AB-650 Administering Microsoft 365 and AI Services Questions and Answers

Microsoft Defender for Office 365 processes email protection policies according to a defined precedence order. When Standard or Strict preset security policies are enabled, they take precedence over custom anti-spam policies.

The general processing order is Strict preset security policy, Standard preset security policy, applicable evaluation policies, custom policies, and finally default protection.

Although Policy1 has priority 0, this value establishes its precedence among custom anti-spam policies only. It cannot override the Standard preset security policy.

Option A is correct because excluding Group1 from Standard protection prevents the Standard preset security policy from applying to the group ' s members. These users can then be evaluated by Policy1, provided its recipient conditions match.

Administrators can configure this exception through the Microsoft Defender portal or the corresponding preset policy rule settings.

Option B is incorrect because disabling Standard protection would affect all recipients, violating the requirement to preserve Standard protection for other users.

Option C is incorrect because changing Policy1 to priority 1 lowers its precedence among custom policies and does not supersede Standard protection.

Option D is incorrect because adding exceptions to Policy1 does not change the precedence of Standard protection.

Therefore, Group1 must be excluded from the Standard preset security policy.

Microsoft References:

https://learn.microsoft.com/en-us/defender-office-365/preset-security-policies

https://learn.microsoft.com/en-us/defender-office-365/how-policies-and-protections-are-combined

Answer Area

Correct Selection

Users

Admin1 only

Portal

The Microsoft 365 Apps admin center

Microsoft 365 provides the Allow web search in Copilot policy to control whether Microsoft 365 Copilot and Microsoft 365 Copilot Chat can retrieve information from public web sources.

First selection: Admin1 only.

According to Microsoft Learn, the Cloud Policy service supports administration by Global Administrators, Office Apps Administrators, and Security Administrators.

In the Contoso scenario:

Admin1: Global Administrator — authorized to manage Cloud Policy.

Admin2: AI Administrator — can manage Copilot settings, but this role is not listed among those authorized to manage Cloud Policy configurations.

Admin3: Privileged Role Administrator — does not have the required Cloud Policy management permissions.

User1: Eligible for AI Administrator, but eligibility does not grant active administrative permissions.

Therefore, Admin1 is the only administrator explicitly authorized for this configuration.

Second selection: The Microsoft 365 Apps admin center.

Administrators configure web search through Microsoft 365 Apps admin center → Customization → Policy Management.

Create a Cloud Policy configuration and select the following setting:

Allow web search in Copilot

Configure it as:

Disabled in Microsoft Copilot and Microsoft Copilot Chat.

This disables public web search for both experiences, satisfying Contoso ' s requirement to use only Microsoft 365 information sources for Copilot responses.

Microsoft confirms that the Microsoft 365 admin center provides a shortcut to this setting, but the Cloud Policy itself is configured in the Microsoft 365 Apps admin center.

Microsoft References:

Microsoft Learn — Cloud Policy service and supported administrator roles .

Microsoft Learn — Manage Copilot settings .

Microsoft Learn — Manage web search for Copilot and Copilot Chat .

Answer Area

Correct Selection

Users that can approve the request

Admin1 and Admin3 only

User1 will be an AI administrator for

24 hours

Microsoft Entra Privileged Identity Management (PIM) enables eligible users to activate privileged administrator roles temporarily, subject to approval and activation-duration restrictions.

First selection: Admin1 and Admin3 only.

The Contoso case study specifies that approval is required to activate the AI Administrator role, but the Approvers setting is None.

According to Microsoft documentation, when no specific approvers are configured, active Global Administrators and Privileged Role Administrators become the default approvers.

Therefore:

Admin1: Global Administrator — can approve.

Admin2: AI Administrator — cannot approve merely by holding this role.

Admin3: Privileged Role Administrator — can approve.

Second selection: 24 hours.

The PIM configuration specifies an Activation maximum duration of 24 hours.

This means User1 can activate the AI Administrator role for a maximum of 24 hours following approval. After the activation period expires, PIM automatically removes the active privileges.

The one-month eligible assignment expiration controls how long User1 remains eligible to request activation. The 15-day active assignment expiration applies to separately assigned time-bound active roles, not this eligible role activation.

User1 can request a shorter activation duration, but 24 hours is the maximum permitted by the configuration.

Microsoft References:

Microsoft Learn — Configure Microsoft Entra PIM role settings .

Microsoft Learn — Activate Microsoft Entra roles in PIM .

Microsoft Foundry agents consume Azure resources, including AI model inference, token processing, and potentially additional services such as Code Interpreter, file search, and storage.

The Contoso case study specifies the following technical requirement:

" The total costs of Agent1 must be monitored and evaluated monthly. "

Option A is correct because Azure Cost Management + Billing provides comprehensive cost monitoring, analysis, and reporting for Azure resources, including resources consumed by Microsoft Foundry agents.

Administrators can navigate to Azure portal → Cost Management + Billing → Cost Management → Cost analysis.

Cost analysis supports monthly reporting, historical comparisons, cost breakdowns, filtering by resources or resource groups, and budget monitoring.

Option B is incorrect because billed messages in the Power Platform admin center relate to Power Platform and Copilot Studio consumption rather than comprehensive Microsoft Foundry agent costs.

Option C is incorrect because Microsoft 365 billing policies manage supported consumption-based Microsoft 365 Copilot services, not the complete Azure infrastructure costs associated with Foundry agents.

Option D is not the best answer because the Microsoft Foundry portal ' s Monitor experience provides valuable agent-level operational metrics and estimated token costs, but it does not necessarily capture every billable Azure resource used by an agent.

Microsoft explicitly recommends Azure Cost Management for comprehensive spending analysis and financial reconciliation.

Therefore, Cost Management + Billing in the Azure portal is the most appropriate solution for monitoring and evaluating Agent1 ' s total monthly costs.

Microsoft References:

Microsoft Learn — Plan and manage costs for Microsoft Foundry .

Microsoft Learn

Microsoft Learn — Monitor Microsoft Foundry costs using Azure Cost Management

Answer Area

Correct Selection

Project documents

Sensitive information type (SIT)

Proposal documents

Exact data match (EDM)

Microsoft Purview provides different classification technologies for identifying sensitive information. The appropriate method depends on whether the information follows a predictable pattern or must match known values.

First selection: Sensitive information type (SIT).

Contoso ' s project documents contain project codes consisting of the letters PR, followed by a hyphen and nine digits, such as PR-123456789.

This predictable structure makes a custom sensitive information type appropriate. Administrators can create a SIT using a regular expression such as:

PR-[0-9]{9}

Supporting keywords and confidence requirements can further minimize false positives.

Second selection: Exact data match (EDM).

Contoso ' s proposal documents contain customer IDs consisting of seven to ten alphanumeric characters. The valid customer IDs are recorded in an Excel workbook.

EDM is appropriate because it compares detected values against a known reference dataset. Administrators can prepare the customer IDs for EDM classification using a supported data format.

EDM reduces false positives by identifying customer IDs that actually exist in the reference data rather than matching arbitrary alphanumeric strings.

A fingerprint SIT would be more appropriate for Contoso ' s feedback forms, which use a standardized template. Trainable classifiers are better suited to content requiring contextual or semantic classification.

Therefore, a custom SIT for project codes and EDM for known customer IDs satisfy both requirements.

Microsoft References:

Microsoft Learn — Create custom sensitive information types .

Microsoft Learn — Exact Data Match sensitive information types .

Microsoft Learn — Reduce false positives using classifiers

Microsoft Entra Conditional Access integrates with Microsoft Entra ID Protection to evaluate risky authentication attempts and enforce additional security requirements based on detected sign-in risk.

The Contoso technical requirement states that Group3 users accessing Microsoft 365 resources from anonymous IP addresses must complete multifactor authentication (MFA).

First selection: Sign-in risk.

Microsoft Entra ID Protection recognizes sign-ins from anonymous IP addresses, including anonymous VPNs and Tor exit nodes, as potential security risks.

Administrators can configure the Sign-in risk condition in CAPolicy1 to evaluate authentication attempts based on their calculated risk level. Microsoft recommends considering medium and high sign-in risk for MFA enforcement, depending on the organization ' s security requirements.

Second selection: Grant.

The Grant access control determines which authentication requirements must be satisfied before access is permitted.

Administrators should select Grant access and configure the built-in Multifactor authentication strength, or the corresponding Require multifactor authentication control where available.

Because CAPolicy1 already targets Group3 and all resources, these selections apply risk-based MFA requirements to the intended users and Microsoft 365 resources.

Technical qualification: Sign-in risk policies operate on calculated risk levels, not exclusively on anonymous IP addresses. Consequently, they may also trigger for other risky sign-ins and cannot guarantee that every anonymous-IP connection is challenged.

The Contoso case study specifies the following planned change:

Add an additional email address alias for the users in Group4.

Group4 is a Microsoft 365 group containing User1 and User2. Both users have Microsoft Exchange mailboxes.

Option C is correct because the Exchange Online PowerShell Set-Mailbox cmdlet supports modifying mailbox properties, including additional SMTP email addresses.

Administrators can use the -EmailAddresses parameter with the @{Add=...} syntax to add a secondary email address without removing existing addresses or modifying the primary SMTP address.

For example:

Set-Mailbox -Identity " user1@contoso.com " `

-EmailAddresses @{Add= " user1.alias@contoso.com " }

The same operation can be performed for the other Group4 members.

Option A is incorrect because Set-MailUser manages mail-enabled users who have external mailboxes, rather than Exchange Online user mailboxes.

Option B is incorrect because Update-MgUser manages Microsoft Entra user properties and is not the appropriate cmdlet for modifying Exchange Online mailbox proxy addresses.

Option D is incorrect because Set-ADUser manages on-premises Active Directory user objects. Contoso ' s users are cloud-only, and their Exchange Online mailboxes are managed in Microsoft 365.

Therefore, Set-Mailbox is the appropriate cmdlet for implementing the planned change.

Microsoft References:

Microsoft Learn — Add or remove email addresses for a mailbox in Exchange Online .

Microsoft Learn — Set-Mailbox PowerShell reference .

Microsoft Entra administrative units support dynamic device membership based on device attributes. AU1 uses the following membership rule:

(device.deviceTrustType -eq " Workplace " ) -and

(device.deviceOSType -in [ " Windows " , " Android " ])

Statement 1: No.

Admin1 holds the Global Administrator role. However, AU1 uses dynamic membership, which prevents administrators from manually adding or removing members. Membership is determined exclusively by the configured rule.

Additionally, Device1 is Microsoft Entra joined, corresponding to deviceTrustType = AzureAD, rather than Workplace. Therefore, Device1 does not satisfy the rule.

Statement 2: Yes.

Device2 runs Windows 11 and is Microsoft Entra registered.

Microsoft Entra registered devices have a deviceTrustType value of Workplace, and Windows matches the permitted operating system list. Therefore, Device2 satisfies both conditions and is automatically included in AU1.

Statement 3: Yes.

The planned Android devices must be Microsoft Entra registered. Registered Android devices satisfy the Workplace trust condition and, when their recorded operating system value is Android, also satisfy the OS condition.

Consequently, they are automatically added to AU1 through dynamic membership processing. Group1 membership itself does not determine AU1 membership.

Administrators should verify the actual deviceOSType value because some Android Enterprise registrations use more specific operating system identifiers.

Microsoft References:

Manage dynamic membership in administrative units .

Microsoft Entra dynamic membership rules and device attributes .

First dropdown: Get-MgGroupMember; Second dropdown: Update-MgUser; Third dropdown: -UserId

Microsoft Graph PowerShell provides cmdlets for managing Microsoft Entra groups and modifying user account attributes. This task requires retrieving members of an existing group and updating the EmployeeType property for each user.

First selection: Get-MgGroupMember. This cmdlet retrieves direct members of the specified Microsoft Entra group. The -GroupId parameter identifies the Contractors group using its unique identifier.

Second selection: Update-MgUser. This cmdlet updates Microsoft Entra user properties through Microsoft Graph. It supports the -EmployeeType parameter, which can be used to assign the value Part-time.

Third selection: -UserId. This parameter identifies the specific user object to modify. The foreach statement processes each returned member and applies the corresponding update.

Technical accuracy note: The video uses $person.ObjectId. Current Microsoft Graph PowerShell documentation normally exposes the member identifier as $person.Id. Additionally, the -All switch should be specified when retrieval of every group member is required across result pages.

A production-oriented implementation would therefore use the following structure, assuming the group contains user objects:

$members = Get-MgGroupMember -GroupId " < GroupId > " -All

foreach ($person in $members) { Update-MgUser -UserId $person.Id -EmployeeType " Part-time " }

The three dropdown answers remain correct; this implementation note distinguishes the original examination script from current Microsoft Graph behavior.

Microsoft References:

https://learn.microsoft.com/en-us/powershell/module/microsoft.graph.groups/get-mggroupmember

https://learn.microsoft.com/en-us/powershell/module/microsoft.graph.users/update-mguser

Microsoft Purview Data Security Posture Management (DSPM) provides centralized visibility into sensitive information and its use across Microsoft 365 workloads, including interactions with Microsoft 365 Copilot.

DSPM includes reporting and investigation capabilities that help administrators understand how users interact with organizational information through AI applications.

The AI activity exploration capabilities provide information about Copilot interactions, including associated users, referenced files, sensitivity labels, and potentially sensitive information.

Because the financial files are classified with the Confidential sensitivity label, administrators can investigate activity involving labeled content and determine which users referenced sensitive organizational files during Copilot interactions.

DSPM minimizes administrative effort by providing predefined dashboards, reports, and integrated investigation workflows instead of requiring administrators to construct individual audit searches.

Option A is incorrect because Microsoft Entra audit logs primarily record identity and directory administrative activities, rather than providing detailed Copilot file-reference investigations.

Option C is incorrect because Content explorer identifies and examines classified organizational content. It is not the primary solution for correlating Copilot interactions with users and referenced files.

Option D is incorrect because Data Lifecycle Management manages retention and deletion requirements rather than providing dedicated AI interaction monitoring.

Therefore, Data Security Posture Management is the appropriate solution.

Microsoft References:

https://learn.microsoft.com/en-us/purview/data-security-posture-management-learn-about

https://learn.microsoft.com/en-us/purview/ai-m365-copilot

Microsoft Entra ID supports the ForceDelete operation for removing custom domains that still have references to directory objects. This operation automatically updates supported references, reducing the need for manual administrative changes.

Option C is correct because the domain meets the main ForceDelete prerequisites: it uses managed authentication, is not the default domain, and has fewer than 1,000 references.

However, Microsoft explicitly requires administrators to update or remove domain references from Exchange-managed objects, including mail-enabled security groups and distribution lists, before executing ForceDelete.

After these references are removed, ForceDelete can automatically update eligible user principal names, email addresses, proxy addresses, group addresses, and application identifier URIs to the tenant ' s initial onmicrosoft.com domain.

Option A is incorrect because manually updating all 750 references creates unnecessary administrative effort.

Option B is incorrect because ForceDelete cannot automatically remove all Exchange-managed dependencies. The mail-enabled security group references must be corrected first.

Option D is incorrect because a default domain cannot be deleted while it remains designated as the default.

Therefore, clearing the Exchange-managed references and using ForceDelete is the most efficient supported approach.

Microsoft References:

Microsoft Learn — Manage custom domain names and ForceDelete .

Microsoft Graph — domain: forceDelete .

Microsoft Defender for Office 365 supports outbound spam policies that control sending limits and automatically restrict accounts exhibiting suspicious outbound email activity.

Option A is correct because a custom outbound anti-spam policy can target Group1 specifically, allowing administrators to enforce more restrictive sending limits for marketing department mailboxes without changing the organization-wide configuration.

Administrators can navigate to Microsoft Defender portal → Email & collaboration → Policies & rules → Threat policies → Anti-spam, and create a custom outbound policy.

The policy should include Group1, configure the appropriate external recipient limit, and select Restrict the user from sending mail as the action when a limit is reached.

Option B is incorrect because modifying the default outbound policy would affect other users not covered by a higher-priority custom outbound policy.

Option C is incorrect because Strict preset security policies do not contain configurable outbound spam policies.

Option D is incorrect because Standard preset protection cannot provide a custom outbound sending limit.

Option E is incorrect because inbound anti-spam policies protect incoming email rather than restricting compromised users ' outgoing messages.

Crucially, Microsoft ' s documentation confirms that Strict and Standard preset security policies do not override outbound spam policies. Therefore, the custom outbound policy can apply to Group1 without excluding its members from Strict protection.

Microsoft References:

Microsoft Learn — Configure outbound spam policies .

Microsoft Learn — Preset security policies .

Use: The Microsoft 365 Apps admin center; To configure: Policy Management

Microsoft 365 administrators can control whether Microsoft Copilot uses public web information to generate responses through Microsoft 365 Cloud Policy.

First selection: The Microsoft 365 Apps admin center.

The Microsoft 365 Apps admin center provides access to Cloud Policy, which centrally manages settings for Microsoft 365 applications and supported Copilot experiences.

Administrators can access these settings through config.office.com.

Second selection: Policy Management.

Within the Microsoft 365 Apps admin center, administrators navigate to Customization and Policy Management. They can create or update a policy configuration containing the Allow web search in Copilot setting.

Microsoft documents a configuration that disables web search in Copilot Work mode while preserving web functionality in Copilot Web mode and Copilot Chat.

This distinguishes work-grounded AI interactions from the general web-based experience.

If administrators disable web search entirely, the restriction applies more broadly and does not satisfy the requirement to target only the work-grounded experience.

The Microsoft Purview portal governs information protection, compliance, and data security. It is not the primary interface for this Cloud Policy setting.

Therefore, the Microsoft 365 Apps admin center and Policy Management are the correct selections.

Technical note: Microsoft distinguishes Work mode from Copilot Chat in its current documentation. The relevant setting is Disabled in Microsoft Copilot Work mode; Enabled in Microsoft Copilot Web mode and Microsoft Copilot Chat. This is the documented configuration corresponding most closely to the question ' s intended work-only restriction.

Microsoft References:

https://learn.microsoft.com/en-us/microsoft-365/copilot/manage-public-web-access

Microsoft Purview auto-labeling policies automatically apply sensitivity labels to supported SharePoint and OneDrive documents when configured sensitive information conditions are satisfied.

AutoLabel1 requires the IP Address sensitive information type at High confidence, with a minimum instance count of two.

Statement 1: No.

File1.docx contains only one IP address. Although .docx is supported for automatic sensitivity labeling, the file does not meet the minimum requirement of two matching instances. Therefore, Sensitivity1 is not applied.

Statement 2: No.

File2.txt contains two IP addresses, satisfying the numerical threshold. However, SharePoint auto-labeling policies do not support .txt files. Microsoft documents supported file formats as modern Word, Excel, PowerPoint, and eligible PDF documents. Consequently, Sensitivity1 is not applied to File2.txt.

Statement 3: Yes.

File3.xlsx contains five IP addresses, exceeding the minimum instance count of two. The .xlsx format is supported by SharePoint auto-labeling policies. Assuming at least two IP addresses are detected at the configured High confidence level, the file satisfies the rule and receives Sensitivity1 once the policy processes it.

Thus, file format support and sensitive information matching must both be considered when evaluating auto-labeling behavior.

Microsoft Purview Data Loss Prevention (DLP) policies support different actions depending on the Microsoft 365 workload to which a policy is applied. Administrators must select a supported location before configuring a particular DLP action.

Option B is correct because the Set headers action is supported specifically for Exchange Online email.

This action allows administrators to add or modify an email message header when the message matches the conditions of a DLP rule. For example, an organization can configure a rule that detects sensitive financial information and adds a custom header indicating that the email requires additional handling.

Microsoft documents Set headers as a non-halting action, meaning it does not automatically stop subsequent DLP rule processing.

Option A is incorrect because OneDrive policies govern files and sharing activities rather than Exchange transport message headers.

Option C is incorrect because SharePoint DLP policies primarily protect stored documents and control access to sensitive content.

Option D is incorrect because Teams chat and channel messages do not expose the Exchange-specific Set headers action.

Therefore, the DLP policy must include the Exchange email location to support Set headers.

Microsoft Reference: Microsoft Learn — Data Loss Prevention policy reference .

Answer Area

Correct Selection

Site1

Restricted Content Discovery

Site2

Restricted Access Control

Microsoft SharePoint provides Restricted Content Discovery (RCD) and Restricted Access Control (RAC) to manage content discoverability and user authorization. Although both strengthen information governance, they serve different purposes.

First selection: Restricted Content Discovery.

Site1 requires administrators to prevent Microsoft 365 Copilot from discovering SharePoint content without modifying existing user permissions.

Restricted Content Discovery limits a site ' s visibility in organization-wide search and Microsoft 365 Copilot experiences. Users who already possess SharePoint permissions can continue accessing content directly.

Administrators can enable it through SharePoint admin center → Sites → Active sites → Site1 → Settings → Restrict content from Microsoft Copilot.

Second selection: Restricted Access Control.

Site2 requires an additional authorization boundary so that users must both belong to Group1 and possess existing SharePoint permissions.

Restricted Access Control enforces these two requirements simultaneously. Users outside Group1 are denied access even if they have direct permissions or sharing links.

Administrators can configure RAC through SharePoint admin center → Sites → Active sites → Site2 → Settings → Restricted site access, selecting Group1 as the permitted security group.

Microsoft 365 Archive manages inactive content, while organization-level sharing settings do not provide the site-specific authorization required.

Important distinction: RCD is not a security boundary and retains some discovery exceptions for previously accessed or owned content. RAC enforces access restrictions.

Microsoft References:

Microsoft Learn — Restricted Content Discovery .

Microsoft Learn — Restricted Access Control

Microsoft Purview Information Protection provides data classification, sensitivity labeling, and protection capabilities for organizational information. However, Information Protection alone does not provide the dedicated investigation workflow required to identify individual AI agent interactions containing confidential information in prompts or responses.

The more appropriate solution is Microsoft Purview Data Security Posture Management (DSPM), using its AI observability and Activity explorer capabilities.

Microsoft DSPM supports visibility into AI agent activities, including Microsoft Agent 365. It can identify interactions involving sensitive information through integrated data classification, audit telemetry, sensitive information types, and trainable classifiers.

Administrators can navigate to the Microsoft Purview portal, open DSPM, and use Discover → Activity explorer → AI activities to investigate AI interactions. The AI observability page also provides agent-specific visibility into potentially sensitive interactions and associated risks.

Microsoft Purview Information Protection contributes classification and labeling capabilities to this process, but simply using Information Protection does not independently deliver the required agent-interaction investigation.

Therefore, the proposed solution does not fully satisfy the requirement, making Option B (No) the appropriate answer.

Microsoft References:

https://learn.microsoft.com/en-us/purview/ai-agent-365

https://learn.microsoft.com/en-us/purview/data-security-posture-management-learn-about