Copilot + Claude, Part 1: Check your compliance before you click enable

Claude is in Copilot. Your data might not be.

A practical guide to the compliance and data residency questions you must answer before enabling Claude.

Microsoft has announced that Claude is coming to Copilot Chat, and while the headlines focus on “AI innovation”, the real story sits somewhere far less glamorous: data processing. Copilot used to be straightforward, Microsoft models running inside Microsoft’s cloud, protected by the Microsoft 365 EU Data Boundary. With Claude, that simplicity disappears. You’re no longer choosing between features; you’re choosing where your data goes.

And enabling Claude means your data leaves the Microsoft 365 EU boundary and may be processed anywhere in the world.

Copilot Chat already uses multiple Microsoft‑hosted models, including GPT, which run inside the Microsoft 365 boundary and follow Microsoft’s compliance commitments. Claude is different. It’s an external model, hosted by Anthropic, and it does not run inside the Microsoft 365 EU Data Boundary. Choosing Claude means your prompts and responses are processed by a model outside Microsoft’s controlled environment, and potentially outside your region. That architectural difference is exactly why enabling Claude becomes a compliance decision, not a feature preference.

What Microsoft announced

Microsoft is adding Anthropic’s Claude as an optional model inside Copilot Chat. This means:

• Users can choose Claude for certain tasks
• Claude will run alongside Microsoft’s own models
• Claude may offer better reasoning or summarisation in some scenarios
• Claude is not a Microsoft model
• Claude does not run inside the Microsoft 365 data boundary

This is not a small detail. This is the detail.

Yes, Claude is optional, but optional features have a habit of becoming enabled unless someone owns the decision.

Where Claude is (and isn’t) available

Before anyone gets too excited: Claude is not available everywhere. In fact, the availability map looks suspiciously like a regulatory risk diagram.

Available: Most commercial regions outside Europe
Not available:

• EU
• EFTA
• United Kingdom
• Government clouds
• Sovereign clouds

So if you’re in Europe and wondering why Claude isn’t showing up in Copilot Chat, the answer is simple:

It’s not missing, it’s intentionally off.

This isn’t a licensing delay. It’s not Microsoft forgetting to flip a switch. It’s because enabling Claude changes your data processing story in ways regulators (and your compliance team) care deeply about.

And yes, European customers can enable Claude, but only by explicitly opting in. The moment you do, your data may be processed outside the Microsoft 365 EU Data Boundary, and your DPIA suddenly needs a new chapter.

In other words:

Claude is optional, but optional features have a long history of becoming “accidentally enabled” unless someone owns the decision.

What this actually means for your data

If you use Microsoft-native models in Copilot Chat:

• Data stays within Microsoft’s cloud
• EU Data Boundary applies
• Purview controls behave predictably

If you enable Claude:

• Your data may be processed outside the Microsoft 365 boundary
• EU-only guarantees may no longer apply
• You now depend on two vendors, not one
• Your DPIA may suddenly be out of date

This is why external models are disabled by default in Europe. It’s not a UX choice. It’s a regulatory hint.

The real architectural decision

This is not about which model is “better”. It’s about:

• Where your data is processed
• Who processes it
• Whether your existing controls still apply

Copilot is no longer something you simply enable. It’s something you architect.

Final thoughts

Claude coming to Copilot Chat is exciting, but excitement doesn’t replace governance.

If you don’t decide which model can touch your data, someone else will. Usually by clicking “enable” because it looked useful at the time.