Security Copilot in Purview

November 2024

Audit

• Ignite 2024 update: New guidance for about how audit logs are generated for user interactions and admin activities related to Microsoft Copilot and AI applications.

Communication Compliance

• Ignite 2024 update: New support for detecting generative AI interactions with Microsoft Copilot, non-Microsoft connected generative AI applications, and AI applications from browser and network activity by users in your organization.
• Ignite 2024 update: New support for harmful user-generated and AI-generated content in applications and services. This includes evaluation of user prompts submitted to generative AI services and the inclusion of known text content that may be sensitive to your organization.
• Ignite 2024 update: Expanded integration for Communication Compliance indicators and generative AI policy indicators in Insider Risk Management policies.

Compliance Manager

• In preview: Users can now customize Compliance Manager regulatory templates by adding controls and improvement actions in order to build custom assessments.
• Updated: Compliance Manager has a Reports page, which displays a history of the recent activity impacting an organization's compliance score.
• In preview: Organizations can assess their predeployment compliance (preview) before deploying any new Microsoft services.

Data Governance

• Ignite 2024 update: The Microsoft Purview Data Catalog is changing its name to Microsoft Purview Unified Catalog. All the features will stay the same. Check the name in your region.

Data lifecycle management and records management

• Ignite 2024 - In preview: Retention policies support Teams AI-generated notes when you use the Teams chat location for AI-generated notes in chat, and OneDrive accounts for AI-generated notes in meetings. Retention labels are supported for AI-generated notes in meetings, but don't support cloud attachments.
• Ignite 2024 update: In preview, you can create separate retention policies for Copilot interactions and Teams chats. Newly created policies use separate locations. You can also separate existing policies. Previously, messages from Teams and Microsoft Copilot were automatically included in the retention policy location.

Data Loss Prevention

• Ignite 2024 update: In preview DLP policies for Fabric support semantic models and lakehouses. Get started with Data loss prevention policies for Fabric and Power BI.
• Ignite 2024 update: In preview four new scenarios for applying controls to files that fail classification or are outside the scope of DLP monitoring or are in scope but unknown or subsets of files.
• Apply controls to supported files that fail scanning
• Apply controls to all unsupported files
• Apply controls to some unsupported files
• Disable scanning for some supported files and apply controls
• Ignite 2024 update: In preview, you can use Security Copilot to gain insights into your DLP policies. Get insights with Security Copilot.
• Ignite 2024 update: In preview, you can use Power Automate integration to trigger custom workflows as a DLP rule action: Get started with Power Automate integration
• Ignite 2024 new: In preview, Microsoft 365 Copilot is available as a monitored location in DLP policies. Learn about the Microsoft 365 location (preview).

Data Security Posture Management for AI

• Ignite 2024 - General Availability (GA): Previously named Microsoft Purview AI Hub, Data Security Posture Management for AI (DSPM for AI) is now GA with added preview functionality and changes:
• Data assessments (in preview) to identify and fix potential oversharing risks in your organization. Fixes include helping you to create a DLP policy to prevent Microsoft 365 Copilot from summarizing files with specific sensitivity labels, and specifying which SharePoint sites to be restricted from Microsoft 365 Copilot (SharePoint restricted content discoverability).
• New recommendations and policies that include detecting risky AI usage using an Insider Risk Management policy.
• Extending support for other generative AI apps, including Copilot Studio, ChatGPT - Microsoft Purview integration with ChatGPT Enterprise Compliance API, and new support for Microsoft Teams AI-generated notes, currently in preview. You see these categorized on the Reports page and in activity explorer as Microsoft Copilot Experiences and Enterprise AI apps.
• A version for E3 customers, to help turn on audit, see data related to Microsoft 365 Copilot interactions, and other AI-related events in activity explorer.

Device onboarding

• Ignite 2024 update: In preview, Microsoft Data Loss Prevention just-in-time (JIT) protection is available on the three most recent releases of macOS. Onboarding devices into device management.

eDiscovery (preview)

• Ignite 2024 update: New support and upgrades to the export flow in search and review sets, providing a unified export structure across premium and non-premium feature exports, faster export performance, detailed reporting, and flexible export options.
• Ignite 2024 update: New support to identify web queries in Microsoft 365 Copilot usage.

Insider Risk Management

• Ignite 2024 - In preview: New Risky AI usage policy template to help detect and enable risk scoring for user prompts and system responses across AI tools in your organization.

Security Copilot in Purview

• Ignite 2024 update - In preview multiple new capabilities are available in Microsoft Security Copilot in Purview. Security Copilot integration in Microsoft Purview.
• Ignite 2024 update - In preview Security Copilot is available in Microsoft Purview activity explorer. It can help with data hunting and generation of filters from natural language prompts Security Copilot in activity explorer (preview).

Sensitivity labels

• Ignite 2024 - In preview: SharePoint document libraries can be configured for a sensitivity label to extend permissions to downloaded documents, and protect files from being copied or moved. For more information, see Configure SharePoint with a sensitivity label to extend permissions to downloaded documents.

Trainable classifiers

• Ignite 2024 update: In preview there are two new trainable classifiers:
• Prompt Shields - Detects adversarial user input attacks in LLMs. Trainable classifiers definitions.
• Protected material - Detects known text content that may be protected under copyright or branding laws. Trainable classifiers definitions.