Microsoft to Enable Anthrophic Models by Default

Initial Introduction of Anthrophic Models into Microsoft 365 Not Well Handled

For a very large company with sophisticated marketing capabilities, Microsoft sometimes can’t communicate its way out of a wet paper bag. Take the situation around the use of Anthrophic models in Microsoft 365. When this first came to light in September 2025, many expressed concerns because of the lack of controls available when using Claude instead of ChatGPT.

At the time, Microsoft said: “Microsoft’s customer agreements, including the Product Terms and Data Processing Addendum do not apply. In addition, Microsoft’s data‑residency commitments, audit and compliance requirements, service level agreements, and Customer Copyright Commitment do not apply to your use of Anthropic services.”

The lack of auditing detail when Anthropic was used in the Researcher agent was also concerning. Overall, it seemed like the capability was rushed out.

Anthrophic Enabled by Default

Which brings us to message center notification MC1193290 published on December 8, announcing that Microsoft plans to enable the Anthropic models by default for “many customers” (with Microsoft 365 Copilot licenses, in certain regions). The announcement covers the commercial cloud only and doesn’t apply in the government or other sovereign clouds.

The big change is that Microsoft has onboarded Anthrophic as a subprocessor, meaning that Microsoft’s data protection standards (including the data protection addendum) will apply when the Anthrophic models are used with Microsoft 365 Copilot.

Not Enabled by Default in the European Union

The new arrangement is different for customer tenants in the European Union, EFTA, and UK. Likely because Anthrophic processing is done in the U.S., enablement of Anthrophic by default won’t happen in these tenants and it will be up to each tenant to decide whether their data processing arrangements can accommodate AI processing outside the EU Data Boundary.

If, like me, your tenant is in the excluded regions and you previously enabled access to the Anthrophic models, the old administrative toggle (Figure 1) is deprecated and the new toggle (under User access rather than Data access in the settings section of the Copilot area in the Microsoft 365 admin center) must be set.

Why the need for a new toggle? Well, we have a new arrangement because of the extension of Microsoft enterprise data protection to cover the Anthrophic models, so it’s reasonable to ask tenants that are not covered by default enablement to indicate their willingness to use the Anthrophic models under the new regime. At least, that seems to be the logic.

Choice is Good

Getting back to the original announcement, Microsoft could have avoided much of the fuss and bother if they had simply said that they were working with Anthrophic to make sure that Microsoft 365 tenant enjoyed the benefits of LLM choice with the same protection no matter what model was used.

I like the ability to choose from different AI models and think that it’s good for Microsoft to make the necessary arrangements. When the new toggle to enable Anthrophic turns up in my tenant (supposedly on December 8, but not yet), I’ll certainly enable the Anthrophic models because I like the results produced in Researcher. Although I haven’t tried the Anthrophic models in Word, Excel, and PowerPoint, they can’t be any worse in terms of generating or refining text than ChatGPT is.

Even though all AI LLMs are based on inherently flawed material gathered from the internet, it seems like Claude generates better output than ChatGPT. Both are capable of inserting howlers into generated text, but Claude seems to do better. I guess beauty is in the eye of the beholder!

---

Support the work of the Office 365 for IT Pros team by subscribing to the Office 365 for IT Pros eBook. Your support pays for the time we need to track, analyze, and document the changing world of Microsoft 365 and Office 365. Only humans contribute to our work!