The short answer
Yes, but under conditions that determine the coverage. Microsoft's own documentation lists four prerequisites for prompt detection in DSPM for AI: an Edge configuration policy, device onboarding, Endpoint DLP and the Purview browser extension. Network-side inspection via Entra Global Secure Access is listed by Microsoft as preview.
Sources: learn.microsoft.com/purview/dspm-for-ai-considerations and learn.microsoft.com/purview/ai-other-apps, retrieved 2026-09-01.
What follows from it
Detection depends on the endpoint. That is not a weakness of the classifiers — Purview's classification is strong, and it knows your labels. It is a statement about where the measurement happens:
- A managed device with Edge, onboarding and the extension is covered.
- A personal laptop, a different browser, or a device without onboarding is not.
- Those are precisely the devices staff move to when a tool is blocked.
Where a control layer in front applies
A layer that sits on the network in front of the model sees the prompt regardless of device, browser or onboarding state. It does not replace Purview: Purview classifies your data in your tenant, and that stays its strength. It closes the gap that opens when measurement hangs off the endpoint.
Stated honestly: this page describes the documented state as of 2026-09-01. Microsoft continues to develop DSPM for AI, and a preview can become generally available. Anyone deciding today should check the linked pages.