Skip to content

Control 1.5: Data Loss Prevention (DLP) and Sensitivity Labels - Verification & Testing

This playbook provides verification and testing guidance for Control 1.5.


Verification Steps

Step Action Expected Result
1 Navigate to Purview > DLP > Policies Policies listed
2 Verify AI locations selected Copilot/Copilot Studio included
3 Confirm SIT readiness (Control 1.13) SITs exist and match sanitized test data
4 Create sanitized test content + apply labels Files/messages prepared; labels applied
5 Run AI test prompts and agent scenarios Blocked/warned/audited per zone
6 (Optional) Run Endpoint DLP tests Endpoint actions enforce as configured
7 Check DSPM for AI > Policies AI-related DLP policies visible
8 Check audit logs / incident reports DLP events and reports captured

Test Cases (Copilot/Agent; U.S.-Only)

Use non-production, sanitized data. Do not use real customer data.

Test ID Scenario Input Expected
AI-01 Prompt includes SSN Prompt text includes SSN-formatted value Zone 2: warning + log; Zone 3: block + incident
AI-02 Prompt includes ABA routing Prompt includes routing-formatted value Same as AI-01
AI-03 Retrieval from labeled content Agent grounds on Highly Confidential file Block per label rule; event logged
AI-04 Retrieval from Confidential content Agent grounds on Confidential content Warn or allow-with-audit; event logged
AI-05 Attempted sensitive output Agent summarizes doc with bank account numbers Output blocked/redacted; event logged

Test Cases (Virtual Governance Connectors)

Verify that DLP policies enforce virtual governance connector classification rules. Use a non-production environment for testing.

Zone 3 Virtual Connector Test Cases (Required)

Test ID Scenario Action Expected Result
VC-01 Agent creation with blocked connector Create Copilot Studio agent that uses HTTP Webhook connector (blocked in Zone 3) Agent creation blocked with DLP policy error; event logged in audit log
VC-02 HTTP call to non-allowlisted endpoint Configure agent skill to call external API not in HTTP endpoint allow list (e.g., https://api.external-service.com) Agent skill execution fails with DLP violation; error message displays to user; event logged in Purview audit log
VC-03 Verify all 11 connectors classified Navigate to PPAC > DLP policy > Connectors tab All 11 virtual governance connectors appear with correct Zone 3 classification (see table below)
VC-04 Knowledge Source connector with approved SharePoint Create agent using Copilot Studio Knowledge connector pointing to approved SharePoint site (after Control 1.3 implemented) Access allowed; agent successfully indexes knowledge source; no DLP error
VC-05 Channel connector enforcement Attempt to publish agent to Custom Website Channel (blocked in Zone 3) Publishing blocked with DLP error; user receives notification "This connector is blocked by DLP policy"
VC-06 AI Builder GPT connector usage Create agent using AI Builder GPT connector for text generation If Business-classified: allowed; agent creates successfully and generates text
VC-07 Zone 3 connector classifications match recommended settings Export DLP policy configuration via PowerShell (see PowerShell Setup playbook) All 11 connectors match Zone 3 recommended classifications from Control 1.5
VC-08 HTTP endpoint filtering blocks non-allowlisted external APIs Create agent skill that calls https://api.twitter.com/2/tweets (social media API, not in allow list) Agent skill fails with HTTP endpoint filtering error; event logged with blocked URL
VC-09 DLP audit log captures connector classification violations Review Purview Audit Log for past 7 days; search for DLP events related to virtual connectors Audit log entries show DLP violations with connector name, user, timestamp, and policy name

Expected Zone 3 Classifications (for VC-03 verification):

Connector Expected Classification Verification Method
AI Builder (GPT) Business Visible in Business group in PPAC
AI Builder (Document Processing) Business Visible in Business group in PPAC
Copilot Studio Topics Business Visible in Business group in PPAC
Copilot Studio Skills Business Visible in Business group in PPAC
Copilot Studio Knowledge Business Visible in Business group in PPAC
HTTP with Microsoft Entra ID Business (with endpoint filtering) Visible in Business group; endpoint filtering badge present
HTTP Webhook Blocked Visible in Blocked group in PPAC
Direct Line Business Visible in Business group in PPAC
Microsoft Teams Channel Business Visible in Business group in PPAC
SharePoint Channel Blocked Visible in Blocked group in PPAC
Custom Website Channel Blocked Visible in Blocked group in PPAC

Zone 1-2 Virtual Connector Test Cases (Optional)

Test ID Scenario Action Expected Result
VC-10 Zone 1 HTTP Webhook usage Create agent with HTTP Webhook connector (Non-Business or Blocked in Zone 1) If Non-Business: allowed but cannot mix with Business connectors; If Blocked: creation fails
VC-11 Zone 2 Custom Website publishing Attempt to publish agent to Custom Website Channel (Blocked in Zone 2) Publishing blocked with DLP error
VC-12 Zone 1 HTTP endpoint block list enforcement Create agent skill calling blocked social media API (e.g., https://api.twitter.com/*) Agent skill fails with HTTP endpoint filtering error

HTTP Endpoint Filtering Test Cases (Zone 3 Required)

Zone 3 Requirement

HTTP endpoint filtering test cases (VC-08, VC-13, VC-14, VC-15) are mandatory for Zone 3 environments. Failure to configure endpoint filtering creates data exfiltration risk.

Test ID Scenario Action Expected Result
VC-13 Allow list permits internal API Create agent skill calling internal API in allow list (e.g., https://api.internal.yourbank.com/customer) Agent skill executes successfully; API call completes; response returned to agent
VC-14 Allow list permits regulatory data source Create agent skill calling SEC EDGAR API https://api.sec.gov/submissions/ (in allow list) Agent skill executes successfully; regulatory data retrieved
VC-15 Allow list blocks unapproved external API Create agent skill calling unapproved market data API https://api.unknown-vendor.com/data (not in allow list) Agent skill fails with HTTP endpoint filtering error; no API call made
VC-16 HTTP endpoint filtering audit log Review Purview Audit Log for HTTP endpoint filtering events Audit log shows blocked URL, user, agent name, timestamp

Evidence Collection

For audit readiness, collect and retain the following evidence:

  • Screenshot: PPAC DLP policy showing all 11 virtual governance connectors with zone-appropriate classifications
  • Screenshot: HTTP endpoint filtering configuration (allow list patterns) for HTTP with Microsoft Entra ID connector
  • PowerShell Export: CSV file of virtual connector classifications (from PowerShell Setup playbook)
  • Audit Log: Purview Audit Log query results showing DLP enforcement events for virtual connectors (past 30 days)
  • Test Results: Test execution log with test ID, timestamp, tester account, expected vs actual outcome, and screenshots of errors
  • Change Record: IT change control ticket documenting DLP policy configuration and dual approval (Zone 3 requirement per Control 2.1)

Test Cases (Endpoint DLP; Optional)

Run only if Devices/Endpoint DLP is in scope.

Test ID Scenario Action Expected
EP-01 Copy/paste exfiltration Copy SSN text, paste into AI prompt in browser Block or warn; event logged
EP-02 File exfiltration Upload Highly Confidential file to AI web experience Block or warn; event logged
EP-03 Removable media Copy file with bank account numbers to USB Block or warn; event logged

Evidence to Retain (Audit-Ready)

  • DLP policy export (policy + rules) and change record
  • Screenshot evidence showing AI locations selected and rule conditions
  • SIT validation evidence per Control 1.13
  • Label taxonomy decision + label publication configuration
  • Test execution log: test ID, timestamp, account used, expected vs actual
  • Incident report samples and notification configuration
  • Unified audit log evidence for representative DLP events
  • DSPM for AI policy visibility and oversharing assessment outputs

Confirmation Checklist

  • DLP policies created with AI locations (Copilot, Copilot Studio)
  • All 11 virtual governance connectors classified in Power Platform DLP policy
  • HTTP endpoint filtering configured and tested (Zone 3 requirement)
  • Virtual connector test cases executed with expected results (VC-01 through VC-06)
  • Sensitivity labels created and published
  • Label-based DLP rules configured
  • SITs validated per Control 1.13
  • Standard AI test cases executed (AI-01 through AI-05)
  • Audit logs capture DLP events including virtual connector enforcement
  • DSPM integration verified (if applicable)
  • Evidence artifacts collected and stored

Back to Control 1.5 | Portal Walkthrough | PowerShell Setup | Troubleshooting


Updated: January 2026 | Version: v1.2