A risky AI use alert gets an owner
The sample response names who reviews and who may act.
Coming
See every demonstrationAI GOVERNANCE STARTER
Turn the AI policy you signed into controls the tenant enforces.
The policy says confidential data must not go into AI tools. Nothing in the tenant prevents it, detects it or records it, and the board that asked for the policy will ask for the evidence next. The AI Governance Starter deploys the Microsoft Purview controls you already license, with an operating model for approving tools and recording exceptions, and ends with every AI interaction audited.
01 Source02 Boundary03 Human decision04 Evidence
HOW IT WORKS
The control path in practice.
- Deploy Microsoft Purview Data Security Posture Management for AI to see which AI apps are used, what data categories go to them and by whom.
- Scope DLP policies to the generative AI location so labeled content is blocked or warned before it reaches Copilot, Copilot Studio agents or the browser AI tools endpoint DLP covers.
- Switch on Purview audit for Copilot and agent interactions, Communication Compliance where prompt review is required, and retention on AI interaction records.
- Hand over the AI governance operating model: new-tool approval, review content, exception recording and expiry, and the quarterly review.
THE PROBLEM
What is the AI policy gap?
A rule with no control behind it. The policy forbids confidential data in AI tools; the tenant neither prevents nor detects nor records it, so the first audit question has no evidence to answer with. The starter closes the gap with controls already in Microsoft 365, not another platform.
Most offers in this space deliver a policy document and a risk register. This one delivers working Purview controls and the operating model that keeps them true, and "starter" means scoped to finish, not scoped to everything.
THE MICROSOFT INSIGHT
How do you see which AI apps are in use and what goes to them?
Microsoft Purview Data Security Posture Management for AI runs a weekly data risk assessment of the most used SharePoint sites without activation and reports AI interactions that touch sensitive data, which is the first picture the starter gives the compliance officer.
Microsoft Purview Data Security Posture Management for AI runs a weekly data risk assessment of the 100 most used SharePoint sites without activation and reports AI interactions that touch sensitive data. (Microsoft Learn, checked Oct 31, 2026)
Read the posture report and the unsanctioned apps carrying labeled content are a list with owners. Without it they are a rumor until the audit.
THE METHOD
What gets deployed?
Four controls and one operating model, in the tenant you already run.
- Purview DSPM for AI: which AI apps are used, what data categories go to them, by whom.
- DLP policies scoped to the generative AI location, so labeled content is blocked or warned before it reaches Copilot, Copilot Studio agents or the browser AI tools endpoint DLP covers.
- Purview audit for Copilot and agent interactions, Communication Compliance for prompt and response review where required, retention on the AI interaction records.
- The operating model: new-tool approval, what gets reviewed, how exceptions are recorded and expire, and the quarterly review, as a working document your team owns.
THE SIGNAL
What flags a risky prompt?
Microsoft Purview Insider Risk Management ships a Risky AI usage policy template that scores prompts and responses containing sensitive information in Copilot and agents, and browsing to generative AI sites, and Communication Compliance adds trainable classifiers mapped to obligations such as FINRA rules where a regulated firm needs them.
Microsoft Purview Insider Risk Management ships a Risky AI usage policy template that scores prompts and responses containing sensitive information in Microsoft 365 Copilot and agents, and browsing to generative AI sites. (Microsoft Learn, checked Oct 31, 2026)
Score the prompts and the review queue is short and owned. Skip it and the review is the entire audit log, read by whoever is left.
WHAT IT IS NOT
Does the starter govern everything?
No. It governs AI within and adjacent to the Microsoft estate; unmanaged personal devices outside endpoint DLP are visible only where network or identity telemetry reaches; it does not cover model development, and it does not certify compliance with any AI regulation.
Prerequisites are Microsoft 365 E5 or the E5 Compliance add-on, which DSPM for AI, Communication Compliance and endpoint DLP need, and a written AI policy, or acceptance that drafting one is a supported prerequisite. Access is Compliance Administrator and Global Reader consent.
AFTER TWO WEEKS
What will be working after two weeks?
- Purview DSPM for AI is reporting which AI apps are used, what data categories reach them and by whom.
- DLP for the generative AI location blocks or warns on labeled content before Copilot, agents or covered browser tools process it.
- Copilot and agent interactions are audited and retained, with Communication Compliance review where you required it.
- Your team holds the operating model: tool approval, review scope, exception expiry and the quarterly review.
The work is delivered remotely inside your own Microsoft tenant, wherever your business is, and on site only when the work needs it.
HONEST ABOUT LIMITS
What this does not do yet
It does not block a typed prompt with a generally available control today. Blocking a Copilot or Cowork response when the typed prompt contains chosen sensitive information types is in preview and rolling out to tenants. (Microsoft Learn, checked Oct 31, 2026, preview) Blocking a Copilot or Cowork response when the typed prompt contains chosen sensitive information types is in preview and rolling out; the starter deploys the released controls on labeled files and emails and records the prompt path for the quarterly review.
It does not cover unmanaged personal devices beyond what network or identity telemetry reaches.
It does not govern model development or certify compliance with an AI regulation; the evidence is yours to present.
THE ONE THING TO REMEMBER
A policy with no control behind it is a promise; the starter puts the control behind each clause.
Said another way: see it, block it, audit it, then run the model that keeps it true.
Said a third way: the controls are in the license you already pay for; the work is switching them on in the right order.
SHOW YOUR WORK
See the control path on a fictional bank.
DIRECT ANSWERS
Questions before you decide.
What licenses does it need?
Microsoft 365 E5 or the E5 Compliance add-on, which DSPM for AI, Communication Compliance and endpoint DLP depend on. Gaps are recorded before deployment.
Do we need a written policy first?
Either an existing written AI policy, or acceptance that drafting one is a prerequisite the engagement supports. The AI Use Policy Builder on this site writes a first draft with the control named behind each clause.
What access do you need?
Compliance Administrator and Global Reader consent for the Purview deployment and reporting. Nothing is copied out of your tenant.
Where is it sold?
On Azure Marketplace, bought through the Azure portal and billed on your existing Microsoft agreement; private offers on request. Plans and terms are on the listing. No price appears on this site.
Does it cover agents?
The Purview controls cover Copilot Studio agents and Copilot interactions. Agent identity, inventory and approval paths are the Agent 365 governance engagement, which the starter's operating model hands off to.
NEXT STEP
Start with a boundary you can explain.
Choose a focused check, or speak with the team about the workflow you need to improve.
Buy through Microsoft Marketplace Bought through the Azure portal and billed on your Microsoft agreement; plans and terms are on the listing.