Summary
- Microsoft is adding Claude Fable 5.1 as a new model choice in Microsoft Copilot, available in Copilot Cowork and Copilot Studio.
- The model is only available to tenants that have enabled Anthropic models under the Microsoft 365 admin center setting for AI providers operating as Microsoft subprocessors.
- Existing user and group access assignments continue to control who can use Anthropic models; there is no separate model-specific opt-in.
- Microsoft states prompts and outputs are not retained by Anthropic for this use, and customer data is not used to train Anthropic foundation models.
- Admins can enable, disable, or limit access to Anthropic models depending on organizational policy.
Primary Service: M365 Copilot
Admin Impact: Low
User Impact: Medium
Release Start: 01 Sept 2026
Release End: 02 Sept 2026
Services: M365 Copilot
Category: Plan for change
Tags: Admin Action, New Feature, User Adoption, Highlighted
History
9/1/2026 Item Added to Message Center
Microsoft Message
Updated September 2, 2026: Tenants in GCC, GCCH and DoD can safely disregard this message. We apologize for any inconvenience.
Microsoft is expanding model choice in Microsoft Copilot by introducing Claude Fable 5.1, a new Anthropic Fable-class model designed to support complex, multi-step tasks.
Fable 5.1 is available to your organization in Copilot Cowork and Copilot Studio with Anthropic operating as a Microsoft subprocessor. Prompts and outputs are not retained by Anthropic when Fable 5.1 is used by your organization. Microsoft’s Product Terms and Data Protection Addendum (DPA) apply, except as otherwise disclosed in Microsoft Learn documentation.
The model will be available only when your organization has enabled Anthropic models under the AI providers operating as Microsoft subprocessors setting in the Microsoft 365 admin center. Your existing admin configuration and user or group assignments will continue to govern access. No additional model-specific opt-in is required. For more information about the model and its capabilities, see the announcement blog.
In the future, Anthropic plans to introduce Enterprise Frontier Safeguards (EFS), a set of controls designed to support automated safety monitoring for its most capable models. When EFS is introduced, any data retention required for these safeguards will occur within Microsoft-controlled infrastructure. If changes to data handling practices are required, Microsoft will provide advance notice before those changes take effect.
Rollout Schedule
- Rollout will begin on September 1, 2026 and is expected to complete by September 2, 2026.
- Availability may vary by region, product experience, and tenant configuration.
- The model will become available in Copilot Cowork and Copilot Studio.
Impact on your organization
Who is affected
- Admins managing access to Anthropic models in Microsoft Copilot and Copilot Studio.
- Users who have been granted access to Anthropic models in supported Microsoft Copilot experiences.
Platforms and services
- Copilot Cowork
- Copilot Studio
What will happen
- Fable 5.1 will be available to your organization with Anthropic operating as a Microsoft subprocessor.
- Prompts or outputs are not retained by Anthropic when Fable 5.1 is used by your organization.
- Microsoft’s Product Terms and Data Protection Addendum apply unless otherwise disclosed in Microsoft Learn documentation.
- Customer data submitted to the model is not used to train Anthropic’s foundation models.
- Existing user and group access selections for Anthropic models operating as Microsoft subprocessors will continue to apply.
- Microsoft is working with Anthropic on Enterprise Frontier Safeguards (EFS), a longer-term approach for supporting access to Anthropic’s most capable models. Microsoft will provide advance notice before any change to your organization’s data-handling practices takes effect.
Action required / recommendations
No action is required if your organization has already enabled Anthropic as a Microsoft subprocessor for the intended users and wants those users to access Fable 5.1.
If Anthropic models are not currently enabled, admins can enable them for all users or selected users and groups in the Microsoft 365 admin center.
If your organization does not want users to access Anthropic models, admins can disable or limit access through the applicable setting in the Microsoft 365 admin center.
Learn more
- Available today: Anthropic Claude Fable 5.1 in Microsoft 365 Copilot
- Anthropic models in Microsoft Online Services