groups and sites, users must justify removing a label or replacing it with a lower-priority label; Office apps prompt once per app session, the Information Protection client prompts for each file, and administrators read the reasons in activity explorer, the Purview view of historical labeling activity (Learn about sensitivity labels[1]). For this setting, sublabels share the priority of their parent label, so changing between sublabels under the same parent needs no justification, and neither does changing between sublabels in the same label group; automatic labeling, by contrast, does use sublabel priority (Learn about sensitivity labels[1]). Lower priority means closer to the top of the label list from Scopes, label groups, and label priority above, so put more restrictive labels lower in that list.
  • Custom help page. A Learn More URL appears after the list of available labels in Office apps, for guidance that goes beyond each label's tooltip (Learn about sensitivity labels[1]).
  • Email inherits highest priority label from attachments. On the policy's Default settings for emails page, this setting lets Outlook select the highest-priority label among the attached files that is published to the sender; the attachment must be a physical file, not a link to a file on SharePoint or OneDrive, and the label must have both the Files & other data assets and Emails scopes (Manage sensitivity labels in Office apps[7]). It follows the replacement rule exactly: a manually labeled email keeps its label, a lower-priority automatic or default email label is replaced, and a higher-priority label stays (Manage sensitivity labels in Office apps[7]). Select Recommend users apply the attachments label instead of automatically applying it when users should confirm; without it, the label is applied automatically, and if that label encrypts, the same encryption is applied to the email (Manage sensitivity labels in Office apps[7]).

When a user is in several policies

Publishing policies have their own priority order on the Label policies page: the policy at the top has the lowest order number and lowest priority, and the policy at the bottom the highest (Learn about sensitivity labels[1]). A user included in several policies gets all the labels and settings from those policies, and when settings conflict, the setting from the policy with the highest order number applies, so the highest priority wins for each setting (Learn about sensitivity labels[1]). The figure below shows both halves for a user in an all-users policy (order 0) that offers Public as its default and a legal policy (order 1) that offers Highly Confidential as its default, two labels from the label list in Scopes, label groups, and label priority above: the labels are combined, and the conflicting default label comes from the legal policy.

Two consequences follow. Removing a label from a departmental policy doesn't hide it from a user who still receives it from the baseline policy, because labels combine. And a departmental default label or mandatory setting overrides the baseline only if its policy has the higher order number; if a user doesn't get the expected setting, check the policy order and move the policy down if needed (Learn about sensitivity labels[1]).

This policy order is a different list from label priority, although both put the winner at the bottom. Label priority ranks sensitivity and drives downgrade justification and automatic labeling; policy priority only picks which policy's setting a user receives.

In short, a publishing policy answers three questions: which labels a user can choose, which defaults and requirements apply to them, and, when policies overlap, which policy's setting wins.

All users policyorder 0Label:PublicDefault:PublicLegal policyorder 1Label:Highly ConfidentialDefault:Highly ConfidentialUser in both policiesLabels: all ofPublic,Highly ConfidentialDefault label:order 1 winsHighly Confidential
Publishing policy precedence for a user in two policies: labels combine, and each conflicting setting comes from the higher-order policy

Labeled files in SharePoint and OneDrive

Suppose a team library holds Office files that a sensitivity label encrypted: until file-label integration is enabled, coauthoring, search and eDiscovery don't work for those files (Enable sensitivity labels for files in SharePoint and OneDrive[13]), and the site admin can also configure a default label that SharePoint applies to new files uploaded to the library (Configure a default sensitivity label for a SharePoint document library[14]). Labels in SharePoint and OneDrive depend on separate settings: file-label integration for the tenant, a default label for a document library, and a library option that extends SharePoint permissions to downloaded files.

File-label integration

Enabling sensitivity labels for Office files in SharePoint and OneDrive, called file-label integration in this topic, gives users the Sensitivity button in Office for the web and lets the services process the contents of Office files, and optionally PDF documents, that a sensitivity label has encrypted (Enable sensitivity labels for files in SharePoint and OneDrive[13]). Until it is enabled, coauthoring, eDiscovery, data loss prevention, search and other collaborative features don't work for those encrypted files (Enable sensitivity labels for files in SharePoint and OneDrive[13]). The processing covers encryption with a cloud-based key; files encrypted with Double Key Encryption or an on-premises key (hold your own key, HYOK) still can't be processed (Enable sensitivity labels for files in SharePoint and OneDrive[13]).

You can turn it on with the Turn on now banner in the Microsoft Purview portal, signed in as a global administrator, or with Set-SPOTenant -EnableAIPIntegration $true in the SharePoint Online Management Shell; with Microsoft 365 Multi-Geo you must use PowerShell for each geo-location (Enable sensitivity labels for files in SharePoint and OneDrive[13]). PDF support is a separate switch, Set-SPOTenant -EnableSensitivityLabelforPDF $true (Enable sensitivity labels for files in SharePoint and OneDrive[13]). Files that were already labeled with encryption before you enabled it gain the new capabilities only after they are edited, or downloaded and uploaded again (Enable sensitivity labels for files in SharePoint and OneDrive[13]).

Keep three neighbors apart:

A default label for a document library

A site admin can configure a default sensitivity label in a document library's settings, and SharePoint then labels new files uploaded to the library, and existing files edited there, that have no label or a lower-priority label (Configure a default sensitivity label for a SharePoint document library[14]). It is location-based labeling without content inspection, and against the replacement rule from Scopes, label groups, and label priority above it behaves exactly as the rule predicts (Configure a default sensitivity label for a SharePoint document library[14]):

Existing label Replaced by the library default?
Manually applied, any priority No
Automatically applied or policy default, lower priority Yes
Automatically applied or policy default, higher priority No

A default label from a publishing policy, by contrast, never overrides an existing label (Configure a default sensitivity label for a SharePoint document library[14]). The library default needs a published label with the Files & other data assets scope, file-label integration, and a library without SharePoint Information Rights Management (IRM), the older protection technology that blocks selecting a default label (Configure a default sensitivity label for a SharePoint document library[14]). An encrypting library default can't use user-defined permissions with the Office prompt, expiry set to anything other than Never, or Double Key Encryption (Configure a default sensitivity label for a SharePoint document library[14]).

The library default doesn't apply to existing files at rest; a dedicated auto-labeling policy type extends it to them, described in Auto-labeling policies: locations, eligibility, and limits below (Configure a default sensitivity label for a SharePoint document library[14]). Timing differs by client: Office for the web applies the label without delay, an upload can take a few minutes, and Microsoft 365 desktop and mobile apps apply it after the app is closed (Configure a default sensitivity label for a SharePoint document library[14]).

Extending SharePoint permissions to downloads (preview)

The library settings also offer Extend protection on download, copy or move, which is in preview, supports a different labeling scenario, and must stay unselected for an ordinary library default (Configure a default sensitivity label for a SharePoint document library[14]). When it is selected, files downloaded from the library stay protected with the user's current SharePoint permissions instead of a fixed recipient list: SharePoint Owner maps to full control except that OWNER (Full Control, from the usage-rights table in Configuring a label's encryption above) and OBJMODEL (Allow Macros, the right to run macros or use other programmatic access, per Rights Management usage rights[10]) rights aren't currently applied, so owners can change the label but can't remove it, Edit to editing rights without changing the label, and Read to viewing and copying only (Extend SharePoint permissions to downloaded documents[15]).

The downloaded file stays tied to the original item (Extend SharePoint permissions to downloaded documents[15]):

  • Changes to the user's SharePoint permissions are reflected in the downloaded file.
  • The user can't open it if their permissions are removed, the file is deleted, the site is no longer active, or the file is moved to a different site.
  • Users can't open downloaded files offline, because they must be able to connect to the original site.
  • Labeled files can't be moved or copied to another site, and moving or copying them to another library in the same site requires permission to create or delete lists.

Setup requires Set-SPOTenant -ExtendPermissionsToUnprotectedFiles $true, file-label integration, co-authoring for encrypted files, and a published label with the Files & other data assets scope configured for Let users assign permissions when they apply the label with In Word, PowerPoint, and Excel, prompt users to specify permissions, although in this use users aren't actually prompted (Extend SharePoint permissions to downloaded documents[15]). This option departs from the replacement rule: it overrides any existing label that doesn't apply encryption, including a manually applied one, but never a label that encrypts (Extend SharePoint permissions to downloaded documents[15]).

File-label integration lets the services read encrypted files; for each library, choose either an ordinary default that gives a location baseline and respects manual labels or extended permissions that keep downloads bound to live SharePoint access.

Client-side auto-labeling in Office apps

Client-side auto-labeling is a setting on the label itself, Auto-labeling for files and emails, that supported Word, Excel, PowerPoint and Outlook apps evaluate while users edit documents or compose, reply to or forward email (Automatically apply a sensitivity label[3]). It is one of Microsoft's two automatic methods; the other, the service-side auto-labeling policy, is taught in the sections after this one.

Licensing for both automatic methods

Both client-side auto-labeling and auto-labeling policies are included with Microsoft 365 E5, Microsoft 365 E5 Compliance, Microsoft 365 E5 Information Protection & Governance and Azure Information Protection Premium P2 (standalone); with E3 or Business Premium, sensitivity labels and manual labeling are included but auto-labeling requires an E5 or P2-tier add-on (Automatically apply a sensitivity label[3]). Auto-labeling also isn't available in all regions because of a backend Azure dependency, and then the Auto-labeling page isn't visible in the Microsoft Purview portal (Automatically apply a sensitivity label[3]). When the surface is missing, check license, role and region before you touch the label's conditions.

Conditions and the user's choice

The conditions are SITs or trainable classifiers. For SITs you can refine the instance count and match accuracy (confidence), require all or any of the selected types, and add groups with logical operators between them, as in data loss prevention (DLP) policies (Automatically apply a sensitivity label[3]). The label is then applied automatically or recommended, and in both cases the user decides whether to accept or reject it; with the Word desktop app, Show sensitive content steps the user through each detection (Automatically apply a sensitivity label[3]). Because the client acts as the user works, a document can be labeled even before it is saved, but not all client apps support auto-labeling (Automatically apply a sensitivity label[3]). A publishing-policy default can't do this job, because it applies the same label regardless of content.

Exact data match (EDM) SITs need one extra element: the label must also specify at least one SIT that doesn't use EDM, such as a built-in type like Credit card number, and a label configured with only EDM conditions has its auto-labeling setting turned off automatically (Automatically apply a sensitivity label[3]). If only records confirmed by the EDM reference data should trigger the label, require both types with all-of logic; with any-of logic the generic type alone could trigger it (Automatically apply a sensitivity label[3]).

When several labels qualify

When conditions match more than one label, the label with the highest order number is selected (Automatically apply a sensitivity label[3]). Sublabels under the same parent label behave differently on the client (Automatically apply a sensitivity label[3]):

  • For a file that isn't labeled yet, the highest-order sublabel configured for automatic labeling is always selected over a sublabel configured for recommended labeling; only when none is set to automatic is the highest-order sublabel recommended.
  • For a file that already has a sublabel from the same parent, no action is taken and the existing sublabel remains, even if it was a default or applied automatically.

Auto-labeling policies, by contrast, select the highest-order qualifying sublabel (Automatically apply a sensitivity label[3]). Moving a recommended sublabel lower in the list can't make Office prefer it over an automatic sibling, because automatic versus recommended is decided first. With the classic scheme, never configure a parent label for automatic or recommended labeling, because it can't be applied (Automatically apply a sensitivity label[3]).

Against the replacement rule, the client follows it, except for the same-parent sublabel case above: it doesn't automatically label content that was manually labeled or automatically labeled with a higher-priority label, and it doesn't show a recommendation for content that already has a higher-priority label (Automatically apply a sensitivity label[3]).

Converting label conditions into an auto-labeling policy

If a label's conditions include SITs, the end of the label creation or editing wizard offers to create an auto-labeling policy from the same settings (Automatically apply a sensitivity label[3]). Trainable classifiers don't carry over: a label with only trainable classifiers offers no conversion, and a label with both converts only its SITs (Automatically apply a sensitivity label[3]). The generated policy includes all SharePoint, OneDrive and Exchange locations by default and runs in simulation when saved, and nothing checks that file-label integration is enabled for SharePoint and OneDrive (Automatically apply a sensitivity label[3]). Review the scope and detection logic before you save.

The client-side method fits when users should see, and can override, a content-based label while they work; it never processes content that nobody opens.

Auto-labeling policies: locations, eligibility, and limits

An auto-labeling policy labels content in the service: files already saved in SharePoint or OneDrive, which Microsoft calls data at rest, and email processed by Exchange Online, data in transit (Automatically apply a sensitivity label[3]). It doesn't depend on any app version a user has installed, which suits labeling at scale, and because no user interacts with it, it doesn't support recommended labeling; you run it in simulation first instead (Automatically apply a sensitivity label[3]). For Exchange, it doesn't include email at rest in mailboxes, so it never relabels historical mail (Automatically apply a sensitivity label[3]).

Prerequisites and label eligibility

Missing any prerequisite typically leaves the policy running without labeling any files and with no error in the portal (Automatically apply a sensitivity label[3]):

  • Auditing for Microsoft 365 is turned on, which simulation requires.
  • File-label integration is enabled for SharePoint and OneDrive locations, and PDF support is enabled if PDFs should be processed.
  • Auto-labeling is available in your region.
  • The label isn't a parent label, and its scope matches the content: Files & other data assets for files, Emails for email.
  • The label is published to at least one user.

Licensing for auto-labeling policies is described in Licensing for both automatic methods above.

Encrypting labels have further limits (Automatically apply a sensitivity label[3]). When the policy includes SharePoint or OneDrive, the label must use Assign permissions now with User access to content expires set to Never. A policy that is just for Exchange can also use Let users assign permissions with Do Not Forward or Encrypt-Only, but no auto-labeling policy can apply a label configured for S/MIME protection, the certificate-based email encryption option.

What gets processed

In SharePoint and OneDrive, Word (.docx), PowerPoint (.pptx) and Excel (.xlsx) files and PDF documents are supported (Automatically apply a sensitivity label[3]). A file that is open or checked out isn't labeled, attachments to list items aren't supported, and a OneDrive account must be associated with its user account for both simulation and auto-labeling (Automatically apply a sensitivity label[3]). Existing modified, modified by and date values aren't changed, in simulation or when labels are applied, and when the label encrypts, the Rights Management issuer and owner is the account that last modified the file (Automatically apply a sensitivity label[3]).

In Exchange, PDF and Office attachments are scanned for the conditions, but a match labels the email, not the attachment (Automatically apply a sensitivity label[3]). If an Exchange mail flow rule or a DLP policy that applies IRM encryption also matches, the label is applied; a label that encrypts causes the rule's IRM settings to be ignored, while a label without encryption is applied in addition to the rule's IRM encryption (Automatically apply a sensitivity label[3]). To predict the protection, check first whether the label itself encrypts; the two rights templates never accumulate.

Scale limits

A tenant can have a maximum of 100 auto-labeling policies (Automatically apply a sensitivity label[3]). In the portal, each policy can explicitly include or exclude up to 1,000 SharePoint sites, or up to 100 OneDrive users or groups, and keeping the default of All is exempt from these maximums (Automatically apply a sensitivity label[3]). The absolute maximum throughput is 500,000 automatically labeled SharePoint and OneDrive files per tenant per day across all policies; Microsoft states this is a ceiling, not a guaranteed or expected daily rate (Automatically apply a sensitivity label[3]).

Choosing locations

Exchange location scope follows the sender, not the recipient (Automatically apply a sensitivity label[3]):

  • Replacing the default of All included with specific users or groups exempts email sent from outside your organization.
  • Keeping All included but excluding users or groups exempts email those users send, but not email they receive.

Keep All included and None excluded when the policy must evaluate incoming email from external senders, and narrow internal recipients with advanced rule conditions, described in Auto-labeling policy rules and label replacement below, instead (Automatically apply a sensitivity label[3]). Including only a Finance group therefore never catches external mail sent to Finance. To encrypt matching email from outside senders, also select Apply encryption to email received from outside your organization and assign a Rights Management owner: a single user identified by an email address your organization owns, not a mail contact, a shared mailbox or any group (Automatically apply a sensitivity label[3]).

For OneDrive you specify users or groups; for SharePoint you can select up to 1,000 sites, or target sites dynamically with a SharePoint adaptive scope (Automatically apply a sensitivity label[3]). An adaptive scope defines its membership with a query you specify, and that query runs daily against the attributes or properties you choose; policies for retention can use adaptive scopes too (Adaptive scopes[16]; see Implement and manage retention). You attach the adaptive scope in PowerShell with New-AutoSensitivityLabelPolicy -SharePointAdaptiveScopes, which replaces static site enumeration with dynamic membership rather than raising the portal limit (Automatically apply a sensitivity label[3]). In PowerShell the policy and its rules are separate objects: New-AutoSensitivityLabelPolicy creates the policy and New-AutoSensitivityLabelRule -Policy adds a rule to it (Automatically apply a sensitivity label[3]). When a policy uses administrative units, SharePoint can't use the Included or Excluded options and must keep All (Automatically apply a sensitivity label[3]).

The library-default policy type

One auto-labeling policy type inspects no content: it applies each selected SharePoint document library's default sensitivity label to eligible existing files at rest, which the ordinary library default from Labeled files in SharePoint and OneDrive above can leave unlabeled (Automatically apply a sensitivity label[3]). It gives a location-defined baseline, not content-based differentiation. Its limits (Automatically apply a sensitivity label[3]):

  • Select specific SharePoint sites rather than All; administrative-unit scoping isn't supported.
  • Only one policy of this type per tenant, with up to 10 SharePoint sites; for more sites, let labeling finish, review the results, then edit the policy to replace the selection with the next batch of up to 10.
  • If a site can't be added, the error distinguishes a missing library default (configure one, or choose another site) from defaults that couldn't be verified (check your access to the site and library metadata).

The general 1,000-site allowance doesn't apply to this policy type. Taken together, an auto-labeling policy is the unattended, scalable path, but only for eligible labels, supported files and a scope chosen with the sender-based Exchange semantics in mind.

Auto-labeling policy rules and label replacement

Once the locations are set, rules decide which items a policy labels, and replacement settings decide what happens to labels that are already there.

Common or advanced rules

Common rules define one set of rules across all selected locations; Advanced rules let you define different rules per location, including rules only available for Exchange or for SharePoint sites and OneDrive accounts (Automatically apply a sensitivity label[3]). Conditions include Content contains (SITs or trainable classifiers), Content is shared (only with people inside my organization, or with people outside my organization), and attachment or file extension, name, property and size conditions (Automatically apply a sensitivity label[3]). Advanced rules for Exchange add conditions such as Sender is, Sender domain is, Recipient is, Recipient domain is, Header matches patterns and Subject matches patterns, among others; SharePoint and OneDrive add Document created by (Automatically apply a sensitivity label[3]). Exchange rules have condition-specific limits: Sender domain is accepts up to 600 domain entries, and the FromAddressMatchesPatterns parameter accepts at most 10 regular expressions of up to 128 characters each; neither is a combined limit for the whole policy (Automatically apply a sensitivity label[3]). Policies also support exceptions and nested AND, OR and NOT logic; the label's client-side setting supports neither but offers all-or-any matching and logical-operator groups, described in Client-side auto-labeling in Office apps above (Automatically apply a sensitivity label[3]). A subject or sender condition lets a policy target message context instead of relying only on SIT counts, as in Microsoft's example that labels email whose subject matches a project code name (Automatically apply a sensitivity label[3]).

Document property is uses SharePoint managed properties and needs exact string matches; regular expressions aren't supported (Automatically apply a sensitivity label[3]). Mapped values that contain an ampersand (&) might not match, even when the strings agree, so Microsoft suggests a classification value without ampersands, used consistently in the document metadata and the rule, then reviewing simulation and confirming actual labeling in a limited scope (Automatically apply a sensitivity label[3]).

Applying versus removing labels

When you create a policy you choose Automatically apply label only or Automatically remove label only (Automatically apply a sensitivity label[3]). A removal policy works on SharePoint and OneDrive, can remove the selected label even when it was applied manually, and removes any encryption that label applied, without the policy creator needing the Export or Full Control usage right (Automatically apply a sensitivity label[3]). It doesn't support password-protected files or files protected with external or non-Microsoft encryption, even when simulation finds them (Automatically apply a sensitivity label[3]).

An apply policy never removes a label to leave content unlabeled, and by default it follows the replacement rule from Scopes, label groups, and label priority above: manual labels and higher-priority labels stay, and lower-priority automatic or default labels are replaced (Automatically apply a sensitivity label[3]). Its one delta is optional: on the Additional label settings page you can let the policy override an existing label of lower priority, including a manually applied one, for Emails only or for All locations (Automatically apply a sensitivity label[3]). Moving a label lower in the list alone never replaces manual labels.

Conflicts between active policies resolve in favor of keeping or applying a label (Automatically apply a sensitivity label[3]):

  • When an apply policy and a removal policy both match a file, the label is applied rather than removed, even if the apply policy applies a different label than the one selected for removal.
  • When a policy configured to replace manual labels meets another active policy that applies a label without that option, the manual label isn't replaced.

Simulation shows the result of one policy at a time, so neither conflict appears in simulation; it is resolved only when all policies run (Automatically apply a sensitivity label[3]).

The result: rules choose the items, the apply-or-remove choice sets the direction, and only the explicit override option lets a policy replace a manual label.

Simulate, turn on, and monitor auto-labeling policies

Simulation is a required gate in the life of every auto-labeling policy: you can't automatically label documents and emails, or remove labels, until the policy has run at least one simulation (Automatically apply a sensitivity label[3]). The figure below traces the lifecycle this section walks through: create, simulate, review and refine, turn on, monitor.

Running a simulation

Simulation evaluates and logs matches without applying or removing labels, can take 12 hours to complete, and sends an email notification to the user configured to receive activity alerts when it finishes (Automatically apply a sensitivity label[3]). It isn't fully silent: alert policies configured on auto-labeling or SIT activity still fire on matches during simulation (Automatically apply a sensitivity label[3]). If more than 20,000,000 files match, you can't turn the policy on; reconfigure it to match fewer files and rerun simulation (Automatically apply a sensitivity label[3]). For Exchange, simulation evaluates email sent and received during the run, so results are comparable only if the same messages are sent and received again (Automatically apply a sensitivity label[3]). Duplicating a policy creates a new policy in simulation regardless of the source policy's state, so the copy passes its own gate (Automatically apply a sensitivity label[3]).

In the wizard, Run policy in simulation mode can be combined with turning the policy on automatically if it isn't edited for 7 days (Automatically apply a sensitivity label[3]). To start or rerun simulation from PowerShell, connect with Connect-IPPSSession -EnableSearchOnlySession; the switch is required so that the simulation job can be submitted correctly (Automatically apply a sensitivity label[3]).

Reading simulation results

The Matched items count is an estimate, not a count of successfully labeled files (Review simulation results for auto-labeling policies[17]). Simulation shows the result of a single policy using the classification available at the time, so overlapping policies, files that change, and labeling failures can all make enforcement differ (Automatically apply a sensitivity label[3]). Displayed items are capped at 100 per site and CSV exports at 50,000 records, so neither view lists every match (Automatically apply a sensitivity label[3]). Sample items are retained for 30 days, but that retention doesn't guarantee access to earlier runs, so export the results you need for comparison, and check the exported file, before you edit the policy or restart simulation (Review simulation results for auto-labeling policies[17]).

Missing evidence is not a zero (Review simulation results for auto-labeling policies[17]):

  • Simulation couldn't be started and Simulation failed don't mean that the policy completed with no matches.
  • A workload marked Not available isn't a zero-match result, and a count marked (partial) includes only the available workload results.
  • Restart simulation tries again but doesn't guarantee a fix; review the new run's status and results before turning the policy on.

Turning the policy on and monitoring it

When results meet your requirements, select Turn on policy; if that action is unavailable, check the policy status, the simulation status and your permissions (Automatically apply a sensitivity label[3]). Monitoring then differs by workload (Automatically apply a sensitivity label[3]):

  • SharePoint and OneDrive files. Labeling progress, Labeled items, Labeling failures and coverage reports cover files only. Progress updates every 48 hours, so a newly enabled policy first shows 0 files to be labeled. Review each file's Failure reason: transient infrastructure failures are retried automatically, file format, protection or label configuration failures might need reason-specific action, and some unsupported files need no action. A file that is missing from Labeling failures isn't proof that it was labeled.
  • Coverage by simulation context (preview). It reports files as labeled during enforcement, pending, not yet attempted, or failed to apply a label, relative to the latest simulation run. It is available for 30 days after that run and isn't an all-time total, so its counts can differ from labeled totals; run simulation again for a new coverage window. Rerunning an Exchange-only policy in simulation doesn't enable file coverage.
  • Exchange email. It isn't included in those views. Filter activity explorer for Sensitivity label applied and check How applied for automatic labeling, knowing that the event doesn't identify which auto-labeling policy or rule applied the label.

When an Exchange rule stops matching

In rare cases a malformed rule fails to load: Exchange messages stop matching an enabled policy, and there is no item-level failure to review because no content was evaluated against that rule (Automatically apply a sensitivity label[3]). Contact Microsoft Support, and delete and recreate a rule only after Support identifies that specific rule (Automatically apply a sensitivity label[3]). The repair sequence: record the rule's policy, name, target label, workload and scope, conditions, exceptions, priority and enabled state; delete only that rule; recreate it with the same configuration; run simulation with newly sent or received messages; then turn the policy on and confirm that matching resumes (Automatically apply a sensitivity label[3]). Enforcement from the rule pauses until the recreated rule propagates, and deleting other rules as a troubleshooting step only removes their enforcement (Automatically apply a sensitivity label[3]).

The whole lifecycle comes down to evidence: simulation estimates, enforcement reports measure, and each workload has its own place to look.

Create the policyRun simulationcan take 12 hoursReview the resultsexport before editsRefineand rerunTurn on policyMonitor by workloadfiles and Exchange
Auto-labeling policy lifecycle: simulation is required before the policy can be turned on, and monitoring differs for files and Exchange email

Container labels for Teams, groups, and SharePoint sites

A container label protects the workspace rather than the files in it. It controls settings such as privacy, external user access, external sharing and access from unmanaged devices for Microsoft Teams sites, Microsoft 365 groups, SharePoint sites, Viva Engage communities and Loop workspaces (Use sensitivity labels with Teams, groups, and sites[8]). Items in the container don't inherit the label, so no item-level setting such as content marking or encryption reaches the files (Use sensitivity labels with Teams, groups, and sites[8]). A private-team label is therefore not a way to encrypt every uploaded document; label the files with the item-labeling methods from earlier sections.

Turning container labels on

Enabling container labels is a one-time sequence, traced in the figure below (Use sensitivity labels with Teams, groups, and sites[8]):

  1. Enable sensitivity label support for Microsoft 365 groups in Microsoft Entra ID, which uses the EnableMIPLabels directory setting.
  2. Connect to Security & Compliance PowerShell and run Execute-AzureAdLabelSync to synchronize your labels to Microsoft Entra ID.
  3. Create or edit a label with the Groups & sites scope and configure its settings; before steps 1 and 2, these settings are visible but can't be configured.
  4. Publish the label to the workspace owners and the users who create teams, groups and sites.

After a label is applied, changing it takes a Microsoft 365 group owner for a group-connected site, and a SharePoint site admin for a site that isn't group-connected (Use sensitivity labels with Teams, groups, and sites[8]). Microsoft advises not changing the site and group settings of a label after it is applied; if you do, wait at least 24 hours for the change to replicate (Use sensitivity labels with Teams, groups, and sites[8]). If a document uploaded to a labeled site has a higher-priority label than the site, SharePoint generates an audit event and an email, so label order matters across item and container labels too (Use sensitivity labels with Teams, groups, and sites[8]).

Privacy and guest access

Privacy can be Public (anyone in the organization can access the container), Private (only approved members), or None (users choose the privacy setting themselves); Public and Private set the privacy setting and lock it so it can be changed only by first removing the label (Use sensitivity labels with Teams, groups, and sites[8]). External user access controls whether the owner can add guests to the container (Use sensitivity labels with Teams, groups, and sites[8]). Tightening it applies to new users, not existing ones: guests who already accessed the site keep access after the setting is cleared, so review existing membership separately (Use sensitivity labels with Teams, groups, and sites[8]).

External sharing and Conditional Access

Control external sharing from labeled SharePoint sites sets external sharing to anyone, new and existing guests, existing guests, or only people in your organization (Use sensitivity labels with Teams, groups, and sites[8]). Two further settings work with Microsoft Entra Conditional Access, the Entra policy engine that grants or limits access based on conditions, and each needs its dependent Conditional Access configuration (Use sensitivity labels with Teams, groups, and sites[8]):

  • Unmanaged devices. The label can block or limit access from unmanaged devices, but it has no effect unless the SharePoint app-enforced restrictions Conditional Access policy is configured, and no effect if it is less restrictive than the tenant-level setting; in a tenant set to allow limited, web-only access, a label that allows full access does nothing.
  • Authentication context. The label can enforce stricter conditions, such as multifactor authentication (MFA) or terms of use, through an existing authentication context that has been created and published in Conditional Access; users who don't meet the conditions, or whose apps don't support authentication contexts, are denied access.

These settings can hand power to site owners: once external sharing or authentication context settings are published in a label, a site owner can change them for a site by applying or changing the label (Use sensitivity labels with Teams, groups, and sites[8]).

Teams shared channels

A team's shared channels automatically inherit the team's sensitivity label settings, and that label can't be removed or replaced with a different one (Use sensitivity labels with Teams, groups, and sites[8]). A container label can restrict which other teams can be invited to the team's shared channels with Internal only, Same label only or Private team only, subject to compatible privacy and external user access settings; only Private team only can remove previously invited teams, and none of the options affect invitations to individual users (Use sensitivity labels with Teams, groups, and sites[8]).

Sharing permissions and default sharing links (PowerShell only)

Site sharing settings and the default sharing link are configured only in PowerShell, with Set-Label -AdvancedSettings (Use sensitivity labels with Teams, groups, and sites[8]). The MembersCanShare advanced setting sets the site's sharing permissions (Use sensitivity labels with Teams, groups, and sites[8]):

Value Who can share
MemberShareAll Site owners and members can share files, folders and the site
MemberShareFileAndFolder Members and people with Edit permissions can share files and folders, but only owners can share the site
MemberShareNone Only site owners can share files, folders and the site

The default sharing link type is the scope and permission preselected when users share from SharePoint and OneDrive (Configure the default sharing link type[18]):

Advanced setting Values
DefaultSharingScope SpecificPeople, Organization or Anyone
DefaultShareLinkPermission View or Edit
DefaultShareLinkToExistingAccess True overrides both settings above and selects people with existing access

Users can always change these defaults before sending a link, so they set a baseline, not a prohibition; when broader sharing must be blocked, use the external sharing controls instead (Configure the default sharing link type[18]). A label with these settings can also be applied to individual documents, and when a site default and a document default differ, the more restrictive scope wins and brings its own permission: a site default of specific people with edit beats a document default of anybody in the organization with view, so the link is for specific people with edit (Configure the default sharing link type[18]). Site-level defaults require container labels to be enabled (Configure the default sharing link type[18]).

In summary, a container label is enabled through Microsoft Entra ID, governs membership, sharing and access conditions for the workspace, and leaves item protection to item labels.

1. Enable group labelsin Microsoft Entra ID2. Synchronize labelsExecute-AzureAdLabelSync3. Configure a labelwith Groups & sites4. Publish the labelto workspace owners
Container label enablement: Microsoft Entra group label support, label synchronization, a Groups & sites label, then publishing

Meetings, calendar items, and Teams channel meetings

A meeting label has to protect several items at once: the calendar event, the invitation and responses, related attachments and emails, and the Teams meeting itself, which is why the Meetings scope also needs Files & other data assets and Emails selected (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]).

Three protection settings, three jobs

On the Choose protection settings for labeled items page, a meeting label offers three settings (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]):

  • Apply or remove encryption encrypts meeting invites and responses and any Office attachment, so only authorized people can see the invite and access the meeting link; usage rights can further restrict it, for example by preventing forwarding.
  • Apply content marking adds headers or footers to meeting invites and responses.
  • Protect Teams meetings and chat displays the label in Teams meetings and enforces Teams-specific settings, including who can bypass the lobby, who can present, who can record and transcribe, encryption for meeting video and audio, automatic recording, video watermarks for screen sharing and camera streams, and preventing copying and forwarding of meeting chat, live captions and transcripts, among others.

Invite encryption doesn't configure any of the Teams options: Do Not Forward on an invitation restricts what recipients can do with the message, not who bypasses the lobby. The meeting inherits the invite's label and is protected even before it starts, and users see a lock where a label enforces a setting (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]). Some options can't be configured without a Teams Premium license, and if participants or organizers don't use a Teams client that meets the minimum versions, the label's meeting options can't be enforced for them (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]). Calendar items support only Azure Rights Management encryption (not S/MIME or Double Key Encryption), don't support automatic and recommended labeling, and require mailboxes in Exchange Online (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]).

Publishing policy settings for meetings

Meeting settings in the publishing policy are independent of the email settings: if you set a default label for emails but not for meetings and calendar events, no default label is applied to calendar events (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]). The settings are (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]):

  • Apply a default label to meetings and calendar events, and Require users to apply a label to their meetings and calendar events.
  • Inherit label from files shared to meetings. This publishing-policy setting for Teams meetings is separate from the content-based automatic and recommended labeling that calendar items don't support: Teams automatically applies or recommends the highest-priority label, published to the organizer, of files shared in the meeting chat, uploaded to the Shared tab, or presented with Live share; files shown by screen or window sharing don't count. A meeting that already has a higher-priority label keeps it, but a manually applied lower-priority label can be overridden, a deliberate delta from the replacement rule in Scopes, label groups, and label priority above.
  • Apply meeting label to artifacts (preview). Teams applies the meeting's label to recordings and their transcripts (.mp4 files) and to meeting notes (.loop files) at the time each artifact is saved to SharePoint or OneDrive. If the meeting label changes later, the artifact's label doesn't change automatically, and saved artifacts can be relabeled in SharePoint or OneDrive by someone with permission to change the label.

MP4 files are labeled only after you enable label support for them with Set-SPOTenant -EnableSensitivityLabelForVideoFiles $true (Enable sensitivity labels for files in SharePoint and OneDrive[13]). The two inheritance settings run in opposite directions: files shared into a meeting can raise the meeting's label, and the meeting's label flows out to the artifacts it produces.

Channel meetings

Channel meetings and chats get their label from the team's container label: edit an existing Groups & sites label and associate a meeting label whose scopes include Files & other data assets, Emails and Meetings, and all channel meetings and chats inherit it (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]). The association option appears only when you edit an existing label, not when you create one, and users can't change the inherited label (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]). If the associated label encrypts, it can use administrator-defined permissions or the Do Not Forward and Encrypt-Only options (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]). The chat-copy setting is then enforced for all channel chats, even outside channel meetings (Use sensitivity labels to protect calendar items, Teams meetings, and chat[19]).

What to remember: encryption protects the invitation, Protect Teams meetings and chat protects the meeting, the policy's inheritance settings move labels into and out of the meeting, and channel meetings take their label from the container label's association.

Sensitivity labels in Power BI

Power BI uses the same Microsoft Purview labels, but inside the Power BI service a label classifies content without controlling access to it.

Enabling labels for Power BI

A Fabric administrator enables Allow users to apply sensitivity labels for content in the tenant settings, for the whole organization or for specific security groups, with an optional list of excepted groups (Enable sensitivity labels in Fabric and Power BI[20]). Before that, labels must be defined and published to the relevant users and groups (Enable sensitivity labels in Fabric and Power BI[20]). Users need an Azure Information Protection Premium P1 or Premium P2 license to apply or view labels, and a Power BI Pro or Premium Per User (PPU) license in addition to apply labels to Power BI items (Enable sensitivity labels in Fabric and Power BI[20]). Users allowed by the tenant setting must also have Create and Edit permissions on any item they label; the tenant setting enables the capability but doesn't grant item permissions (Enable sensitivity labels in Fabric and Power BI[20]).

Labels classify in the service and protect on export

In the Power BI service, sensitivity labeling doesn't affect access to content: Power BI permissions alone manage access, and a label's encryption settings aren't applied while data stays in the service (Sensitivity labels in Power BI[21]). Users granted access to a report get access to the entire underlying semantic model unless row-level security (RLS) limits them, so limiting which rows a user sees is an RLS job, not a label job (Sensitivity labels in Power BI[21]).

Protection applies when data leaves the service through a supported export path (Sensitivity labels in Power BI[21]):

  • export to Excel, PDF files and PowerPoint (for PowerPoint, only with Export as image);
  • Analyze in Excel, which downloads an Excel file with a live connection to a semantic model;
  • Download the .pbix from the service.

Exports to .csv and any other unsupported path carry no label or protection (Sensitivity labels in Power BI[21]). In Power BI Desktop, by contrast, a label with encryption does affect access: a user without sufficient permissions under the label's encryption can't open the .pbix file (Sensitivity labels in Power BI[21]). Protection design therefore covers two controls: Power BI permissions and RLS in the service, and labels on what leaves it.

Inheritance follows lineage and keeps the stronger label

Power BI moves labels along content relationships (Sensitivity labels in Power BI[21]):

  • New reports and dashboards created in the service inherit the label of their parent semantic model or report.
  • A label applied in Desktop travels with a published or uploaded .pbix and is applied to both the resulting semantic model and report.
  • When you download a .pbix whose report and semantic model have different labels, the more restrictive label is applied to the file.
  • Analyze in Excel applies the semantic model's label and protection to the Excel file. If the model's label later becomes more restrictive, the Excel label updates on data refresh, but a manually set Excel label isn't overwritten, and an Excel file never inherits a less restrictive label.

In short, a Power BI label is a classification that follows the data and becomes protection only when the data leaves the service.

Sensitivity labels with Microsoft Defender for Cloud Apps

Microsoft Defender for Cloud Apps can apply Purview sensitivity labels to files in cloud apps in two ways, and they differ in which copy changes: labeling the stored file itself, or labeling only the copy a user downloads during a monitored session. Pick the approach by asking whether the original in the cloud app should carry the label.

Prerequisites and label discovery

The integration needs both a Defender for Cloud Apps license and a Microsoft Purview license, the Microsoft 365 app connector, and labels published as part of a publishing policy in Microsoft Purview (Integrate with Microsoft Purview for information protection[22]). A label is recognized only if its scopes include at least Files & other data assets and Emails, so a Groups & sites-only label is invisible to the integration (Integrate with Microsoft Purview for information protection[22]). Defender for Cloud Apps retrieves the list of the tenant's labels every hour (Integrate with Microsoft Purview for information protection[22]).

In the Microsoft Defender portal, the Microsoft Information Protection settings for Cloud Apps hold two choices: automatically scan new files for labels, and scan only labels from this tenant, which ignores labels set by other organizations (Integrate with Microsoft Purview for information protection[22]). Automatic scan discovers labels; it doesn't apply any.

Labeling the stored file

You can apply a label to one file from Cloud Apps > Files with Apply sensitivity label, or to matching files with a file policy whose governance action is Apply sensitivity label; either way the label is applied to the original file (Integrate with Microsoft Purview for information protection[22]). Supported file types are Word, Excel, PowerPoint and PDF, up to 30 MB, in Box, Google Workspace, SharePoint Online and OneDrive (Integrate with Microsoft Purview for information protection[22]). The documented limits (Integrate with Microsoft Purview for information protection[22]):

  • Empty files and files in libraries that require checkout aren't labeled, and labels on password-protected files can't be read.
  • Unprotected labels applied outside Defender for Cloud Apps can be overridden but not removed, and protected labels applied outside it can't be removed. These integration limits, not the Purview replacement rule from Scopes, label groups, and label priority above, decide what Defender for Cloud Apps overrides.
  • Labels that Defender for Cloud Apps applied itself aren't overridden by a later labeling action, so a second file policy can't overwrite the first one's label.
  • As a safety precaution, file policies allow 100 Apply label actions per app, per tenant, per day; labeling then pauses and continues automatically the next day after 12:00 UTC, and disabling a policy suspends all its pending labeling tasks.
  • For the service to read label information, the label's encryption permissions must be assigned to any authenticated user or to all users in your organization.

File policies in Defender for Cloud Apps retire on January 6, 2027, and Microsoft directs customers to recreate them as Microsoft Purview DLP or auto-labeling policies, with a migration wizard available from the retirement banner on the Policies page (Migrate file policies to Microsoft Purview[23]).

Labeling the downloaded copy with a session policy

A session policy with the Control file download (with inspection) session control type and the Protect action applies a label and its protections to a file as the user downloads it, while the original file stays as it was in the cloud app (Session policies[24]). Only labels configured to apply encryption appear as an option, supported types are Word, Excel, PowerPoint and PDF, and Protect doesn't overwrite a file that already has a label (Session policies[24]). The session must be routed through Conditional Access App Control by a Microsoft Entra Conditional Access policy, which requires a Defender for Cloud Apps license and a Microsoft Entra ID P1 license (Session policies[24]). Conditional Access App Control is what lets a session policy allow access while monitoring the session and limiting specific activities, such as downloading sensitive files (Session policies[24]). The figure below traces that path.

When two session policies conflict, the more restrictive policy wins, so a matching policy that blocks downloads prevents the download even when another policy would label it (Session policies[24]). The option Always apply the selected action even if the data cannot be scanned extends the action to content that can't be scanned; it doesn't add support for other file types (Session policies[24]).

Policies created for a host app aren't connected to its resource apps: a policy for Teams or Exchange doesn't cover files served by SharePoint or OneDrive, so create a separate policy for the resource app (Session policies[24]). To test, sign out of all existing sessions, reauthenticate to the apps as an in-scope user, perform the download, and then open the policy under Cloud Apps > Policies > Policy management to see its report of redirected sign-ins and file actions (Session policies[24]). A successful sign-in to the host app proves nothing until the report shows the protected download.

The choice comes down to the copy: file labeling changes the original in the cloud app, and session protection leaves the original alone and protects only what leaves through a monitored session.

1. User signs into the cloud app2. Session routed throughConditional AccessApp Control3. Download matchesthe session policy4. Protect action labelsand encrypts thedownloaded copyOriginal file in thecloud app is unchanged
Defender for Cloud Apps session protection: the Protect action labels and encrypts only the downloaded copy, leaving the cloud original unchanged

Removing a label from publishing versus deleting it

Retiring a label has two very different tools, and Microsoft describes removing a label from a publishing policy as less risky than deleting it (Create and publish sensitivity labels[5]).

Removing a label from publishing policies

When you remove a label from a publishing policy, users in that policy no longer see the label to select in their Office apps after the next policy refresh (Create and publish sensitivity labels[5]). Because labels combine across policies, as shown in Publishing policies: availability, defaults, and precedence above, remove it from every policy that offers it to those users. A label that is already applied isn't removed from the content or container, and an applied container label continues to protect the Teams or SharePoint site (Create and publish sensitivity labels[5]). You can add the label back to a policy later, and you can't delete a label while it is still in a publishing policy (Create and publish sensitivity labels[5]). A removal-type auto-labeling policy is not a way to retire a menu choice: it removes labels from matching files.

Deleting a label

Deleting a label has effects that differ by workload (Create and publish sensitivity labels[5]):

  • Labels that applied encryption. The protection template is archived, not destroyed, so previously protected content can still be opened. The archived template prevents creating a new label with the same name, and deleting the template with PowerShell risks losing access to content encrypted with it.
  • Documents in SharePoint and OneDrive with file-label integration. The label no longer displays, and if the deleted label applied encryption that the services can process, the encryption is removed; download, copy, move and opening in a desktop or mobile app have the same outcome.
  • Documents outside SharePoint and OneDrive, or without file-label integration, and emails. The label information stays in the metadata, but apps can't map the label ID to a name, so users don't see it; encryption remains, and users see the name and description of the archived protection template.
  • Containers. The label is removed and the settings it configured are no longer enforced, which typically takes 48 to 72 hours for SharePoint sites and can be quicker for Teams and Microsoft 365 groups.

Without the ID-to-name mapping, deleted labels can appear as GUIDs instead of names in Content explorer and activity explorer, the Purview views of labeled items and labeling activity (Create and publish sensitivity labels[5]). A familiar template name on an old file is therefore no proof that its label still exists. Microsoft notes that deleting labels is rare in production and more likely during initial testing (Create and publish sensitivity labels[5]).

The decision is simple: to take a label out of people's menus, unpublish it; delete only when you understand the archived template and the per-workload effects.

Four main ways to label content without waiting for the user

Labeling methodWhere it runsHow it chooses the labelCan it only recommend?Existing manual label
Publishing-policy defaultApps that support labels, on unlabeled itemsFixed label from the publishing policy, no content inspectionNoKept: a policy default never replaces an existing label
SharePoint library defaultOne SharePoint document library, new and edited files; a library-default auto-labeling policy type extends it to existing files at restLocation: the label configured for the library, no content inspectionNoKept
Client-side auto-labelingSupported Office apps while users edit or composeSensitive information types or trainable classifiers in the label settingYesKept
Auto-labeling policyService side: SharePoint and OneDrive files at rest, Exchange mail in transitContent-based policies: rules with sensitive information types, trainable classifiers and other conditions; the library-default policy type inspects no contentNo: simulation replaces recommendationsKept by default; a lower-priority manual label is replaced only with the policy option

Decision tree

Should detected contentchoose the label?NoYesBaseline set by aSharePoint library?Must it run in theservice, unattended?YesNoYesNoSharePointlibrarydefaultAuto-labelingpolicyPublishing-policydefaultClient-sideauto-labeling

Sharp facts the exam loves — give these one last read before exam day.

Cheat sheet

Sharp facts the exam loves — scan these before test day.

Sensitivity label administration can be delegated without Global Administrator

Use a Purview role group containing Sensitivity Label Administrator for custom delegation of label and label-policy management, or use an appropriate built-in group such as Information Protection Admins. For read-only configuration access, use Sensitivity Label Reader. These are administrative permissions; users applying a published label in an app do not need the label-administration role.

Trap Assigning Content Explorer Content Viewer to a label-policy author, because permission to inspect classified documents does not confer label configuration rights.

3 questions test this
Administrative units can bound both delegated label administration and policy eligibility

Assign Microsoft Entra administrative units to individual members of applicable Purview role groups to restrict the populations they manage. When creating label or auto-labeling policies, select administrative units to constrain eligible Exchange and OneDrive users or SharePoint sites. Refine user and group scope as needed, but SharePoint policies apply to all sites associated with the selected units and do not allow individual site inclusions or exclusions. Keep unit membership accurate because it determines the effective boundary even when the policy selects all users or sites within that scope.

Trap Using a sensitivity label's encryption recipient list to limit which division an administrator may manage, because content usage rights do not scope administrative authority.

3 questions test this
An item carries one organizational sensitivity label as persistent metadata

A supported item can carry one sensitivity label from your organization, stored in clear-text metadata that travels with the content. Supported applications use that identity to display the classification and enforce the configured settings even when the file changes location. A sensitivity label can coexist with a retention label because protection classification and retention lifecycle are separate dimensions.

2 questions test this
A label's selected scopes determine where it can be used and which settings it can configure

Select Files & other data assets for files and Power BI content, Emails for mail, and Groups & sites for supported containers. A scope that is not selected leaves its settings unavailable and prevents the label from being offered for that resource type. Meetings also requires both Files & other data assets and Emails, so an administrator must satisfy the scope dependency before configuring meeting protection.

Trap Republishing a Groups & sites-only label to an Outlook user to make it available for email, because audience changes do not add the Emails scope.

3 questions test this
A label grouping organizes choices rather than supplying inherited protection settings

Use label groups in the modern scheme, or parent labels and sublabels in the classic scheme, to organize related choices in a two-tier menu. The child labels do not inherit the parent's protection settings; configure each applicable label independently, with color as the documented inheritance exception. A classic parent with published sublabels cannot itself be applied or used as a default, and modern label groups cannot be published as labels. For classic sublabels, select and publish both the parent label and the sublabel. In the modern scheme, label display names must be unique within each group and among standalone labels, so moving a label into a group fails if another label there has the same display name.

Trap Using encryption configured only on the Confidential parent to protect its departmental sublabels, because the sublabels do not acquire those encryption settings.

3 questions test this
Meeting labels enforce Teams options separately from encrypting the invitation

Select Protect Teams meetings and chat on a meeting-capable label to enforce lobby access, presenter and recording permissions, audio/video encryption, automatic recording, video watermarks, and chat-copy restrictions. Invite encryption separately restricts who can read the invitation and meeting link; it does not configure those meeting options. Teams-specific protection depends on Teams Premium licensing and supported Teams clients, because the configured options cannot be enforced for users on unsupported clients. Calendar items require Exchange Online mailboxes and do not support ordinary automatic or recommended labeling, so use manual selection or the meeting-specific default and inheritance settings instead.

Trap Using Do Not Forward on the invitation as the lobby-admission control, because invitation usage rights restrict message reuse rather than deciding who bypasses the Teams lobby.

3 questions test this
Sensitivity-label scripts manage label definitions separately from publishing policies

Use New-Label and Set-Label for sensitivity-label definitions, and New-LabelPolicy and Set-LabelPolicy for their publishing policies in Security & Compliance PowerShell. Retrieve Guid, Name, DisplayName, and ContentType with Get-Label and target the GUID when modifying a label, because display names need not be unique. For multilingual deployments, supply translated DisplayName and Tooltip entries through Set-Label -LocaleSettings rather than expecting the Office display language to translate administrator-authored text. These operations let scripts change a label's identity-specific settings without confusing the label object with the policy that distributes it.

Trap Using Set-LabelPolicy to translate an individual label's display name, because localized DisplayName and Tooltip entries belong to the label's LocaleSettings rather than its publishing policy.

3 questions test this
Sensitivity label encryption can use centrally assigned permissions or user-selected permissions

Choose Assign permissions now when every item in the category should use administrator-defined readers and usage rights. Choose Let users assign permissions when collaborators differ per item; supported Office apps can prompt for recipients, while Outlook can use Do Not Forward or Encrypt-Only. Rights-management encryption requires an activated Azure Rights Management service, and the chosen client must support the configured permission mode. A tenant supports at most 500 labels that specify encryption users and permissions even though the overall catalog can exceed 1,000 labels, so keep the user-facing taxonomy small rather than designing toward the catalog maximum.

Trap Creating a new centrally permissioned label for every ad hoc contract recipient combination, because a user-permission mode can express changing collaborators without multiplying labels.

4 questions test this
A label's encryption recipients determine content access independently of who can apply the label

Configure the encryption audience explicitly, using supported users or groups, organizational membership, authenticated users, or an external organization's domain as appropriate. Everyone in your organization excludes guest accounts, while Any authenticated users permits a broad audience subject to the configured usage rights. Specifying a domain belonging to a Microsoft Entra tenant includes that tenant's verified domains, and publishing exclusions or non-email-enabled security groups do not become equivalent Azure Rights Management access controls.

Trap Excluding a guest from the label publishing policy to revoke access to an already encrypted file, because publication eligibility does not remove rights granted by the protection configuration.

4 questions test this
Encrypted-content permissions distinguish editing from exporting or removing protection

Grant the usage rights needed for the intended workflow rather than treating all document access as Full Control. DOCEDIT allows content changes, EDIT permits saving in protected formats, and EXPORT or Full Control supports operations that can remove existing encryption. Authorized Azure Rights Management super users provide a separate recovery and inspection path when ordinary recipients lack those rights.

Trap Assigning a tenant label-administration role as the way to decrypt every protected document, because management permission is not itself a content usage right or super-user grant.

3 questions test this
Content expiration and offline access impose different limits on encrypted content

User access to content expires determines when recipients should stop opening an item, while Allow offline access determines how long a cached use license can be used before reauthentication. Labels with Assign permissions now can set offline access to never, always, or a limited number of days; user-assigned permissions use the tenant use-license validity, which defaults to 30 days. Previously authorized readers can retain cached rights until their use license expires, so select the offline interval with access-change responsiveness in mind.

Trap Setting only a document expiration date to require authentication on every opening, because lifetime expiry does not define the use-license renewal interval.

3 questions test this
Double Key Encryption reserves decryption control for data that can forgo cloud processing

Use Double Key Encryption for highly sensitive content requiring an organization-controlled key in addition to the Microsoft-held key; access to both keys is required to decrypt. Deploy the organization's DKE service before configuring the label, and plan to use supported Windows clients for consumption. SharePoint and OneDrive can store DKE files but cannot process them for Office for the web, coauthoring, search, DLP, or eDiscovery, so ordinary rights-management encryption is the better fit when those cloud capabilities are required. Bring Your Own Key (BYOK) instead provides customer-managed key provisioning and lifecycle for Azure Rights Management and requires an Azure subscription; it does not introduce DKE's separate organizational key service.

Trap Using BYOK alone when Microsoft must be unable to decrypt with its service-side key material, because BYOK does not add the separate organization-controlled DKE service key.

3 questions test this
Changing a published label's encryption does not affect every previously labeled item in the same way

Adding encryption to a published label protects newly labeled items after the update reaches clients. Existing files in SharePoint and OneDrive with labeling integration enabled acquire the new encryption status when next accessed, but other previously labeled documents and emails generally need the label removed and reapplied to change encryption status. Changes to administrator-assigned users and rights take effect at authentication, subject to the offline use-license boundary; changing user-assigned permission mode applies to newly labeled or relabeled items.

Trap Recreating the publishing policy to force all offline desktop files to adopt newly added encryption immediately, because policy distribution does not rewrite each existing file's protection.

3 questions test this
SharePoint file-label integration enables service processing of supported encrypted content

Enable SharePoint and OneDrive file-label integration with the Purview portal's Turn on now banner or Set-SPOTenant -EnableAIPIntegration $true; Multi-Geo deployments require running PowerShell for each geo. This allows the services to process supported label-encrypted files that use a cloud-based key for Office for the web, search, eDiscovery, and DLP, rather than treating the contents as unreadable. Existing encrypted files must be edited or downloaded and uploaded again to gain these processing capabilities. File integration is separate from Microsoft Entra container-label enablement and from the Co-authoring for files with sensitivity labels setting used for encrypted-file collaboration in desktop and mobile Office apps.

Trap Using Execute-AzureAdLabelSync to make encrypted cloud documents searchable, because synchronizing container labels does not enable SharePoint's processing of file contents.

3 questions test this
Extended SharePoint protection keeps downloaded documents tied to the original item's permissions

Select Extend protection on download, copy or move for a SharePoint library to protect downloaded unprotected documents with the original SharePoint permissions rather than an independent fixed recipient list. Enable Set-SPOTenant -ExtendPermissionsToUnprotectedFiles $true, file-label integration, and encrypted-file coauthoring, then select a published file label using Let users assign permissions with the Office permission-prompt option. Ordinary recipients must connect to the original SharePoint item to open the downloaded file, so offline access is unavailable and removal of their source access or deletion of the source prevents opening. These labeled files cannot be moved or copied to another site, and moving or copying between libraries in the same site requires permission to create or delete lists.

Trap Using an ordinary administrator-permissioned library default to track later SharePoint access changes on downloads, because its fixed usage-right grants do not perform this original-item permission check.

3 questions test this
Document tracking and revocation target a registered ContentID rather than every copy of a file

Track and revoke a protected Office document by its registered ContentID; supported Office apps register local label-protected documents when opened, and password-protected documents are unsupported. After Connect-AipService, use Get-AipServiceDocumentLog to find the ContentID by filename or protection owner and Get-AipServiceTrackingLog -ContentId to inspect successful and denied access. Revoke with Set-AipServiceDocumentRevoked -ContentId -IssuerName, or undo a mistaken revocation with Clear-AipServiceDocumentRevoked using the same values; cached offline rights can continue until their allowed period expires. For documents protected with administrator-defined permissions, SharePoint or OneDrive uploads remove ContentID and downloads receive a new one, so revoking the original identity does not revoke downloaded copies.

Trap Using an Advanced Message Encryption email-revocation command for a registered local Office document, because that workflow targets eligible encrypted messages rather than a document ContentID.

3 questions test this
Sensitivity label content markings depend on item type and application support

Office content markings can add headers, footers, and document watermarks, with variables such as the label or document name. Email and meeting invites support headers and footers but not watermarks. Watermarks allow 255 characters and headers or footers generally allow 1,024, while Excel has a combined 255-character header-and-footer limit that includes formatting codes; service-side auto-labeling does not itself insert these visual markings into documents.

Trap Using encryption alone to meet a requirement for visible handling instructions on printed documents, because restricting access does not insert those instructions into the content.

3 questions test this
Dynamic watermarks personalize the displayed document and require encrypted labels

Dynamic watermarks insert the viewing user's UPN by default and can include customized text, date, and time. They require label encryption, persist in print but not export, and Microsoft recommends setting offline access to Never. Except for documented exceptions such as the Rights Management owner, Outlook attachment preview, and PowerPoint Live, unsupported Office clients deny access rather than silently omitting the watermark, so verify client capability before deployment.

Trap Using a static Confidential watermark for a requirement to identify which reader captured a page, because fixed content marking does not personalize itself for the current viewer.

2 questions test this
A sensitivity label must be published to become available to its intended users

Create and configure a sensitivity label first, then include it in a label policy targeted to users or supported groups. One label can be reused in several policies, allowing a pilot audience and later organization-wide publication without duplicating the label. Editing an already published label does not require a new policy; allow up to 24 hours for changes to replicate across apps and services. The publishing policy can also supply a custom help-page URL, displayed as Learn More after the available labels in Office apps, to explain label choices beyond individual label tooltips.

Trap Creating an auto-labeling policy solely to populate a user's Sensitivity menu, because menu availability comes from a publishing policy.

4 questions test this
Default labels provide a starting classification while mandatory labeling requires coverage

A publishing policy can assign separate defaults for supported item types such as documents, email, meetings, containers, and Power BI content. Mandatory labeling requires an applicable label before the supported save, send, or creation operation; the requirement can be satisfied manually, automatically, or by a default. Combine these settings when coverage is required with fewer prompts, while allowing users to choose the classification that accurately fits the content.

Trap Using mandatory labeling alone to guarantee that sensitive documents receive the most restrictive category, because requiring a label does not determine which classification is correct.

3 questions test this
Users receive labels from all applicable policies and conflicting settings from the highest-priority policy

Sensitivity label policies are ordered with the lowest priority at the top and the highest order number at the bottom. A user in multiple policies receives their combined available labels, while each conflicting setting comes from the highest-priority applicable policy. Reorder the policies when a departmental default or mandatory-label setting must override the organization-wide baseline.

Trap Removing a baseline label from a departmental policy to hide it from a user who still receives the baseline policy, because label availability is combined across applicable policies.

3 questions test this
Downgrade justification follows label priority but does not apply to every label change

Configure the label policy to require a justification when users remove a label or replace it with a lower-priority label on files, email, or meetings. Put more restrictive labels lower in the label list so the priority represents the intended sensitivity order. The setting does not apply to container labels or changes between sublabels under the same parent or label group; Office prompts once per app session, while the Information Protection client prompts for each file.

Trap Using a publishing policy's priority number as the sensitivity ranking for downgrade detection, because policy precedence and the labels' own sensitivity order are different lists.

3 questions test this
Unpublishing a sensitivity label withdraws future selection without removing applied labels

Remove a sensitivity label from every publishing policy that offers it to the intended users to withdraw it from their selection menus after policy refresh. Labels already applied to content or containers remain, and an applied container label continues protecting its Teams workspace or SharePoint site. Publication can be restored by adding the label back to a policy; deletion is a separate operation and is blocked while the label remains in any label policy.

Trap Using a SharePoint auto-labeling removal policy to retire a menu choice, because that action changes labels on matching files rather than withdrawing publication to users.

3 questions test this
Meeting inheritance uses separate settings for shared files and saved artifacts

Configure Inherit label from files shared to meetings separately from Apply meeting label to artifacts, because they control opposite directions of label inheritance. The first setting automatically applies or recommends the highest-priority supported file label published to the organizer for files shared through meeting chat, the Shared tab, or Live share, not screen or window sharing; it can replace a lower manual meeting label but preserves a higher label. The artifact setting applies the meeting label to recordings and transcripts in MP4 files and notes in Loop files when saved to SharePoint or OneDrive, with MP4 labeling support enabled through Set-SPOTenant -EnableSensitivityLabelForVideoFiles $true. Changing the meeting label afterward does not relabel saved artifacts, which require a separate permitted file-label change.

Trap Using the shared-file inheritance option to relabel recordings saved before a meeting's classification changed, because artifact labeling has its own setting and occurs when each artifact is saved.

3 questions test this
Outlook can derive an email label from eligible physical attachments

Enable Email inherits highest priority label from attachments in the publishing policy's Default settings for emails to select the highest-priority supported attachment label published to the sender. Attachments must be physical files rather than SharePoint or OneDrive links, and the label must support both files and emails. Inheritance preserves manually labeled email and higher-priority email labels, but can replace lower-priority automatic or default labels. Select Recommend users apply the attachments label instead of automatically applying it when users should confirm the change; without that option, inheritance applies the selected label and its configured encryption automatically.

Trap Assuming a pasted OneDrive sharing link will trigger attachment-label inheritance, because this Outlook policy evaluates physical attachments rather than cloud-file references.

3 questions test this
Deleting a sensitivity label changes protection differently across workloads

Deleting an encrypting sensitivity label archives its protection template so previously protected content can still be opened; it does not destroy the template. The archived template prevents creating a new label with the same name; deleting that template risks losing access to content encrypted with it. In SharePoint and OneDrive with file-label integration enabled, opening or exporting a document removes the deleted label's encryption if the services can process it, whereas email and documents outside that integration retain encryption. Deleting a container label removes it and ends enforcement of its configured settings. Label metadata can remain in files after deletion, but losing the label-ID-to-name mapping hides the applied name and can make deleted labels appear as GUIDs in Content explorer or Activity explorer.

Trap Using a familiar protection-template name as proof that a deleted sensitivity label still exists, because archived encryption templates can outlive their label definitions.

3 questions test this
Client-side label conditions classify content while users edit documents or compose email

Configure Auto-labeling for files and emails on the label to evaluate sensitive information types or trainable classifiers in supported Office clients. Select automatic application or recommendation according to whether users should see a proposed label, and refine SIT conditions with instance counts, confidence, and all-versus-any logic. This path can label a document before it is saved and depends on the Office app's feature support; a service-side auto-labeling policy is the separate choice for unattended processing.

Trap Configuring only a publishing-policy default label for a requirement to recommend classifications based on detected payroll content, because a default is not content-dependent.

4 questions test this
Office auto-labeling conditions using EDM must also include a non-EDM sensitive information type

When configuring a label's client-side sensitive-information conditions with EDM, also specify at least one non-EDM type, such as the corresponding built-in identifier type. An EDM-only configuration turns off that label's auto-labeling setting. Require both types in an all-of group when only reference-confirmed records should trigger labeling; an any-of grouping would also permit the generic type to trigger without EDM confirmation.

4 questions test this
Automatic label selection uses sensitivity order with special client-side sibling behavior

When multiple labels qualify, selection normally favors the highest-priority label, which is the one with the highest order number. For an unlabeled file matching sibling sublabels in Office, automatic application is preferred to recommendation before comparing sibling order. If the file already has a sublabel under that same parent, client-side labeling keeps it; service-side policy evaluation instead selects the highest-order qualifying sibling.

Trap Moving a recommended sibling below an automatic sibling to guarantee that Office chooses the recommendation, because automatic-versus-recommended behavior is evaluated before that sibling ordering.

3 questions test this
Service-side auto-labeling processes cloud files at rest and Exchange mail in transit

Use an auto-labeling policy for saved SharePoint or OneDrive files and email processed by Exchange Online, without depending on an installed Office client. Exchange policy evaluation does not retroactively process mail already at rest in mailboxes. Supported Office and PDF attachments can supply detection evidence, but the policy labels the email rather than assigning the same label to its attachment; encryption inheritance is a separate behavior. For SharePoint and OneDrive, simulation and label application preserve existing modified, modified-by, and date metadata. When the label encrypts a cloud file, its Rights Management issuer and owner are the account that last modified the file.

Trap Using an Exchange auto-labeling policy to relabel all historical mailbox messages, because its Exchange evaluation point is mail transport rather than mailbox-at-rest scanning.

4 questions test this
Auto-labeling policies must run simulation and be eligible to turn on before they can apply or remove labels

Run at least one simulation before turning on a service-side auto-labeling policy, and review its matches before enforcement; simulation records expected outcomes without changing labels, can take about 12 hours, and blocks activation if more than 20 million files match, requiring a narrower policy and another run. Before submitting or rerunning simulation through PowerShell, connect with Connect-IPPSSession -EnableSearchOnlySession so the simulation job can be submitted correctly. Results describe one policy and available classification at the time, so assess overlapping active policies and actual enforcement separately; configured activity alerts can still fire during simulation. When selecting Run policy in simulation mode, optionally enable automatic turn-on if the policy is not edited for seven days, rather than requiring a separate manual activation. Duplicating a policy creates a new policy in simulation regardless of the source policy's active state, so the copy must pass its own simulation gate before enforcement.

Trap Assuming a simulation match count proves successful relabeling of all matching files, because evaluation and label-writing enforcement are separate stages.

4 questions test this
Service-side labeling requires eligible labels, supported files, and realistic processing scope

The selected label must be published to at least one user; SharePoint and OneDrive encryption requires Assign permissions now with content expiration set to Never, whereas Exchange-only policies also allow Let users assign permissions with Do Not Forward or Encrypt-Only but cannot auto-apply S/MIME-protecting labels. Supported files include DOCX, PPTX, XLSX, and PDFs after tenant PDF support is enabled, while open or checked-out files are ineligible during processing. Plan within 100 policies and a tenant-wide ceiling of 500,000 automatically labeled files per day, not a promised rate; explicit portal targeting allows 1,000 SharePoint sites or 100 OneDrive users or groups, while All is exempt from those selection caps. SharePoint list-item attachments are unsupported, and OneDrive accounts must be associated with their corresponding users for both simulation and auto-labeling. Label-removal policies do not support password-protected files or files protected with external or non-Microsoft encryption, even when simulation identifies matching content.

Trap Using a label that prompts the user applying the label to choose collaborators for unattended SharePoint auto-labeling, because the service needs administrator-defined permissions and nonexpiring content access.

3 questions test this
Service-side policies distinguish label replacement from explicit label removal

Choose an apply-label policy to assign or upgrade classification, or the separate SharePoint and OneDrive removal action to clear a selected label, including a manual label, and its label-applied encryption. Automatic application normally preserves manual labels and higher-priority labels while replacing lower-priority automatic or default labels; service-side replacement of a lower-priority manual label requires the explicit policy option. When matching active policies conflict, application wins over removal even if a different label is applied. A matching apply-label policy without the manual-replacement option can also prevent another policy from replacing a manual label.

Trap Raising a label's position alone to ensure that all manually labeled files are replaced by an auto-labeling policy, because replacing manual classifications requires the additional policy option.

3 questions test this
A library-default auto-labeling policy can extend SharePoint defaults to existing files

Use the auto-labeling policy option that applies SharePoint document libraries' default sensitivity labels to eligible existing files at rest, which the ordinary new-or-edited-file default can leave untouched. This policy selects each library's configured default rather than inspecting content to choose a classification, so it supplies a location-defined baseline rather than content-dependent differentiation. Select specific SharePoint sites rather than All, without administrative-unit scoping; only one such policy is supported per tenant, with up to 10 sites selected at a time. After labeling completes and results are verified, edit that policy to replace the selection with the next batch of up to 10 sites. If a site cannot be added, distinguish a missing library default, which needs configuration, from an inability to verify defaults, which requires checking access to site and library metadata.

Trap Using the general 1,000-site auto-labeling selection allowance for a library-default backfill policy, because this policy type accepts only 10 sites at a time in its single tenant-wide policy.

4 questions test this
Auto-labeling rollout requires licensing and permissions beyond manual labeling

Verify an eligible E5 or P2-tier auto-labeling entitlement rather than assuming that a license supporting manual labeling also permits automatic or recommended labeling. Service-side simulation requires Microsoft 365 auditing, and reading matched source contents additionally requires Data Classification Content Viewer, which Global Administrator does not receive by default. Turning a successfully simulated policy on requires Compliance Administrator or Compliance Data Administrator; content-preview permission is not policy-activation authority. If the auto-labeling surface is absent, verify regional availability as well as the license and administrative assignment before changing the label's detection conditions.

Trap Assigning only Data Classification Content Viewer to an operator who must activate a simulated policy, because source-content visibility does not grant the administrative permission to turn enforcement on.

3 questions test this
Advanced auto-labeling rules expose workload-specific conditions and exceptions

Select Common rules for a shared rule definition across selected workloads, or Advanced rules when Exchange, SharePoint, and OneDrive need different conditions or exceptions. Use Content is shared to distinguish only with people inside my organization from with people outside my organization, and supported attachment or document name, file-extension, and size conditions to refine which items qualify without changing the content detector. Exchange advanced rules can evaluate subjects, headers, senders, and recipients, allowing a message-context requirement without relying only on sensitive-information counts. SharePoint and OneDrive rules can use document properties, but Document property is compares managed-property values as exact strings rather than regular expressions; mapped values containing an ampersand can fail even when the strings agree, so use a consistent alternative classification value without ampersands in the document metadata and rule, then verify simulation matches and actual labeling. For Exchange, Sender domain is accepts up to 600 domain entries in that condition, while FromAddressMatchesPatterns supports at most 10 regular expressions of up to 128 characters each; these are condition- or parameter-specific limits, not combined policy-wide allowances.

Trap Using a regular expression in Document property is to match several SharePoint metadata values, because that managed-property comparison requires exact strings rather than regex matching.

3 questions test this
Exchange auto-labeling location scope follows senders rather than recipient membership

Keep Exchange at All included and None excluded when an auto-labeling policy must evaluate incoming messages from external senders, and narrow internal recipients with advanced rule conditions instead. Selecting specific included users or groups exempts external senders, while exclusions exempt messages those users send rather than messages they receive. To encrypt matching externally sent messages, also enable Apply encryption to email received from outside your organization and assign a Rights Management owner. That owner must be one internal user identified by an organization-owned email address, not a mail contact, shared mailbox, or group.

Trap Including only the Finance group in Exchange location scope to encrypt external mail addressed to Finance, because explicit location inclusions select senders and exempt messages sent from outside the organization.

3 questions test this
An encrypting auto-label overrides competing Exchange rule-based IRM settings

When an Exchange auto-labeling policy and a mail flow or DLP rule both match, the applied label's encryption takes precedence over the rule's IRM encryption settings. If the applied label does not encrypt, the rule's IRM encryption can be applied alongside that classification label instead. Determine whether the selected label itself supplies encryption before predicting the protection outcome, rather than assuming that every matching rights template accumulates on the message.

Trap Assuming a mail flow rule's broader rights template remains effective beneath a stricter encrypting auto-label, because the matching label's encryption displaces the rule's IRM settings rather than layering the grants.

3 questions test this
PowerShell auto-labeling can target adaptive SharePoint membership instead of enumerated sites

Associate an existing SharePoint adaptive scope through New-AutoSensitivityLabelPolicy -SharePointAdaptiveScopes when sites must be targeted dynamically or exceed the portal's 1,000 explicitly selected sites. This replaces static enumeration with adaptive membership rather than increasing the portal selection limit. Create the policy and its conditions as separate objects with New-AutoSensitivityLabelPolicy and New-AutoSensitivityLabelRule -Policy.

Trap Using repeated portal site selections to configure a single policy for more than 1,000 explicit SharePoint sites, because the portal selection cap still applies instead of being replaced by adaptive-scope membership.

3 questions test this
File labeling progress and Exchange labeling verification require different evidence

Use policy-level progress, Labeled items, Labeling failures, and coverage reports to verify SharePoint and OneDrive file enforcement, not Exchange mail; progress updates every 48 hours, so a newly enabled policy can initially show zero. Distinguish successful labeling from Pending, Not yet attempted, and Failed to apply label outcomes; absence from the failure list does not establish success. Coverage by simulation context reports labeling activity after the latest simulation and is available for 30 days, so its counts can differ from all-time labeled totals; rerun simulation to obtain a new coverage context when that window expires. Review each file's Failure reason to distinguish automatically retried transient infrastructure failures from format, existing-protection, or label-configuration failures needing reason-specific administrative remediation, while recognizing that some unsupported files require no action. For recent Exchange results, filter Activity explorer for Sensitivity label applied and check How applied for automatic application, recognizing that the event does not identify the specific auto-labeling policy or rule.

Trap Using an Exchange-only simulation rerun to populate file coverage reports, because those enforcement and coverage views report SharePoint and OneDrive files rather than email.

2 questions test this
A SharePoint library default can upgrade automatic classifications but preserves manual labels

An ordinary SharePoint library default replaces lower-priority automatic or publishing-policy default labels, but preserves manual labels and higher-priority classifications. Configure it as a site administrator in Library settings using a published Files & other data assets label, with file-label integration enabled and library IRM disabled. If the default label encrypts, use administrator-defined permissions with content expiration set to Never, not user-defined permissions or Double Key Encryption. A publishing-policy default never replaces an existing label, and the separate Extend protection on download, copy or move option must remain unselected for this ordinary library-default scenario.

Trap Raising a user's publishing-policy default to overwrite an existing automatic label in a library, because a publishing-policy default does not perform the library default's priority-based replacement.

3 questions test this
Comparable auto-labeling simulations require preserved results and complete workload evidence

Export the simulation results needed for comparison and check the exported file before editing the policy or restarting simulation, because sample retention does not guarantee earlier-run availability. Simulation displays up to 100 items per site and CSV exports have a 50,000-record ceiling, so displayed or exported rows need not enumerate every policy match. A failed simulation or a workload marked Not available is missing evidence rather than a completed zero-match test, and a partial match count includes only available workload results. For comparable Exchange runs, send and receive the same representative messages during each simulation instead of assuming that earlier test traffic will be evaluated again. Restarting simulation retries evaluation without guaranteeing a fix, so inspect the new status and source results before deciding that testing succeeded.

Trap Assuming retained simulation samples guarantee recovery of earlier-run exports after policy editing, because sample retention and prior-run result availability are separate boundaries.

3 questions test this
Recreating an auto-labeling rule requires confirmation of the specific rule-load failure

Contact Microsoft Support when an enabled Exchange auto-labeling policy stops matching without corresponding item-level failures, and recreate a rule only after Support identifies it as failing to load. A rule-load failure prevents evaluation against that rule, unlike a file-level failure after a match, so an empty labeling-failure list does not establish that rule evaluation is healthy. Preserve the affected rule's policy, name, target label, workload and scope, conditions, exceptions, priority, and enabled state, then delete and recreate only that confirmed rule with the same configuration. Plan for interrupted enforcement from that rule during repair and propagation, rerun simulation with newly sent or received Exchange messages, and turn on the policy only after validation before verifying that matching resumes.

Trap Using bulk deletion and recreation of unrelated policy rules to troubleshoot missing Exchange matches, because deleting each rule removes its enforcement without identifying the rule-load fault.

3 questions test this
Converting label settings creates a service-side policy from sensitive-information-type conditions only

Use the conversion option at the end of label creation or editing to populate an auto-labeling policy from the label's sensitive-information-type conditions, not its trainable classifiers. Classifier-only labels do not offer conversion, and mixed conditions transfer only the sensitive-information-type portion, so review the resulting detection logic before saving. The generated policy includes all SharePoint, OneDrive, and Exchange locations by default and starts in simulation when saved; edit its scope and conditions when those defaults do not match the intended rollout. Conversion does not check whether file-label integration is enabled for SharePoint and OneDrive, so verify that prerequisite separately before expecting those locations to apply labels.

Trap Assuming conversion preserves a mixed label's trainable-classifier criteria, because the generated service-side policy copies only its sensitive-information-type conditions.

3 questions test this
Container label settings require Microsoft Entra enablement and label synchronization

Enable sensitivity-label support for Microsoft 365 groups in Microsoft Entra ID, then run Execute-AzureAdLabelSync in Security & Compliance PowerShell to synchronize the labels. Select Groups & sites on the label and publish it to users who need to choose it for supported workspaces. Group owners can change labels for group-connected sites, while a non-group-connected SharePoint site requires a site admin.

Trap Enabling sensitivity labels for Office files in SharePoint as the only prerequisite for labeling Microsoft 365 groups, because file integration and group-label support are separate configurations.

2 questions test this
Container labels govern workspace access without assigning labels to the files inside

A container sensitivity label can set privacy, guest admission, and SharePoint external-sharing controls, but documents inside do not inherit its item-level label or encryption. Public and Private privacy values set and lock the workspace choice, while None leaves owners able to choose it. The guest-access setting controls adding guests rather than automatically removing existing guests, so review existing membership separately when tightening admission.

Trap Using a private Teams label as the mechanism to encrypt every document uploaded by team members, because workspace access controls do not assign file-level protection.

2 questions test this
Container labels can invoke Conditional Access but do not replace its required configuration

For labeled SharePoint sites, choose an unmanaged-device restriction or an existing published authentication context when access must depend on device or authentication conditions. The unmanaged-device option requires the documented app-enforced-restrictions policy and cannot relax a stricter tenant setting. Authentication contexts must be configured and published in Microsoft Entra Conditional Access, and clients that do not support the selected context are denied access.

Trap Assigning a more permissive site label to bypass the tenant's unmanaged-device restriction for one project, because the tenant-level restriction remains the effective lower bound.

2 questions test this
Teams shared channels inherit the parent label and expose team-invitation controls

A shared channel inherits the sensitivity label settings of its parent team and cannot replace or remove that inherited label. Container labels can constrain invitations to other teams with Internal only, Same label only, or Private team only controls, subject to compatible privacy and guest settings. These controls govern team invitations rather than invitations to individual users, and only Private team only can potentially remove previously invited teams.

Trap Applying a separate more permissive sensitivity label to one shared channel to allow a partner team, because that channel's label is inherited from its parent team.

2 questions test this
Container labels can standardize who may share a site and its content

Use the MembersCanShare advanced setting on a sensitivity label to control sharing permissions for the labeled SharePoint site. MemberShareAll allows the broader owner-and-member sharing model, MemberShareFileAndFolder reserves site sharing for owners, and MemberShareNone reserves sharing of files, folders, and the site for owners. Configure this with Set-Label -AdvancedSettings and distinguish it from external-sharing audience controls or the default sharing-link type.

Trap Changing the default sharing-link type to restrict the ability of members to share the site, because a default link choice does not set the sharing-permission model.

2 questions test this
Channel meeting protection requires an associated meeting label on the container label

Edit an existing Groups & sites label and associate a configured meeting-protection label to make channel meetings and chats inherit its meeting controls. The association option is available when editing rather than initially creating the label, and the required label scopes include files, emails, and meetings. Users cannot change the associated label inherited by channel meetings and chats. This association supplies meeting and chat protection rather than merely relying on the parent team's container privacy settings.

Trap Reaching for a standalone label on each channel meeting to enforce the channel's baseline, because channel meeting protection is inherited through the container label's configured association.

2 questions test this

Use Set-Label -AdvancedSettings to configure DefaultSharingScope as SpecificPeople, Organization, or Anyone and DefaultShareLinkPermission as View or Edit for supported labeled sites or documents. Setting DefaultShareLinkToExistingAccess to True overrides both settings and selects people with existing access instead. Users can change the proposed link settings before sharing, so use access restrictions rather than these defaults when broader sharing must be prohibited. When site and document defaults conflict, the more restrictive audience scope wins, with the permissions attached to that selected default rather than independently combining the most restrictive permission.

Trap Assuming a document's View default always overrides a site's Edit default, because a narrower site audience can select the site's default together with its Edit permission.

2 questions test this
Power BI labeling requires more than publishing a Purview label to a user

In Fabric tenant settings, enable Allow users to apply sensitivity labels for content and scope it to the intended organization or security groups. Users also need the labels published to them, appropriate information-protection licensing, Power BI Pro or Premium Per User for labeling Power BI items, and Create and Edit permissions on the item. The tenant setting permits labeling capability but does not supply missing item permissions.

Trap Granting only access to the Fabric tenant-labeling setting to a report viewer who must relabel content, because the item still requires Create and Edit rights.

3 questions test this
Power BI service permissions govern access while sensitivity labels protect supported exported files

Within the Power BI service, sensitivity labels classify content but do not replace Power BI access permissions or row-level security. Supported exports to Excel, PDF, PowerPoint as an image, and downloaded PBIX files carry the label and its configured file encryption; CSV and other unsupported paths do not. Power BI Desktop enforces encryption on a protected PBIX, so consider both in-service authorization and allowed export paths when designing protection.

Trap Assigning a restrictive sensitivity label to a semantic model as the way to limit which rows a service user can view, because service-level row filtering requires the relevant Power BI security controls.

3 questions test this
Power BI label inheritance follows content relationships and preserves stronger classifications

New reports and dashboards inherit the label of their parent semantic model or report, and a label on a published PBIX travels to the resulting service report and model. Downloading a PBIX when the service report and model have different labels selects the more restrictive one. Analyze in Excel also carries the model label and protection to the workbook; refresh can upgrade an inherited label but does not silently replace a manually applied label or downgrade a stronger Excel classification.

3 questions test this
Defender for Cloud Apps label integration requires licensing, connectors, and suitable published labels

Provide both Defender for Cloud Apps and Microsoft Purview licensing, enable the Microsoft 365 app connector, and publish labels whose scope includes files and emails. Enable automatic label scanning in the Microsoft Information Protection section of Cloud Apps settings in the Microsoft Defender portal. Defender for Cloud Apps retrieves the tenant's label list hourly, and can be configured to ignore externally defined labels when only the organization's classifications should be considered.

Trap Adding a Groups & sites-only label to a publishing policy for cloud-file governance, because the integration's label recognition requires the item scopes.

2 questions test this
Defender for Cloud Apps can label supported originals directly or through file-policy governance

For an individual file, use Cloud Apps > Files and Apply sensitivity label; for matching populations, configure a file policy with the sensitivity-label governance action. Applying labels is supported for suitable Office and PDF files up to 30 MB in Box, Google Workspace, SharePoint Online, and OneDrive. Empty files and libraries requiring checkout cannot be labeled, and password-protected files do not expose readable labels to the service. File-policy application has a safety limit of 100 Apply label actions per app per tenant per day, after which labeling pauses and resumes automatically the next day after 12:00 UTC. Disabling a policy suspends all its pending labeling tasks rather than accelerating the remaining work.

Trap Using the automatic label-discovery checkbox alone to assign a new classification to every matching file, because discovery does not configure a label-application governance action.

3 questions test this
Defender for Cloud Apps label remediation depends on where protection was applied

Defender for Cloud Apps can override unprotected labels applied outside the service but cannot remove them, and it cannot remove protected labels applied outside the service. It also does not override labels it already applied itself. Where protected content needs inspection, configure the documented permission to inspect protected files and choose remediation compatible with the label's origin rather than assuming every label can be replaced or cleared.

Trap Using a second Defender for Cloud Apps file policy to overwrite a label assigned by the first one, because labels already applied by that service are not overridden by its subsequent labeling action.

3 questions test this
Session-policy protection labels the downloaded copy without changing the cloud original

Use the Protect action in a Defender for Cloud Apps session policy with Control file download (with inspection) to label and encrypt a downloaded copy in an eligible Word, Excel, PowerPoint, or PDF format while leaving the cloud original unchanged. Only sensitivity labels configured to apply encryption are offered for this action, and Protect does not overwrite an existing label. The app session must be routed through Conditional Access App Control, with Defender for Cloud Apps and Microsoft Entra ID P1 licensing and the required Purview labeling entitlement in place. When matching session policies conflict, the more restrictive policy wins, so a matching Block prevents download even when another policy would protect it. Select Always apply the selected action even if the data cannot be scanned when the chosen action must also apply to unscannable content, without treating that setting as support for labeling arbitrary file types.

Trap Using a file-policy Apply sensitivity label governance action when only downloaded copies should change, because file-policy governance applies the label to the original cloud file.

3 questions test this
Session-labeling validation must exercise the app that actually serves the file

Target the resource app that serves a download separately when needed, because a session policy for a host such as Teams or Exchange does not automatically cover SharePoint or OneDrive. Sign out of existing sessions, reauthenticate to each configured app as an in-scope user, and perform the intended file download. In the Defender portal, open Cloud Apps > Policies > Policy management and select the session policy to check its report for redirected sign-ins and the resulting file action. Confirm both the controlled session and download outcome rather than treating successful host-app sign-in as proof that file protection ran.

Trap Assuming a Teams-targeted session policy controls a file served by SharePoint, because the host app and its resource app require their own applicable policy coverage.

3 questions test this

References

  1. https://learn.microsoft.com/en-us/purview/sensitivity-labels
  2. https://learn.microsoft.com/en-us/credentials/certifications/resources/study-guides/sc-401
  3. https://learn.microsoft.com/en-us/purview/apply-sensitivity-label-automatically
  4. https://learn.microsoft.com/en-us/purview/get-started-with-sensitivity-labels
  5. https://learn.microsoft.com/en-us/purview/create-sensitivity-labels
  6. https://learn.microsoft.com/en-us/purview/purview-admin-units
  7. https://learn.microsoft.com/en-us/purview/sensitivity-labels-office-apps
  8. https://learn.microsoft.com/en-us/purview/sensitivity-labels-teams-groups-sites
  9. https://learn.microsoft.com/en-us/purview/encryption-sensitivity-labels
  10. https://learn.microsoft.com/en-us/purview/rights-management-usage-rights
  11. https://learn.microsoft.com/en-us/purview/double-key-encryption
  12. https://learn.microsoft.com/en-us/purview/track-and-revoke-admin
  13. https://learn.microsoft.com/en-us/purview/sensitivity-labels-sharepoint-onedrive-files
  14. https://learn.microsoft.com/en-us/purview/sensitivity-labels-sharepoint-default-label
  15. https://learn.microsoft.com/en-us/purview/sensitivity-labels-sharepoint-extend-permissions
  16. https://learn.microsoft.com/en-us/purview/purview-adaptive-scopes
  17. https://learn.microsoft.com/en-us/purview/auto-label-simulation-results
  18. https://learn.microsoft.com/en-us/purview/sensitivity-labels-default-sharing-link
  19. https://learn.microsoft.com/en-us/purview/sensitivity-labels-meetings
  20. https://learn.microsoft.com/en-us/power-bi/admin/service-security-enable-data-sensitivity-labels
  21. https://learn.microsoft.com/en-us/power-bi/enterprise/service-security-sensitivity-label-overview
  22. https://learn.microsoft.com/en-us/defender-cloud-apps/azip-integration
  23. https://learn.microsoft.com/en-us/defender-cloud-apps/migrate-file-policies-to-purview
  24. https://learn.microsoft.com/en-us/defender-cloud-apps/session-policy-aad