Control 2.9: Defender for Cloud Apps — Copilot Session Controls — Troubleshooting
Common issues and resolution steps for Defender for Cloud Apps session controls.
Common Issues
Issue 1: Routed Browser Sessions Not Appearing in Policy Reports or Activity Log
- Symptoms: Expected Microsoft 365 browser sessions are not visible in the Defender for Cloud Apps policy report or activity log despite session policies being configured
- Root Cause: Conditional Access App Control may not be fully configured for the target web apps, the session might not be browser-based, or investigators might be expecting Monitor only to record more than the routed login activity.
- Resolution:
- Verify Conditional Access App Control is enabled for the target Microsoft 365 web apps in Defender settings
- Check the Conditional Access policy routes traffic through Defender for Cloud Apps
- Verify the user's session was proxied (check for the Defender for Cloud Apps lock icon in Edge or the
.mcassuffix in other browsers) - If the policy uses Monitor only, expect routed Login activity only. For detailed file or action monitoring, use Block activities or Control file download/upload (with inspection) with Audit or Block
- Confirm the user tested in a supported browser session rather than a native desktop client such as Teams desktop
Issue 2: Session Control Causing Performance Issues
- Symptoms: Users report slower Copilot responses or Office application performance degradation when session controls are active
- Root Cause: Proxying sessions through Defender for Cloud Apps adds latency. Content inspection adds additional processing time.
- Resolution:
- Review session policy scope — limit content inspection to specific conditions rather than all traffic
- Optimize content inspection rules to target only high-risk activities
- Consider monitoring-only policies (no blocking) for general use and blocking policies for specific conditions
- Work with Microsoft support to optimize proxy routing if latency is significant
Issue 3: False Positive Alerts Overwhelming Security Team
- Symptoms: High volume of alerts for normal Microsoft 365 web-session activity, creating alert fatigue
- Root Cause: Alert thresholds may be too sensitive, normal business activity may match the conditions, or the anomaly-detection learning period may not have elapsed.
- Resolution:
- Review alert patterns and identify the most common false positive types
- Adjust alert thresholds to reduce noise (e.g., increase the activity count threshold)
- Create suppression rules for known-good activity patterns
- Use alert aggregation to group related alerts into single incidents
- Allow the documented seven-day anomaly-detection learning period before concluding that the baseline is stable
- Implement a tiered alerting approach (informational vs. actionable)
Issue 4: Content Inspection Missing Sensitive Data
- Symptoms: Known sensitive data passes through Copilot sessions without triggering content inspection alerts
- Root Cause: Content inspection patterns may not match the specific data format, the policy may be using Monitor only instead of a content-inspecting session control type, or inspection may not be configured for the specific supported activity.
- Resolution:
- Review content inspection sensitive information type definitions
- Test with known data patterns to verify detection capability
- Verify the session policy uses Block activities or Control file download/upload (with inspection) for the supported activity where data was missed
- Add custom inspection rules for organization-specific data patterns
Issue 5: Agent Threat Detection Alerts Not Appearing
- Symptoms: Copilot agents are deployed but no agent-related alerts appear in Microsoft Defender; the Assets > AI agents and Incidents & alerts views show no agent data despite agent activity
- Root Cause: Microsoft identifies Copilot Studio AI agent protection, not Microsoft Foundry coverage, as the Preview item in its November 2025 AI-security release notes. As of July 1, 2026, the licensing cutover applies to both Copilot Studio and Microsoft Foundry agent coverage: the tenant must hold a Microsoft Agent 365-eligible license and complete Agent 365 onboarding. Microsoft 365 E5 is not an exclusive prerequisite; tenants without an eligible license no longer receive the transitioned agent threat-detection coverage.
- Resolution:
- Verify the tenant holds a Microsoft Agent 365-eligible license using the Control 2.9 qualifying-license guidance and has completed Agent 365 onboarding (required as of July 1, 2026 for Copilot Studio/Foundry agent coverage)
- Navigate to Defender portal > Settings > Security for AI > Get started and confirm the Security for AI Agents toggle and the Copilot Studio real-time protection step are both enabled
- Check that agents appear in Defender portal > Assets > AI agents with a current risk level — agents not listed there may not be fully onboarded for threat detection coverage
- Verify no filter is hiding agent incidents: in Incidents & alerts, clear all filters and search for recent agent-related events
- Allow 24-48 hours after enabling detection — initial alert generation may be delayed
- If alerts still do not appear after enabling: open a Microsoft support ticket referencing Microsoft Defender Security for AI agent threat detection
Issue 6: Generative AI App Catalog Not Showing Discovered Apps
- Symptoms: The Cloud app catalog Generative AI filter shows apps but no "discovered" usage data, or the catalog shows fewer apps than expected
- Root Cause: Discovery data requires traffic analysis either through Microsoft Defender for Endpoint integration or manual log upload. Without endpoint agents or log upload configured, only the catalog (not the Discovered apps usage evidence) is visible.
- Resolution:
- Navigate to Defender portal > Cloud Apps > Settings > Cloud Discovery to check discovery configuration
- If Microsoft Defender for Endpoint is deployed: verify the "Defender for Endpoint integration" toggle is enabled in Cloud Discovery settings — this enables automatic traffic analysis for discovery
- If Defender for Endpoint is not deployed: configure manual log upload from proxy or firewall logs to enable discovery
- Allow 24-48 hours after enabling discovery for initial data to appear
- Verify the correct data source is connected: Defender portal > Cloud Apps > Cloud Discovery > Data sources
Diagnostic Steps
- Verify proxy routing: Check browser URL for Defender for Cloud Apps session control indicators during Copilot use
- Check policy status: Defender > Policies > verify all session policies are "Enabled"
- Review policy reports and activity log: Search for routed sign-ins and any monitored file or action events from the last 24 hours
- Test detection: Use a test document with known sensitive data patterns
- Check CA integration: Verify the Conditional Access policy references Defender for Cloud Apps and that native-client blocking is configured where required
Escalation
| Severity | Condition | Escalation Path |
|---|---|---|
| Low | Minor logging gaps | Security Operations |
| Medium | Session controls causing user-impacting latency | Security Operations and Microsoft support |
| High | Content inspection failing to detect sensitive data | Security Operations and Information Protection team |
| Critical | Session controls non-functional — no monitoring on Copilot | CISO and Security Operations immediately |
Related Resources
- Portal Walkthrough — Session control configuration
- PowerShell Setup — Monitoring scripts
- Verification & Testing — Session control validation
- Back to Control 2.9