Skip to content
All projects
Compliance · AI GovernanceProduction

Microsoft Purview DSPM for AI (Classic)

Discovering, auditing and governing Microsoft 365 Copilot and generative AI activity, with auto-provisioned Data Loss Prevention and Insider Risk Management policies.

Microsoft PurviewDSPM for AIMicrosoft 365 CopilotPurview AuditData Loss PreventionInsider Risk Management

Difficulty

Advanced

Environment

Microsoft 365 / Purview

Deployment

Production

Est. Implementation

45–60 minutes

Related certifications:SC-400SC-300SC-900
GitHub
Overview

Executive Summary

This project documents an end-to-end configuration of Microsoft Purview Data Security Posture Management (DSPM) for AI (Classic) — the control plane Microsoft 365 tenants use to discover, audit and govern how users and agents interact with Microsoft 365 Copilot and other generative AI applications.

The lab completes the mandatory onboarding checklist, validates live AI interaction and sensitive-information-type telemetry in Activity Explorer, reviews risk-focused Reports, and confirms the Data Loss Prevention and Insider Risk Management policies that extending AI data discovery insights auto-provisions, closing with a governance maturity scorecard from Purview’s Recommendations engine.

Context

Business Problem

The organization had no way to see what sensitive data — PII, financial data, national identifiers — was being shared with Microsoft 365 Copilot, Copilot agents, or external generative AI applications accessed through the browser.

Microsoft 365 Copilot and browser-based AI tools sit directly on top of the same SharePoint, Exchange and Teams content stores that traditional DLP programs were built to protect, and prompts submitted to those tools cannot be retrieved once sent — a governance blind spot that grows with every stage of Copilot adoption.

Business Risks

  • Sensitive identifiers (Aadhaar numbers, Japanese My Number data and other regulated PII) shared with Copilot without any monitoring
  • No structured signal for unauthorized or unethical AI interactions
  • Governance and audit gaps as Copilot usage scales across the tenant

Compliance Concerns

  • SC-400 / SC-300 aligned data security and insider risk scenarios
  • Zero Trust audit-first governance posture
Approach

Solution Overview

DSPM for AI (Classic) was activated from the Purview Solutions catalog and the four-step onboarding checklist was completed in sequence — activating Microsoft Purview Audit, installing the Purview browser extension, onboarding devices, and extending insights for data discovery.

Extending insights auto-provisioned three governance policies across Data Loss Prevention and Insider Risk Management in a single action, while Activity Explorer and the Reports dashboard gave a unified view of Copilot interactions, sensitive information type matches and a dedicated "unethical AI interaction" risk signal.

Design

Architecture

Architecture Diagram
100%

Architecture diagram placeholder

Users interacting with Microsoft 365 Copilot and browser-based AI tools feed Purview Audit telemetry into DSPM for AI, which aggregates activity into Activity Explorer and Reports and auto-provisions enforcement policies in Data Loss Prevention and Insider Risk Management.
Stack

Technology Stack

Microsoft Purview DSPM for AI (Classic)

Discovery, telemetry and governance control plane for AI activity

Microsoft Purview Audit

Foundational telemetry layer feeding Copilot interaction data

Purview browser extension & device onboarding

Extends visibility to AI sites reached through Edge, Chrome and Firefox

Activity Explorer

Event-level detail on AI interactions and sensitive info type matches

Data Loss Prevention / Insider Risk Management

Auto-provisioned enforcement policies triggered by extending AI insights

Environment

Lab Environment

Portal
Microsoft Purview compliance portal (purview.microsoft.com)
Tenant admin account
admin365lab
Test user
testuser1@securem365lsb...
AI app in scope
Microsoft 365 Copilot (Copilot Chat, WebChat surface)
Lab date range observed
30 June 2026 – 7 July 2026
Build

Implementation

A phased, expandable walkthrough. Screenshots and evidence can be attached to each phase.

  1. Locate DSPM for AI (Classic) in the Purview Solutions catalog and review the four-step "Get started" onboarding checklist.

    • Select DSPM for AI (classic) from the Purview Solutions catalog
    • Review the Overview dashboard and the four required onboarding steps
    • Confirm Purview Audit, the browser extension and device onboarding show completed
    Screenshots for this phase can be added here.
Automation

PowerShell

Illustrative example snippets — copy or download each script. Production values are placeholders.

Confirm Microsoft Purview Audit status

Get-PurviewAuditStatus.ps1
# Illustrative scaffold — DSPM for AI has a hard dependency on Purview Audit being active.
Connect-IPPSSession -UserPrincipalName admin@yourtenant.onmicrosoft.com

# Confirm unified audit logging is enabled tenant-wide
Get-AdminAuditLogConfig | Select-Object UnifiedAuditLogIngestionEnabled
Evidence

Configuration Screenshots

Click any tile to open the lightbox. Real screenshots will replace these placeholders.

Proof

Validation

  • Onboarding checklist completed

    Purview Audit, browser extension and device onboarding confirmed complete before extending insights — Pass

  • Telemetry validated

    Activity Explorer returned 9 events across AI Interaction and Sensitive info types activity types — Pass

  • Policies auto-provisioned

    Extend Insights panel confirmed policies created; independently verified Status: On for all three in the Policies list — Pass

  • Sensitive interactions confirmed

    Reports Data section showed 12 sensitive interactions and 1 unethical interaction against Microsoft 365 Copilot — Pass

Reality

Challenges

Sequenced onboarding dependency

DSPM for AI cannot surface any Copilot interaction telemetry until Purview Audit is active first — the onboarding checklist must be completed in order, not in parallel.

One action, multiple solution areas

"Extend insights for data discovery" provisions policies across Data Loss Prevention and Insider Risk Management simultaneously, so a single DSPM for AI action has a footprint that other solution owners need to review.

Insight

Lessons Learned

  • Visibility precedes governance — meaningful AI activity data only became visible after the first three onboarding steps were completed.
  • A single "Extend insights for data discovery" action provisioned policies across three different Purview solution areas at once, which matters for change management.
  • The Recommendations completion count is a leading indicator of governance maturity, not a lagging measure of actual protection already in place.
  • DSPM for AI (Classic) is scheduled for retirement on 30 September 2026, with all classic data carrying forward into the new DSPM experience — a migration every Microsoft 365 security team using it will need to plan for.
Outcome

Business Impact

12

Sensitive Interactions Detected

3

Policies Auto-Provisioned

9 Activity Events

Telemetry Validated

1 of 14 Recommendations

Governance Maturity

Capability

Skills Demonstrated

Microsoft Purview DSPM for AI ConfigurationMicrosoft 365 Copilot Activity MonitoringActivity Explorer & Reports AnalysisData Loss Prevention Policy ReviewInsider Risk Management Policy ReviewAI Governance Maturity Tracking
Mapping

Related Certifications

Certifications this project maps to are highlighted.

MS-102SC-200SC-300SC-400AZ-900AZ-500SC-100
Reading

Related Blog Articles

Placeholder cards — future TechCertGuide integration.

Source

GitHub Repository

Microsoft-Purview-DSPM-for-AI

Public

Enterprise configuration and validation of Microsoft Purview DSPM for AI (Classic) for Microsoft 365 Copilot governance.

Resources

Download Center

Project Case Study

PDF · Available soon

Architecture Diagram

PDF · Available soon

PowerShell Scripts

ZIP · Available soon

Technical Documentation

PDF · Available soon