GDPR check

Is Claude GDPR compliant? What universities should check

Claude, Anthropic's AI assistant, is widely used for writing, coding and research, and Anthropic now sells Claude for Education to universities. This page sets out what Anthropic publicly documents for each plan, what that means for a university acting as controller, and what to ask before an institutional rollout.

Published 7 October 2026 · Sources checked 7 October 2026

Short answer

Claude is made by Anthropic. For people in the EEA, Anthropic Ireland, Limited (Dublin) is the controller under the privacy policy, and it is also the contracting party for commercial customers in the EEA, Switzerland and the UK. Anthropic sells consumer plans (Free, Pro and Max) and commercial plans (Team, Enterprise, the API and Claude for Education), and the data protection position differs sharply between them. On consumer plans Anthropic may train on conversations unless the user opts out. On commercial plans Anthropic acts as processor under a public DPA with SCCs, does not train on customer content by default, and holds ISO 27001, ISO 42001 and SOC 2 reports. Data is stored in the US, and we found no EU residency option. Whether a university can use Claude depends on the plan, the contract and what people put into it.

Prefer a tool built in the EU? Kahubi, from Avidemic AB in Sweden, covers an AI assistant for research work with EU hosting and, for institutions, only European subprocessors. See how Kahubi handles research data

What Claude documents publicly

Everything below comes from Anthropic's own legal pages, privacy centre, trust centre and developer documentation, and from the official Data Privacy Framework List, all read on 7 October 2026. "Commercial plans" means Team, Enterprise, the API and Claude for Education. "Not found publicly" means we did not find it in the sources listed; it does not mean the safeguard does not exist.

TopicWhat the vendor statesSource
Company and establishmentDocumented The privacy policy (effective 10 September 2026) names Anthropic Ireland, Limited, 6th Floor, South Bank House, Barrow Street, Dublin 4, as controller for EEA users, with a DPO reachable at dpo@anthropic.com. The Commercial Terms name Anthropic Ireland, Limited as the contracting party for customers in the EEA, Switzerland or the UK.[1], [3]
Where data is stored and processedPartly documented Anthropic says it uses several cloud providers with processing in "select countries in the US, Europe, Asia and Australia", and that "data is stored in the US". For the API, the only workspace storage location is the US, and inference can be set to "global" (the default) or "us". Enterprise plan users on usage-based billing can select a traffic routing location. Not found publicly An EU storage or EU-only processing option.[12], [13]
Data processing agreementPlan-dependent A public Data Processing Addendum (effective 24 February 2025) is automatically incorporated into the Commercial Terms and covers Claude for Work and the API. Anthropic acts as processor. It does not apply to consumer Free, Pro or Max accounts, where Anthropic is controller under its privacy policy.[4], [15], [1]
SubprocessorsDocumented A public list in the trust centre with 20 entries. Google Cloud, AWS and Microsoft Azure provide cloud infrastructure listed as "Worldwide". Others include Cloudflare, Stripe, WorkOS (single sign-on), Intercom, Brave Search (web search), ElevenLabs (text to speech for Claude for Work) and support providers in the US, South Africa and Canada. Under the DPA, customers get notice of new subprocessors and 15 days to object.[5], [4]
International transfersDocumented The privacy policy relies on adequacy decisions and Standard Contractual Clauses. The DPA uses the 2021 SCCs (Modules 2 and 3) governed by Irish law, plus UK and Swiss addenda. We found no entry for Anthropic on the official Data Privacy Framework List.[1], [4], [18]
AI model training on customer contentPlan-dependent Consumer plans: Anthropic "may use your Inputs and Outputs to train" its models "unless you opt out"; the setting is called "Help improve our AI models", opting out does not remove data from training already under way, and conversations flagged by safety classifiers may still be used for safety models. Commercial plans: "we will not use your inputs or outputs from our commercial products to train our models", unless a user submits thumbs up or down feedback or the organisation opts in. The Commercial Terms say Anthropic "may not train models on Customer Content".[1], [7], [9], [8], [3]
Retention and deletionPlan-dependent Consumer: deleted chats leave back-end storage within 30 days; with model improvement on, data may be kept de-identified in training pipelines for up to five years. Commercial: deleted chats leave back-end storage within 30 days; API inputs and outputs are deleted within 30 days. In both cases, content flagged for policy violations may be kept up to two years (classification scores seven years) and feedback five years. Enterprise admins can set custom retention. The DPA provides for return or deletion within 30 days after the contract ends.[10], [11], [16], [4]
Security certificationsDocumented ISO 27001:2022, ISO/IEC 42001:2023, SOC 2 Type I and Type II, and a HIPAA-ready configuration, for commercial products such as Claude for Work and the API. Reports are requested through the trust centre, so their scope is not public.[6]
Institution and enterprise controlsPlan-dependent Team (2 to 150 people) includes SSO and domain verification. Enterprise adds SCIM, audit logs, role-based access and custom data retention. Claude for Education accounts are governed by the agreement between Anthropic and the university, whose Primary Owner manages accounts and data, with Anthropic acting as processor.[16], [14]

What this means for a university

Validemic's analysis

The plan decides the legal set-up. On Free, Pro or Max, Anthropic is the controller and the university has no contract with it. Staff using personal accounts for university work cannot be covered by institutional instructions, retention rules or a training opt-out set centrally; the individual's own setting decides whether conversations feed into training. On Team, Enterprise and Claude for Education, Anthropic is a processor under the DPA, training on customer content is excluded by the Commercial Terms, and the institution controls membership and (on Enterprise) retention. Article 28 GDPR requires a processor contract [22], which only the commercial plans provide.

Claude for Education is a contract, not a discount. Anthropic describes Claude for Education as an agreement with the university, under which Anthropic acts as processor and the university's Primary Owner manages the data [14]. That is the right structure for institutional use. The Education help article does not itself mention training, so ask Anthropic to confirm in the contract that the commercial no-training terms apply in full.

Transfers are the main open point. Anthropic Ireland is the EU counterparty, but Anthropic says data is stored in the US, and developer documentation offers only US or global inference, with US the only storage location [12], [13]. We found no entry for Anthropic on the DPF List [18], so transfers rely on the SCCs in the DPA. A transfer impact assessment is the expected next step. Support providers in South Africa and Canada are further destinations to note [5].

Feedback is a training route on every plan. On commercial plans, a thumbs up or down sends the whole conversation for possible training and five-year retention [8]. Enterprise and Team owners can switch off the "Rate chats" setting [8]; a university handling research data should consider doing so. On Claude for Education, Anthropic says thumbs feedback is disabled by default [14].

Students. The consumer terms require users to be at least 18 or the local age of digital consent, whichever is higher [2]. Students below 18, for example on bridging courses, cannot use consumer Claude under those terms; an institutional agreement should say how they are handled.

Credit where due. A named EU controller and contracting entity, a public DPA with SCCs, a public subprocessor list, a clear contractual no-training commitment for commercial customers, and ISO 27001, ISO 42001 and SOC 2 Type II reports are a solid basis for a vendor assessment.

Reviewing a vendor right now? Validemic checks the vendor's documents against GDPR and the EU AI Act and cites every finding. Try the demo workspace

Questions to ask Anthropic before approving Claude

  1. Can you offer EU storage or EU-only inference for Claude for Education or Enterprise, and if not, is it on the roadmap?
  2. Does the commercial no-training commitment apply in full to Claude for Education, and can the feedback route be switched off for all users?
  3. Which subprocessors process conversation content (not only account data) for our workspace, and in which countries?
  4. What transfer impact assessment do you provide for transfers to the US under the SCCs?
  5. Can we set a short retention period for all users, and how do we export conversations for access requests?
  6. What is the scope of the SOC 2 Type II report and the ISO 27001 and ISO 42001 certificates: do they cover the Claude apps used by Education customers?
  7. How do you handle staff who already have personal Pro accounts on a university email address when we roll out an institutional plan?
  8. Which connectors and integrations (for example a learning management system) can admins disable, and what data do they pull in?

The EU AI Act angle

Using Claude for writing, coding or research support is normally not a high-risk use under the AI Act, Regulation (EU) 2024/1689 [19]. Annex III, point 3 lists the education uses that are high-risk: deciding admission or access, evaluating learning outcomes, assessing the level of education a person will receive, and monitoring students during tests. Using any tool for those purposes would change the assessment.

A university that rolls out Claude is a deployer. Article 4 on AI literacy has applied since 2 February 2025. Regulation (EU) 2026/1744 (the Digital Omnibus on AI), in force since 27 July 2026, replaced that article with a duty to take measures to support AI literacy, and moved the application date of the Annex III high-risk rules to 2 December 2027 [20]. Obligations for general-purpose AI models have applied since 2 August 2025 and fall on Anthropic as model provider. Anthropic is listed by the European Commission as a signatory of the General-Purpose AI Code of Practice [21].

Sources

  1. Anthropic, Privacy Policy (effective 10 September 2026), retrieved 7 October 2026
  2. Anthropic, Consumer Terms of Service (effective 8 October 2025), retrieved 7 October 2026
  3. Anthropic, Commercial Terms of Service (effective 17 June 2025), retrieved 7 October 2026
  4. Anthropic, Data Processing Addendum (effective 24 February 2025), retrieved 7 October 2026
  5. Anthropic Trust Center, Subprocessors, retrieved 7 October 2026
  6. Anthropic Privacy Center, "What Certifications has Anthropic obtained?", retrieved 7 October 2026
  7. Anthropic Privacy Center (consumers), "Is my data used for model training?", retrieved 7 October 2026
  8. Anthropic Privacy Center (commercial customers), "Is my data used for model training?", retrieved 7 October 2026
  9. Anthropic Privacy Center, "How do I change my model improvement privacy settings?", retrieved 7 October 2026
  10. Anthropic Privacy Center, "How long do you store my data?", retrieved 7 October 2026
  11. Anthropic Privacy Center, "How long do you store my organization's data?", retrieved 7 October 2026
  12. Anthropic Privacy Center, "Where are your servers located? Do you host your models on EU servers?", retrieved 7 October 2026
  13. Claude Developer Platform documentation, "Data residency", retrieved 7 October 2026
  14. Anthropic Privacy Center, "Who owns and manages the data of my Claude for Education account?", retrieved 7 October 2026
  15. Anthropic Privacy Center, "How do I view and sign your Data Processing Addendum (DPA)?", retrieved 7 October 2026
  16. Claude, Pricing, retrieved 7 October 2026
  17. Claude, Higher education, retrieved 7 October 2026
  18. U.S. Department of Commerce, Data Privacy Framework List (searched for "Anthropic"; no result), retrieved 7 October 2026
  19. Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 4 and 113 and Annex III, Official Journal text read via the Publications Office, retrieved 7 October 2026
  20. Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal text read via the Publications Office, retrieved 7 October 2026
  21. European Commission, The General-Purpose AI Code of Practice (signatories), retrieved 7 October 2026
  22. Regulation (EU) 2016/679 (General Data Protection Regulation), Article 28, retrieved 7 October 2026

About this page

We read Anthropic's privacy policy, consumer and commercial terms, DPA, trust centre, privacy centre articles, pricing page and developer documentation, searched the official Data Privacy Framework List, and read the EU legal texts on 7 October 2026. We did not test the product or review any non-public contract. This page describes public documentation only. It is not legal advice and not a verdict on whether Anthropic or any university's use of Claude complies with the GDPR; that depends on your plan, your contract and your use.

If you work at Anthropic and something here is out of date or incomplete, please contact us and we will review it promptly.

Frequently asked questions

Does Claude train on my conversations?

It depends on the plan. On the consumer Free, Pro and Max plans, Anthropic's privacy policy says it may use inputs and outputs to train its models unless you opt out in your account settings. For commercial products such as Team, Enterprise and the API, Anthropic says it does not use inputs or outputs for training, except where a user submits feedback or the organisation opts in (checked 7 October 2026).

Which Anthropic company is responsible for EU users?

Anthropic's privacy policy names Anthropic Ireland, Limited in Dublin as controller for users in the EEA. The Commercial Terms say that Anthropic Ireland, Limited is the contracting party for commercial customers in the EEA, Switzerland and the UK.

Does Anthropic sign a DPA for Claude?

Yes, for commercial products. Anthropic's Data Processing Addendum, with Standard Contractual Clauses, is incorporated into its Commercial Terms, which cover Claude for Work (Team and Enterprise) and the API. It does not apply to consumer Free, Pro or Max accounts.

Is Claude data stored in the EU?

Anthropic's privacy centre states that data is stored in the US. For the API, the only workspace storage location currently offered is the US, and inference can be set to global or US only. We found no EU storage or EU-only processing option in public documentation (checked 7 October 2026).

Is Anthropic certified under the EU-US Data Privacy Framework?

We found no entry for Anthropic when searching the official Data Privacy Framework List on 7 October 2026. Anthropic's privacy policy and DPA rely on adequacy decisions and Standard Contractual Clauses for transfers instead.

Related reading