Microsoft to Integrate Anthropic AI Models into Word, Offering Copilot Users More Choices
Microsoft announced that it will introduce Anthropic's large language models into Microsoft 365 Copilot in late May, providing users with new options for intelligent refinement and document organization within Word. This integration is described as an AI provider operated by Microsoft as a subprocessor, meaning Anthropic models will operate under Microsoft’s existing security, compliance, and governance framework, while existing OpenAI models will remain available to users.
This update will be automatically rolled out to most Microsoft 365 Copilot (Premium) tenants globally, with the exception of tenants in the European Union, European Free Trade Association (EFTA), and the United Kingdom, where the Anthropic model will not be enabled by default. In these regions, organizations will decide whether to enable Anthropic access, reflecting stricter local data and privacy regulations and providing businesses with more granular control.
For users already subscribed to Microsoft 365 Copilot (Premium), the Anthropic model will appear as a new AI option in the Copilot model selection interface within Word, alongside existing models such as OpenAI. As long as Anthropic is enabled at the tenant level, end users can call different models when writing, refining, or reorganizing documents without additional deployment, to match their individual preferences and content needs. If administrators choose to keep Anthropic disabled, the end-user experience will remain unchanged, and only existing model options will be visible.
Microsoft emphasizes that the introduction of the Anthropic model will not change Microsoft 365 Copilot’s existing commitments to security, compliance, and governance, including data processing paths, access controls, and compliance audit mechanisms. From a technical architecture perspective, Anthropic will be integrated into the Copilot system as an “AI provider” and viewed by Microsoft as its subprocessor. This design helps clarify boundaries of responsibility at the legal and compliance levels and also leaves room for future access to more third-party models.
For organizations located in the EU, EFTA, and the UK, Microsoft provides a clear configuration path: administrators need to log in to the Microsoft 365 admin center, find “AI providers operated by Microsoft as a subprocessor” under “Copilot–Settings–View all,” and then select “Anthropic” from the “Subprocessors available to your organization” list and set it to enabled or disabled. Microsoft recommends that organizations notify their internal help desk teams and update internal documentation in advance to reduce support pressure before and after the feature launches, but with only about two weeks remaining until the official launch, the window for enterprise preparation is relatively limited.
Industry observers are generally concerned about whether this move signifies a further shift in Microsoft’s strategy in the large language model space from a “single partner” to a “multi-model ecosystem.” Neowin points out that after Anthropic joins the Copilot system, observers will focus on whether Microsoft will introduce more models from companies such as Google and Meta in the future, providing businesses and individual users with a wider range of AI choices. This not only concerns product functionality diversity but will also directly affect the new round of competition in the cloud services and office suite markets centered around AI capabilities.
Administrators interested in learning more about the technical details and implementation timeline of this adjustment can consult Microsoft’s official Roadmap ID 558440 for the latest roadmap and change notes:
https://www.microsoft.com/en-US/microsoft-365/roadmap?filters=&searchterms=558440