The AI week, distilled.
Week 35 · 2026
This week in Microsoft AI

Microsoft 365 extensibility and GitHub Copilot governance both moved forward

Microsoft opened a broader preview path for building interactive Copilot experiences inside Microsoft 365 via SPFx. GitHub shipped multiple Copilot admin and product changes that affect how enterprises govern models, standardize integrations, and plan licensing and rollout.

01

Copilot Components open public preview for SPFx developers

Microsoft moved Copilot Components (formerly SharePoint Copilot Apps) into public preview and enabled it worldwide by default. Microsoft targets general availability for October 2026.

  • Use existing SPFx engineering practices to build internal Copilot canvas experiences without standing up a separate app platform.
  • Pilot Copilot-driven workflows in M365 with standard SharePoint governance controls, which helps security teams keep experimentation inside known boundaries.
  • Plan budget and release management for an October 2026 GA window if you intend to productionize Copilot canvas experiences in HR, finance, or operations.
02

GitHub Copilot global model policy reaches GA

GitHub made a global model policy generally available for Copilot Business and Enterprise and will gradually enforce it through September 1, 2026. New or unconfigured models inherit the global policy, and open-weight or data-retaining models are disabled by default.

  • Set one enterprise-wide approval stance for Copilot models to reduce inconsistent usage across Czech subsidiaries and dev teams.
  • Rely on safer defaults that block open-weight and data-retaining models unless administrators explicitly approve them for controlled use cases.
  • Prepare for enforcement timing through 2026-09-01 by reviewing model allowlists and exception processes with security and legal stakeholders.
03

Copilot app Customize tab is generally available

GitHub made the Copilot app Customize tab generally available to centralize management of MCP servers, plugins, skills, and canvases. The change consolidates configuration that previously required multiple surfaces.

  • Centralize which plugins and MCP servers developers can use, which helps enforce data-access rules for internal systems.
  • Standardize organization-wide Copilot integrations to reduce per-team configuration drift and support overhead.
  • Use the single control point to roll out curated internal skills and canvases that align with Czech compliance and audit requirements.
04

Copilot weekly release adds CLI improvements and WSL support

GitHub’s Copilot weekly release shipped a GA Customize tab, improved multi-session management, a new Rust runtime for Copilot CLI, experimental WSL support, and deeper integration across MCP servers, plugins, skills, and canvases. The updates target both developer experience and operational reliability.

  • Evaluate Copilot CLI for controlled terminal and CI/CD scenarios because the Rust runtime and session management aim to improve stability and performance.
  • Support mixed Windows/Linux developer environments with experimental WSL support while keeping a single Copilot workflow across platforms.
  • Treat the expanded MCP/plugin/skill/canvas integration as a governance task by defining which internal endpoints and tools Copilot can reach.
05

GitHub announces Copilot billing and unified experience changes

GitHub said it will re-enable new Copilot Business and Enterprise sign-ups via credit card or PayPal starting September 1, 2026. GitHub also announced a unified Copilot experience across web, mobile, and the cloud agent no earlier than September 28, plus a default code review effort change from Lite to Balanced.

  • Re-check procurement and onboarding paths for smaller Czech entities because self-serve purchasing may remove reliance on broader group agreements.
  • Plan change management for a unified Copilot experience because consolidated surfaces typically require updated training and internal support playbooks.
  • Review Copilot code review policies and cost/performance expectations because the default effort level change can alter review depth and runtime for existing teams.