SharePoint Restricted Search retirement and new Copilot governance controls led this week’s Message Center updates.

Microsoft 365 Copilot Cowork is generally available worldwide (Tier 1 languages), while Frontier remains the early access channel. Frontier management also introduces usage-based billing, which may affect cost controls and governance for pilot users. Review who is in Frontier, validate Purview/governance settings, and align budgeting with the new billing model.

• GA includes multi-model capabilities, plugins, updated navigation, skill management, and Purview integration
• Frontier: GPT-5.5 Thinking and Edge browser-use features are admin-enabled / default-off
• Frontier-specific: usage-based billing to scale access based on consumption

Microsoft is retiring Restricted SharePoint Search (RSS) in SharePoint Online, which some organizations use to limit how SharePoint content appears in Microsoft Search and Copilot during reviews. Organizations should transition to Restricted Content Discovery (RCD) or other controls to manage discoverability and governance. Assess current reliance on RSS and plan the migration to RCD policies.

• Change impacts content discoverability in Microsoft Search and Microsoft 365 Copilot scenarios
• RCD provides more granular and scalable discovery controls

Starting July 2026, Microsoft is adding an admin control to permanently remove all People Skills data from your tenant. This supports data governance, compliance, and offboarding scenarios where skills data must be removed. Decide whether you need this capability and update internal processes and documentation accordingly.

• Applies to People Skills used across Copilot, profile cards, and Viva experiences
• Designed for tenant-wide removal and permanent deletion scenarios

Microsoft Copilot Studio will support using Model Context Protocol (MCP)-compliant tools and knowledge servers as steps inside agent workflows. This enables standardized, reusable tool integration across multiple agents with structured inputs/outputs. Platform owners should review governance, connector/tool approval, and lifecycle management for MCP endpoints before enabling broad use.

• Workflow steps can call MCP tools deterministically using structured inputs/outputs
• Supports reuse of the same MCP server across multiple agents

Copilot Studio is adding evaluation capabilities to measure, validate, and improve AI actions inside workflow automations. Teams can assess whether critical AI-driven steps behave as intended before shipping and understand downstream impact of quality issues. Establish evaluation criteria, testing pipelines, and release gates for production workflows.

• Node-level evaluations for agent actions in workflows
• Targets improved reliability before deployment

Microsoft Teams is enhancing recap sharing so organizers can grant access to meeting recordings, transcripts, AI-generated summaries, and notes when sharing or copying a recap link. This changes how meeting artifacts can be shared with people who were not in the original meeting. Review sharing policies, sensitivity labeling, and user guidance for recap links.

• Applies to recap links for recordings, transcripts, summaries, and notes
• Improves external-to-meeting sharing flows (subject to policy)

Edge for Business will support summarization and contextual grounding with Microsoft 365 Copilot Chat, including multiple open tabs and YouTube videos, plus Microsoft 365 documents. This expands what users can reference directly from the browser side pane, affecting data access patterns and user expectations. Validate tenant and Edge policies, and update guidance for acceptable data usage and compliance.

• Grounding across multiple browser tabs and supported content types
• Designed for use directly in the Edge side pane

Microsoft is introducing Copilot Chat in Microsoft Forms along with a Surveys Agent to help create, distribute, and analyze forms. The existing Copilot refinement bar and insights summary on the results page will be replaced by a new Copilot dynamic action button. Update end-user training and any internal documentation that references the previous UI.

• AI assistance for authoring and analysis in Forms
• Results-page Copilot experience changes to a single dynamic action entry point

Outlook emails can be added as references in Microsoft 365 Copilot Notebooks, enabling notebooks to be grounded in email threads and decisions alongside files. This increases the likelihood that email content is used as source material for generated briefs, summaries, and presentations. Review information governance, user guidance, and sensitivity expectations when referencing emails in notebooks.

• Email threads become selectable notebook references
• Improves context for generation across notebook scenarios

Microsoft is updating agents in SharePoint to simplify how site-scoped agents are launched and to add new site AI settings. This affects both ready-made agents automatically scoped to SharePoint sites and custom-built agents created by users with site permissions. Review site-level AI settings, helpdesk guidance, and rollout communications for site owners.

• Impacts the experience of selecting and launching site agents
• Adds/updates site-level controls for AI behavior

Excel Agent in Microsoft 365 Copilot is adding enterprise search across Microsoft 365 content and a new file selector to find and choose files to process. This expands what users can pull into Excel Agent for grounding and reduces friction in selecting sources. Confirm search/data access expectations, and update user guidance on using enterprise content in Excel Agent outputs.

• Enterprise search can include files, email, meetings, chats, people, transcripts
• New file selector streamlines choosing inputs for Excel Agent

Microsoft Purview is expanding the Data Security Posture Agent with credential scanning to discover exposed credentials across scoped locations. The feature provides risk scoring and AI-generated insights to support investigation and remediation. Security teams should plan rollout, confirm scoped locations, and align response playbooks for detected credentials.

• Detects credential types such as Entra credentials, private keys, and API tokens
• Provides risk scores, confidence ratings, and categorization