Microsoft is changing the default AI model for Microsoft 365 Copilot customers in the UK. Starting December 9, organizations that haven't set their own preference will have Copilot running Anthropic's Claude model behind the scenes. While it may look like a minor settings tweak, the move points to a bigger shift in Microsoft's model-sourcing strategy.

What happened?

The default engine for UK Microsoft 365 Copilot users is changing on December 9. Any organization that hasn't set a preference by that date will automatically have Claude switched on; administrators can turn this setting off whenever they like. For organizations that don't change the default, request data will be processed on Anthropic's US-based infrastructure.

Why does it matter?

Microsoft's Copilot history is built on its extremely strong tie with OpenAI. The Copilot family, launched in 2023, was for a long time built almost entirely on GPT models, and Microsoft's massive investment in OpenAI only cemented that dependency. But over the past year or two, the picture has gradually shifted: Azure AI Foundry's model catalog has added alternatives like Claude, Llama, Mistral, and DeepSeek, and in GitHub Copilot, developers can choose Claude or Gemini instead of GPT. This default change in the UK is the boldest step yet — because without requiring any action from the customer, it shifts the engine of a product they've already subscribed to over to a competing provider.

Reading this move as simply a favor to Anthropic would be misleading. On the OpenAI side, the dependency is loosening too: as the company transitions to a for-profit structure, it's striking capacity deals with cloud providers like Oracle and CoreWeave outside of Azure. Neither side is holding onto the other as tightly as before. Microsoft's real goal is to be able to route workloads to whichever model best fits performance and cost, rather than being locked into a single supplier's roadmap and pricing policy. That's a standard risk-diversification reflex in enterprise software, not a special gesture toward Anthropic.

Who does this affect, and how?

The real point of contention is data location. Anthropic's Claude processes requests coming from the UK on servers in the US. For Azure and Microsoft 365, Microsoft guarantees EU customers that their data will stay within the EU under its "EU Data Boundary" commitment; but post-Brexit, the UK isn't automatically covered by that program. As a result, organizations subject to UK GDPR and public-sector or financial-sector compliance requirements in particular could find their data crossing the Atlantic simply because they took no action.

StatusDefault modelWhere data is processed
Before December 9OpenAI model familyMicrosoft/OpenAI infrastructure
After December 9, if setting isn't changedAnthropic ClaudeAnthropic's US infrastructure

This table sums up the practical effect of a "default is on, opting out is on you" design. Every organization that doesn't make a conscious choice ends up unknowingly redirecting the flow of its data.

What's next?

There are a few threads worth watching: will Microsoft extend this default change to the EU and US markets as well; will Anthropic offer a local data-processing option for the UK or EU; will organizations in the financial and public sectors revert back to the old setting en masse. Whether OpenAI responds to this development with a countermove on pricing or data processing is another open question.

Framing this change as proof of Microsoft's confidence in Claude would be an overstatement. The more realistic reading is that it's a maneuver by the company to reduce single-supplier dependency. But whatever the rationale, a design where things change automatically unless you take action deserves the attention of organizations sensitive about data location.