Claude in business: data protection, EU operation, first steps

Anthropic's Claude can be used in a business in line with data protection rules. Which access route fits depends on which data you use and where it may be processed. For most companies, this includes a company account, a data processing agreement and clear rules on which data may be entered. If you need processing in Europe, AWS Bedrock or Google Vertex AI are options.

This article reflects the position as of 15 October 2026. Check the linked provider information before making a decision, as the offerings may change.

Overview of the options

Claude Team and Enterprise

Team and Enterprise offer companies central account management, a data processing agreement and a commitment that inputs are not used for training. According to our information, Anthropic does not commit to processing exclusively in the EU for these options. For many tasks this is acceptable, such as writing text, reviewing code or analysing public information. For personnel data, health data or data subject to professional secrecy, you should examine the conditions more closely.

Running Claude in Europe

For processing in Europe, you can use Claude through a cloud provider:

• AWS Bedrock: Claude models in European AWS regions, billed through your AWS account.
• Google Vertex AI: Claude models in European Google Cloud regions.
• Microsoft Foundry: according to Anthropic, processing in Europe has been announced there.

These routes are mainly suitable for your own applications, for example an AI solution that checks invoices or prepares enquiries. You do not readily get a ready-made chat tool for all employees this way. For that, you need your own user interface.

Claude in Microsoft 365 Copilot

Anthropic models can be enabled in Microsoft 365 Copilot. Microsoft points out that they are currently excluded from the EU Data Boundary. Admins decide whether to enable the models for the organisation. If you chose Copilot precisely because of the EU Data Boundary, you should take this exception into account when enabling them.

Five steps to get started

1. Define the purpose. Decide what Claude is to be used for: writing, analysis, programming or a solution for a specific process.
2. Classify the data. Determine which types of data will be processed. This decides whether Team or Enterprise is sufficient or whether processing in Europe is necessary.
3. Review the contracts. Check the data processing agreement, the subprocessors and the deletion periods. The Datenschutzkonferenz (German data protection authorities) has published guidance on this, including a paper from October 2025 on systems using retrieval-augmented generation, which retrieve additional information from documents or databases for their answers.
4. Write the rules. An AI policy defines which data may be entered and who checks results. A template is in the article AI policy for companies.
5. Train and document. Anyone who uses Claude needs an introduction. How to document it is covered in the article AI training record.

Skills: teaching Claude your company's knowledge

Claude can use skills: packages of instructions, examples and templates for recurring tasks, such as a quote in the company format. In Team and Enterprise, they can be made available to the whole organisation. With our Skill Hub, you can maintain and approve skills for Claude and Microsoft 365 Copilot in one place.

We are happy to clarify which option suits your data in an initial consultation.