Is Claude GDPR compliant?

Which Claude is meant for work?

Anthropic separates two groups. The consumer plans are Claude Free, Pro and Max. The business offerings are Claude for Work with the Team and Enterprise plans, and the Anthropic API, through which your own software calls Claude.

The consumer terms Anthropic publishes for the EEA apply to people acting mainly outside their trade or profession. In them the user agrees not to use the services for business purposes (version of 8 October 2025).

For your company, that leaves Claude for Work or the API. Both are governed by the Commercial Terms of Service, under which the contracting party for customers in the EEA is Anthropic Ireland, Limited.

Is Anthropic a processor, and where is the DPA?

For the business offerings, yes. Under Anthropic's Data Processing Addendum (DPA), the customer is the controller of personal data and Anthropic its processor (Art. 4(7) and (8) GDPR). The DPA is part of the Commercial Terms: according to Anthropic, accepting them means accepting the DPA.

The GDPR allows such a contract in electronic form (Art. 28(9)). Still check that it covers the points of Art. 28(3), and file the version that applied when you signed up.

From the version of 24 February 2025: Anthropic notifies a security breach affecting your personal data without undue delay and within 48 hours at the latest. It assists you with a data protection impact assessment, and after the contract ends it deletes your data within 30 days, with exceptions such as legal obligations.

Art. 4(7) and (8) · Art. 28(3) and (9) GDPR

Does Anthropic train on your data?

For the business offerings, Anthropic writes that it does not use inputs and outputs for training by default. The Commercial Terms put it more briefly: Anthropic may not train models on customer content from the services.

Feedback is an exception. When someone clicks thumbs up or down, Anthropic stores the whole conversation for up to five years and, by its own account, may also use it for training. Owners of a Team or Enterprise plan can switch this off for everyone with the “Rate chats” setting under “Data and Privacy”.

On Free, Pro and Max, each person decides in their settings whether Anthropic may use their chats for training; if so, Anthropic keeps them for that purpose for up to five years. Chats flagged by safety classifiers may be used for Anthropic's safety models even without that permission.

Where does Anthropic process and store the data?

According to Anthropic, traffic from the business offerings is routed by default to select countries in the US, Europe, Asia and Australia unless agreed otherwise. The data is stored in the US (as of October 2026).

For the API, Anthropic names “global”, the default, or “us” as the place of processing, and currently only “us” as a workspace's storage location. The pages we read describe no setting that keeps data in the EU.

For the transfer, the DPA incorporates the Commission's standard data protection clauses, usually called standard contractual clauses, Modules Two and Three (Art. 46(2)(c) GDPR). The US as destination belongs in your record of processing activities (Art. 30(1)(e)), the safeguard in your privacy notice (Art. 13(1)(f)).

Art. 44 · Art. 46(2)(c) · Art. 13(1)(f) · Art. 30(1)(e) GDPR

An example, and what often goes wrong

Kemper Lohse GmbH, a made-up maker of conveyor belts, wants to use Claude in engineering, for example for maintenance manuals. It books Claude Team, files the DPA with its contracts and enters Anthropic, with the transfer to the US, in its record.

Its internal rule says: technical documents yes, personnel files and health data no. The owner switches off “Rate chats” so that no conversations reach Anthropic through the feedback button.

Common mistakes: engineers keep working with private Pro accounts, although the consumer terms exclude business purposes and Germany's data protection authorities (DSK) advise work accounts (para. 41). Or the company account is in place, but the record and the privacy notice do not yet name the new recipient (Art. 13(1)(e)).

Art. 13(1)(e) · Art. 30(1) GDPR

With an AI assistant

An assistant, Claude itself included, can compare Anthropic's DPA with Art. 28(3), draft an entry for your record and write a usage rule.

Whether the result is right and which data is allowed is decided and answered for by your company (Art. 5(2) GDPR). Give the assistant contracts and policies for this, not personnel data, because its provider processes whatever it reads.

A GDPR walkthrough on nisd2.eu is in the works, with the same MCP server as the NIS 2 walkthrough. Through that server an AI assistant already enters information into the NIS 2 walkthrough today, with the rights of the person who connected it.