What each test proves
A DLP rule that matches the right item is only half the test. Enforcement waits until the rule also ignores near-misses, leaves people outside the scope alone and shows users the message you approved.
| Test | What it proves | Policy state | Blocks enforcement? |
|---|---|---|---|
| True match | The rule catches the sensitive item | Simulation mode | Yes |
| Near-miss | Look-alike content does not match | Simulation mode | Yes |
| Scope | Users outside the pilot group see nothing | Simulation with policy tips | Yes |
| User experience | Policy tip wording and the override flow work | Simulation with policy tips, then Block with override | No, tune it |
| Enforcement | The action happens and the event is recorded | Turned on | Yes |
| Copilot prompts | No response, or no web search, for prompts with the chosen sensitive information types | Turned on, Copilot location | Yes |
| Copilot labeled content | Labeled files and emails are not used in responses | Turned on, Copilot location | Yes |
| Known gap | The gap is owned by a named control | Any | No, but it must be recorded |
Where the evidence shows up
- Simulation mode. The simulation overview shows scan status per location, items scanned and matches; Items for review lists each matched item; simulation alerts appear only on their own tab. Simulations run up to 15 days and keep data for 30 days.
- Which items a simulation sees. SharePoint and OneDrive evaluate existing and new or changed items. Exchange, Teams and Devices evaluate only items new during the simulation, so create fresh test items after it starts.
- Simulation with policy tips. Nothing is enforced, but users get policy tips and notification emails. Microsoft recommends narrowing the scope to a pilot group at this stage.
- Turned on. Actions are enforced; monitor DLP alerts and Activity explorer. Start with Audit only (Allow for Devices) or Block with override, because override reasons show you the false positives before a full Block.
Testing the Microsoft 365 Copilot location
- Its own policy. The Microsoft 365 Copilot and Copilot Chat location is available only in the Custom template, and selecting it disables every other location in that policy.
- One condition type per rule. A sensitivity label condition and a sensitive information type condition cannot share a rule; use two rules in the same policy.
- Wait before you test. Policy updates can take up to four hours to reach Copilot. In Word, Excel and PowerPoint the policy is checked when a file is opened.
- Check preview status. Blocking prompts that contain sensitive information types, and excluding external email, are listed by Microsoft as preview. Confirm they have reached your tenant before you write pass rules for them.
- Test as the prompting user. Copilot DLP scoping follows the user who submits the prompt, not the owner of the file or email Copilot reads.
- Record the gaps. DLP does not scan files uploaded into a prompt, calendar invites are not covered, and labeled-email coverage applies to emails sent on or after January 1, 2025.
Run the Copilot-location tests before licenses go out, as part of a Microsoft 365 Copilot readiness assessment.
Acceptance rule
Write the expected result and pass rule for every row before the first run. Any failed true-match, near-miss or scope test keeps the policy in simulation; tune it, copy it to a new version if it is already in production, and rerun. Move to enforcement only when the policy owner signs off on the evidence. For Copilot agents built on the same data, pair this with the Copilot Studio agent acceptance test pack.
Download the DLP policy test matrix (CSV). Its twelve example rows cover Exchange, SharePoint and OneDrive, Devices and the Copilot location. They are synthetic and labeled, not client data; use a payment provider’s published test card number or other test values, never real personal data.
Primary sources
- Learn about data loss prevention simulation mode — Microsoft Learn, accessed 2026-10-28
- Create and deploy a data loss prevention policy — Microsoft Learn, accessed 2026-10-28
- Microsoft Purview DLP for Microsoft 365 Copilot and Cowork — Microsoft Learn, accessed 2026-10-28
Frequently Asked Questions
How long does a Purview DLP simulation run?
Up to 15 days, so results build up over time rather than being a point-in-time snapshot. Simulation data is kept for 30 days. For SharePoint and OneDrive, existing and new or changed items are evaluated; for Exchange, Teams and Devices, only items that are new during the simulation are evaluated, so send fresh test items after the simulation starts.
Do simulation alerts appear in the DLP alerts console?
No. While a policy is in simulation mode its alerts appear only on the simulation alerts tab. They do not show in the DLP alerts console and do not flow into the Microsoft Defender portal, so plan alert-routing tests for after the policy is turned on.
How long until a Copilot DLP policy change takes effect?
Microsoft says updates to a DLP policy can take up to four hours to reflect in Microsoft 365 Copilot and Copilot Chat. Test after that window. In Word, Excel and PowerPoint the policy is evaluated when a file is opened, so a label applied mid-session is enforced from the next open.
Does DLP check files uploaded into a Copilot prompt?
No. DLP checks the text typed into the prompt, not the contents of files uploaded directly into it. Record this as a known gap in the test plan and name the control that covers it, rather than writing a test that expects DLP to catch it.
Talk to a senior architect
A one-hour call with a senior EPC Group architect covers your DLP rules, the tests that should gate their enforcement and how to stage them by pilot group. Email contact@epcgroup.net or call 888-381-9725.