Microsoft has onboarded OpenAI as a subprocessor for Microsoft 365 Copilot and Copilot Studio, enabling access to OpenAI-operated models running on OpenAI infrastructure, separate from the existing Azure OpenAI Service integration.
For that reason, a new setting in the Microsoft 365 admin center lets AI administrators control whether these external OpenAI models are available to their organization.
This new OpenAI subprocessor control applies to Microsoft 365 Copilot, Copilot in Microsoft 365 apps, and Copilot Studio, but does not affect OpenAI models operated by Microsoft (Azure OpenAI).
Timeline
- The new OpenAI subprocessor setting has been available since 9 July and is currently disabled by default.
- The setting will be enabled on 24 July 2026 for all users unless an AI administrator updates it.
How does this affect your organization?
Microsoft has added OpenAI as a subprocessor to provide the OpenAI model families as they become available (beginning with GPT-5.6), alongside the existing Azure OpenAI models. This setting is currently disabled by default until 24 July 2026.
If no admin action is taken before 24 July 2026, the setting automatically switches to enabled for all users. Microsoft will update the setting if an AI administrator has not touched it by then.
An organization should update the setting if it does not want OpenAI-operated models available to its users, or wants to restrict them to specific users.
Administrators must go to the Microsoft 365 admin center > Copilot > Settings > View all > AI providers operating as Microsoft subprocessors and select “No users” or restrict it to specific users/groups. This setting does not affect OpenAI models operated by Microsoft through Azure OpenAI Service, which remain governed separately and are unaffected by this toggle.
Difference between OpenAI models operated by Microsoft and OpenAI as a subprocessor:
Azure OpenAI Service (Microsoft-operated)
These are OpenAI’s models, but they are hosted and run entirely within Microsoft’s own infrastructure. Microsoft operates the deployment, so data processed by these models stays inside Microsoft’s Online Services environment. This is the setup Copilot has used since GPT-4/GPT-5 became the default underlying models for Copilot Chat and Agent Builder.
OpenAI as a subprocessor (OpenAI-operated)
This is a separate delivery path in which OpenAI operates the model infrastructure, with Microsoft acting in an oversight role under contractual terms rather than hosting the model directly.
- Processing happens on OpenAI’s infrastructure, not Microsoft’s.
- Governed by the same Microsoft Product Terms and DPA, with the exclusions disclosed. Notably, the OpenAI subprocessor models are currently excluded from in-country processing commitments where applicable, since the data may leave the region even though Microsoft’s contractual protections still apply. Administrators should read the information on OpenAI as a subprocessor before activating the models.
- Included in the EU Data Boundary, with some exceptions noted in Microsoft’s EU Data Boundary documentation.
- Not available in government or sovereign clouds (GCC, GCC High, DoD, or other sovereign cloud environments). The setting does not appear for those tenants.
- Requires explicit administrative control via the new AI providers operating as Microsoft subprocessors setting. This is the same admin center pattern already used for Anthropic’s Claude models as a subprocessor.
This is also a reminder that enabling or disabling OpenAI subprocessor models affects the external models in the Power Platform admin center and Copilot Studio agents.
