Microsoft Kills Planned Purview Auto-Labeling Simulation Evaluation Before It Hits Preview

Microsoft has scrapped a planned enhancement to its Purview Information Protection auto-labeling simulation experience, catching many compliance teams off guard. The feature, listed under Microsoft 365 Roadmap ID 560707, was cancelled on July 22, 2026—before ever reaching public preview or general availability. Organizations that had incorporated it into their security roadmaps need to adjust immediately.

What Actually Got Cancelled

The now-defunct feature was called Microsoft Purview: Information Protection – Auto-labeling Simulation Evaluation. It was originally expected to enter public preview in May 2026 and become generally available in June 2026. Microsoft’s cancellation notice is blunt: the company “decided not to proceed with the feature as previously planned” after a strategy review, and development resources will shift toward “other classification and labeling investments.”

There are three critical take-homes from this language:

• This is not a delay. “Cancelled” on the Microsoft 365 roadmap means the feature is dead in its described form.
• No one outside Microsoft ever saw the feature. There is no preview build, no partial portal experience, and no supported capability to test.
• Microsoft hasn’t promised a direct replacement. The phrase “other classification and labeling investments” indicates continued work in the Purview portfolio, but not a one-for-one substitution of the evaluation tool.

The Existing Simulation Mode Isn’t Going Anywhere

Easily the most important clarification for Purview administrators: the cancellation does not remove or degrade the current auto-labeling simulation capability. That simulation mode—which lets you test service-side auto-labeling policies against real content before enforcing them—remains fully supported and available. What was cancelled was a “simulation evaluation” enhancement that would have added a new layer of analysis, reporting, or decision support on top of the existing simulation results.

Microsoft Purview’s existing auto-labeling simulation works at scale across SharePoint Online, OneDrive for Business, and Exchange Online. It allows security teams to understand which items would match a policy’s conditions and how a sensitivity label would be applied, without making any actual changes. This is a vital safety net, because auto-labeling can apply labels that encrypt, restrict sharing, or trigger downstream data protection behavior. A poorly configured policy could lock users out of legitimate content or quietly mislabel entire document libraries.

The cancelled evaluation feature was likely intended to make it easier to interpret simulation outcomes—perhaps through a dashboard, a readiness score, or an automated recommendation engine. Without detailed specs from Microsoft, the exact scope remains fuzzy, but the bottom line is that simulation remains, while the evaluation add-on does not.

What This Means for You

For Microsoft 365 and Security Administrators

If your team included Auto-labeling Simulation Evaluation in a project plan, risk assessment, compliance roadmap, or consulting engagement, it’s time for a clean-up. A cancelled roadmap item should be purged from:

• Purview rollout schedules
• Information Protection program charters
• Security architecture decision records
• Internal training decks
• Any dependency map that assumes the feature will arrive

Leaving it in place creates a phantom readiness signal. It can delay decisions that need to be made now—such as building your own evaluation processes or investing in third-party governance tools.

For Compliance and Data Governance Teams

The cancellation does not alter regulatory obligations or internal data security mandates. If your organization relies on sensitive information types, trainable classifiers, or exact data match to auto-label content, the existing simulation workflow is still the correct place to validate policies before activation. The process itself hasn’t changed; only the hoped-for evaluation shortcuts have disappeared.

For IT Leaders and Decision-Makers

This is a lesson in roadmap hygiene. Microsoft 365 roadmap entries are useful planning signals, not contractual commitments. A feature marked for a future date can change scope, slip, merge into another workstream, or evaporate before it ever ships. Build your security program around capabilities that are in production today, not items on a public list.

How We Got Here

Microsoft Purview’s auto-labeling feature has matured steadily since the introduction of service-side labeling policies. Early adopters had to rely on manual testing and limited previews; simulation mode eventually arrived to give organizations a controlled way to predict policy impact. But the evaluation of simulation results remained a manual, often painstaking task—administrators had to sample matched items, manually verify true and false positives, and document decisions across spreadsheets.

The cancelled feature appeared on the roadmap as a potential answer to that friction. Its naming suggested a more structured evaluation layer, perhaps similar to how some DLP solutions offer simulation impact assessments. But for reasons Microsoft has not shared in detail, it fell victim to portfolio prioritization.

The timeline is short and simple:

What to Do Now

1. Remove the feature from all planning artifacts

Go through every document, spreadsheet, and slide deck that mentions Auto-labeling Simulation Evaluation (or Roadmap ID 560707). Mark it “cancelled,” and strike it from delivery timelines. If you’ve made architecture decisions that depend on it, reassess them now.

2. Double down on your existing simulation workflow

Confirm that your team is using the current simulation mode effectively. That means:

• Running simulations on a tightly scoped basis before expanding scope.
• Involving data owners in reviewing matched samples—not just IT security.
• Documenting acceptance criteria (e.g., false positive rate below a certain threshold).
• Understanding the operational limits: simulation is point-in-time and subject to review capacity caps.

3. Build your own evaluation framework

You don’t need a Microsoft product team to improve how you evaluate simulation results. Create a repeatable internal process that records:

• Policy name and business owner.
• Target label and its protection settings.
• Simulation start and end dates, number of matched items.
• A sample review log showing confirmed true positives, false positives, and false negatives.
• Notes on conflicts with existing labels or other policies.
• Go-live approval and rollback owner.

This isn’t about creating paperwork for its own sake. It’s about making policy activation auditable, defensible, and less likely to cause a Monday morning outage.

4. Keep human review in the loop, especially for high-impact labels

A security administrator can tell you why a document matched a sensitive information type, but only a business owner can reliably say whether that document should actually get a “Confidential” label that encrypts it and restricts sharing. For any label that can disrupt user productivity if wrongly applied, insist on a joint review by IT and the data owner.

5. Separate auto-labeling evaluation from DLP simulation

The Purview compliance portal includes simulation modes for both auto-labeling and Data Loss Prevention (DLP). They are different. DLP simulation tests whether a policy would detect and block sensitive data leaving the organization; auto-labeling simulation tests whether a policy would apply the correct sensitivity label. Don’t confuse the two, and don’t assume that a successful DLP simulation replaces the need to test label application.

Outlook

The cancellation signals that Microsoft continues to invest in classification and labeling—just not through this specific evaluation lens. Purview remains central to Microsoft’s data security and compliance stack, and sensitivity labels are foundational to everything from DLP to insider risk management to AI governance. Expect future roadmap entries to address similar needs, possibly under different names or as part of broader Purview capabilities.

In the meantime, the existing tools are fit for purpose if you use them with discipline. The organizations that fare best will be those that treat auto-labeling not as a set-and-forget technology, but as an ongoing governance practice backed by human judgment and robust internal review processes. The cancelled evaluation feature may have made that easier, but its absence doesn’t change the core mission: safe, explainable, well-governed automated classification.