Microsoft 365 Copilot: Anthropic models available when editing Word documents

Microsoft 365 Copilot will add Anthropic AI models as an option for Word document editing, available worldwide by late May 2026. Enabled by default outside the EU, EFTA, and UK, and disabled by default within these regions, admins can control this setting while existing security and compliance standards remain.

[Introduction]

Microsoft will expand AI model options in Microsoft 365 Copilot. With this change, organizations will be able to enable Anthropic models alongside OpenAI models when users edit documents with Copilot in Microsoft Word. This update provides administrators with additional configuration options for AI model providers while maintaining Microsoft’s existing security, privacy, and compliance commitments.

This message is associated with Microsoft 365 Roadmap ID 558440.

[When this will happen]

• General Availability (Worldwide):
  • Rollout will begin in mid-May 2026.
  • Rollout will complete by late May 2026.
• Regional availability and default state
  • EU, EFTA, and UK: Anthropic models will be available but disabled by default and must be explicitly enabled by an administrator.
  • All other regions: Anthropic models will be enabled by default.

[How this affects your organization]

Who is affected

• Microsoft 365 tenants using Microsoft 365 Copilot (Premium)
• Users who edit documents in Microsoft Word using Copilot
• Administrators who manage Copilot and AI provider settings

What will happen

• Anthropic will be available as an AI provider operating as a Microsoft subprocessor.
• Users will be able to select Anthropic models in addition to OpenAI models when using Copilot in Word:

   

• For tenants outside the EU, EFTA, and the UK, Anthropic models will be enabled by default.
• For tenants in the EU, EFTA, and the UK, the feature will be off by default and require explicit admin enablement.
• Administrators can enable or disable Anthropic at any time at the tenant level.
• Existing Microsoft 365 Copilot security, compliance, and governance controls will continue to apply.
• There is no change to the user experience unless Anthropic is enabled (or remains enabled by default, depending on region).

[What you can do to prepare]

• If your tenant is in the EU, EFTA, or the UK and you plan to enable Anthropic:
  • Review your organization’s AI governance, privacy, and compliance requirements.
• If you want to manage or change availability of Anthropic models:
  • In the Microsoft 365 admin center, go to Copilot → Settings → View all.
  • Select AI providers operating as Microsoft subprocessors.
  • Under Available subprocessors for your organization, choose Anthropic and select Enable or Disable.
• Communicate this upcoming change to your security, compliance, and helpdesk teams.
• Update any internal documentation related to Copilot AI model usage.

[Compliance considerations]

Question	Answer
Does the change alter how existing customer data is processed?	Yes. When Anthropic is enabled, Copilot in Word will process document content using Anthropic models to generate responses.
Does the change introduce or modify AI or machine learning capabilities interacting with customer data?	Yes. Anthropic large language models will be introduced as an additional AI option for Copilot in Word.
Does the change provide end users with a new way to interact with generative AI?	Yes. Users will be able to choose Anthropic models when using Copilot in Word, if enabled by an administrator.
Does the change add integration with third‑party software products?	Yes. Anthropic will operate as a third‑party AI provider acting as a Microsoft subprocessor.
Does the change include an admin control?	Yes. Administrators will be able to enable or disable Anthropic at the tenant level through the Microsoft 365 admin center.