Microsoft just handed Anthropic a distribution deal most startups can only dream about. As of June 16, 2026, Claude is available as an alternative AI model inside Microsoft 365 Copilot, giving 420 million monthly active users a one-click path to switching away from GPT. For European organizations, there is a catch: Claude ships disabled by default, and someone needs to flip the switch in your admin center.
What changed on June 16?
Since June 16, 2026, Microsoft 365 Copilot users can select Claude as an alternative to the default GPT models. Two variants are available: Claude Sonnet for everyday tasks and Claude Opus for deeper analysis and structured work.
This is not a wholesale replacement. GPT remains the default in most Copilot features. But where Claude is available, users can now pick their model per conversation. Think of it as Microsoft turning Copilot from a single-brand product into a model marketplace. The company earns the same license revenue whether you use GPT or Claude. What changes is the competitive dynamic underneath.
Worth noting: this is not a sudden pivot. Microsoft began testing Claude in early 2025 through its Frontier program outside Europe. The rollout to EU commercial tenants in June 2026 is the step that matters for organizations covered by GDPR.
Where exactly can you use Claude?
Claude is available in six areas of Microsoft 365. Not all of them give you a choice.
- Copilot Chat: select Claude Opus for complex conversations and long-document analysis.
- Researcher: Opus is available for deep research. The Researcher feature combines multiple sources and produces structured reports.
- Office agentic editing: Claude participates in Word, Excel, and PowerPoint for content editing and generation.
- Copilot Studio: build custom agents with Claude as the underlying model, useful for internal chatbots or automated workflows.
- Copilot Cowork: Claude runs here by default, with no model selection available.
- SharePoint AI: same story, Claude is the default model.
Those last two are worth flagging. In Copilot Cowork and SharePoint AI, Claude is already running whether your users know it or not.
What does it cost?
Nothing extra. Claude is included in your existing Microsoft 365 Copilot license. For organizations with more than 300 users, Copilot runs $30 per user per month. Smaller businesses pay $21 per user per month.
For context: a standalone Claude Pro subscription is $20 per month, and Claude Team is $25 per user per month. Getting Claude bundled with GPT, Teams, SharePoint, and the rest of the Office suite is a different value proposition entirely.
Here's the thing: analysts have flagged that this pricing will not necessarily hold. Microsoft is currently absorbing Anthropic's compute costs. Whether that arrangement survives a Copilot price revision is an open question.
Why is Claude disabled by default in the EU?
This is where European IT teams need to pay attention. Claude is turned off by default for all commercial tenants in the EU, the EEA, and the UK. An IT administrator must manually enable Anthropic models via the Microsoft 365 Admin Center.
The reason is data residency. With standard GPT models in Copilot, Microsoft processes your data within Azure datacenters covered by the EU Data Boundary program. Claude works differently. Anthropic is a subprocessor of Microsoft and processes your data on its own infrastructure at AWS and Google Cloud, with servers located primarily in the United States.
To be concrete: your Word documents, Excel files, and email content leave the Azure ecosystem and travel to an American AI lab. That is exactly why Microsoft does not enable this automatically for European customers.
The EU AI Act adds another layer. For organizations handling high-risk AI applications under the Act's classification framework, routing sensitive workloads through a non-EU subprocessor introduces compliance documentation requirements that most IT teams have not yet built. GDPR's Chapter V restrictions on cross-border data transfers apply regardless of whether Anthropic promises not to train on your data.
What about your company data?
Anthropic will not use your data to train its models. That is spelled out in the Data Processing Addendum that Anthropic and Microsoft signed in January 2026. On that specific point, the arrangement is clean.
But there is an important distinction from how GPT works. With OpenAI's models in Copilot, your data stays within Microsoft's Azure network. With Claude, your data goes to Anthropic's servers outside Azure, specifically to AWS or Google Cloud infrastructure in the US.
Anthropic has not announced plans for European datacenters. If data residency within the EU is a hard requirement for your organization, Claude is not usable today without accepting that constraint.
For most organizations already using US-based cloud services, this is a manageable risk with proper documentation. But it is a deliberate choice your IT department needs to make, not something that should happen without review.
Claude or GPT: which to use for what?
The two model families have different strengths. Here is a practical breakdown by task type.
| Task | Best choice | Why |
|---|---|---|
| Analyzing long documents | Claude Opus | Larger context window, better at maintaining nuance across dozens of pages |
| Quick questions and summaries | GPT | Faster response time, good enough for simple tasks |
| Structured report writing | Claude Opus | Stronger at following complex instructions and specific output formats |
| Email drafting and replies | GPT | Faster, and quality differences on short text are minimal |
| Code and Excel formulas | GPT | Deeper integration with Microsoft products |
| Research and deep reasoning | Claude Opus | Specifically available in the Researcher feature for complex analysis |
The rule of thumb: for fast, everyday tasks, stick with GPT. For work where accuracy and depth matter more than speed, Claude is worth testing.
420 million users, one new competitor on the inside
Microsoft 365 Copilot has 420 million monthly active users globally. That number is what makes this announcement significant for Anthropic. Before June 2026, Claude's reach outside enterprise API contracts ran primarily through Claude.ai and Amazon Bedrock. This integration is a different scale of distribution entirely.
According to TheAIDaily's AI workforce statistics, 37% of workers in AI-augmented roles see average productivity improvements compared to 12% from traditional automation. The question for Copilot users is whether adding Claude to the mix moves that number, or whether default behavior keeps most users on GPT without ever switching.
For context: Microsoft already ran a smaller version of this experiment with 4.7 million GitHub Copilot subscribers, where third-party models from Anthropic and Google have been available since late 2025. The results there gave Microsoft the confidence to expand the model-choice approach to the broader Office suite.
And that broader suite is not just for developers. It reaches the marketers, finance teams, operations managers, and everyone else who runs their work through Word, Excel, and Outlook. The model-choice question stops being a technical preference and becomes an organizational one.
What does this mean for the market?
Microsoft is building Copilot as a platform where AI models compete, not a product tied to a single model. For Microsoft, this is rational: license revenue stays flat whether users pick GPT or Claude. The value is in keeping users inside Microsoft 365, not in which model answers their questions.
For Anthropic, the deal accelerates distribution in a way that no direct-to-consumer campaign could match. Claude was until now primarily accessible through its own channels and Amazon Bedrock. Being baked into the world's most widely deployed office suite changes the adoption curve fundamentally.
For your users, the change brings more choice but also more complexity. Which model for which task? Who decides that at the organizational level? How do you make sure sensitive data does not go to the wrong processor? These are questions most organizations have not yet answered, but they need to be before IT teams enable Anthropic models.
Three concrete steps to take this week
- If you use M365 Copilot: check whether Claude is already available in your Copilot Chat. If the option is not there, ask your IT admin to enable Anthropic models via the M365 Admin Center.
- If you are an IT admin: review whether Anthropic's data processing terms fit within your organization's GDPR documentation. Data goes to Anthropic servers in the US, outside Azure. Consider starting with a pilot group of 5 to 10 users before a full rollout.
- If you handle sensitive data: wait until Anthropic announces European datacenters, or limit Claude to non-confidential documents in the meantime.
Claude in Microsoft 365 is not a revolution in itself. It is the starting point for a world where your employer, not you, decides which AI you use. The vendor of your office suite is becoming the platform on which AI labs compete for your attention.
