3.3.2. DLP Alerts, Activity Explorer, and DSPM for AI
💡 First Principle: When something goes wrong with data — a policy violation, a suspicious action, an AI interaction with sensitive content — you need tools to detect it in near real time and build an audit trail. DLP alerts, Activity Explorer, and DSPM for AI are the operational monitoring layer of Purview.
DLP alerts: When a DLP policy match occurs and the policy is configured to alert, an alert appears in the Microsoft Purview compliance portal. Administrators can review the alert, see the content that triggered it, and decide whether to escalate or dismiss. DLP alerts feed into the broader incident management workflow.
Activity Explorer: A timeline view of data-related user activity across M365. You can see:
- When sensitivity labels were applied, changed, or removed from content
- When DLP policies were triggered (and whether users overrode them)
- When sensitive files were accessed, downloaded, moved, or shared
- Who performed each action and from which device
Activity Explorer is your audit trail for data governance. If a user claims they didn't share a sensitive file externally, Activity Explorer shows the truth.
Activity Explorer vs. Content Explorer and the audit log: Content Explorer (3.3.1) shows what labeled or sensitive content exists now and where it lives; Activity Explorer shows what was done to it over time. Activity Explorer centers on labeling and DLP events; for a complete record of who opened a specific SharePoint or OneDrive file, also search the Purview audit log, which records file-access events.
A run of repeated DLP matches or overrides by one user points to deliberate circumvention rather than accident: investigate it and bring in Insider Risk Management (3.1.3), which correlates signals into a behavioral picture of the user.
DSPM for AI (Data Security Posture Management for AI):
- Microsoft's newest governance capability, purpose-built for AI risks
- Surfaces AI-related data activity: what data Copilot accessed, which sensitive files were referenced in Copilot prompts, which users are generating high volumes of AI interactions with sensitive content
- Identifies oversharing risks that Copilot is likely to expose
- Recommends remediation actions (fix permissions, apply labels, create DLP policies for AI)
- Was previously named Microsoft Purview AI Hub (older material still uses that name); it lives in the Microsoft Purview portal
- Its activity explorer view lists AI interactions (Copilot prompts and responses) and the sensitive information found in them
- Works with DLP for the Microsoft 365 Copilot location, which can stop Copilot from responding to prompts that contain chosen sensitive information types and from using files or emails with chosen sensitivity labels in its answers
- Because Copilot runs with the asking user's permissions, content it surfaces to the wrong person almost always means that person already has access, so investigate permissions (oversharing) first
💡 Key Point: DSPM for AI is a discovery and reporting tool, not a blocking tool. It shows you what's happening; DLP policies and permissions changes are what you do about it.
⚠️ Exam Trap: Activity Explorer shows activity after the fact — it's forensic, not preventive. To prevent a sensitive action, use DLP. To see what happened after the fact, use Activity Explorer.
Reflection Question: After deploying Copilot, your security team wants to know which sensitive files are being referenced in user prompts. Which Microsoft Purview feature should they use, and what follow-up actions might it recommend?