Claude Data Protection Switzerland: How SMEs Use Claude in Line with the revFADP

Jens Margraf ·

Claude can be used in Swiss companies in compliance with data protection law – but not with every plan. What the revFADP requires, where the data ends up and when a Swiss platform is easier.

Monday morning, executive board meeting. Whether to introduce "something like Claude" is on the agenda. What nobody in the room knows: three employees have long been using Claude – with private Pro accounts, fed with quotes, customer e-mails and CVs. This is exactly where the question of Claude data protection in Switzerland begins. Claude is one of the strongest AI models on the market. But buying a licence alone does not make its use compliant with Swiss data protection law.

The good news: it can be done. The less good news: it takes more than choosing the right subscription. This article shows what Swiss SMEs need to watch out for – and when a Swiss platform for Claude is the easier route.

Does data protection law even apply to AI?

Yes, directly. Switzerland has no dedicated AI law, and the Federal Council does not intend to create one modelled on the EU AI Act. However, the Federal Data Protection and Information Commissioner (FDPIC) made it clear as early as November 2023: the revised Swiss Federal Act on Data Protection (revFADP), in force since 1 September 2023, is technology-neutral and applies to any AI-supported processing of personal data.

One detail makes the topic especially relevant for management: unlike under the GDPR, fines under Swiss law are imposed on the responsible individual – up to CHF 250,000 for intentional violations. Anyone who simply lets shadow AI run is therefore taking a personal risk.

Which Claude plan is compliant with the revFADP?

Anthropic, the company behind Claude, separates two worlds: consumer plans and commercial plans. Only the commercial plans include a data processing agreement (DPA) – the contract in which Anthropic commits to processing your data only on your behalf.

PlanContractTraining on your dataSuitable for company data?
Claude FreeConsumer, no DPAuser can optNo
Claude Pro / MaxConsumer, no DPAuser can optNo
Claude TeamCommercial, with DPAexcludedYes, with additional measures
Claude EnterpriseCommercial, with DPAexcludedYes, plus audit logs and custom retention periods
Claude APICommercial, with DPAexcludedYes, for your own applications

Important: Pro and Max are not "Business Light". More computing power does not mean more data protection.

What happens to the data in the USA?

This is the crux. According to Anthropic, Claude requests are processed on US infrastructure – in all plans. There is currently no Swiss or European data residency for claude.ai. Legally, this is a disclosure abroad.

Since 15 September 2024, the USA has been considered a safe country only for companies certified under the Swiss-U.S. Data Privacy Framework. Anthropic is not listed in this official register (as of October 2026). The transfer therefore relies on standard contractual clauses included in the contract with Anthropic. This includes a brief risk assessment: which data goes to the USA, and how sensitive is it?

Think of it like a safe: the contract with Anthropic is the lock. But the safe stands in another country, with different rules for access by authorities.

The five steps to compliant use

  1. Choose a commercial plan. Team, Enterprise or API – and consistently replace private accounts.
  2. Regulate processing on your behalf (Art. 9 FADP). File the DPA and document the list of sub-processors.
  3. Safeguard the disclosure abroad (Art. 16–17 FADP). Document the standard contractual clauses and prepare a risk assessment – Anthropic has no DPF certification.
  4. Create transparency (Art. 19 FADP). Update your privacy policy – including the recipient countries. This is a Swiss particularity that is often forgotten.
  5. Set guardrails. AI policy, SSO, training – and clear rules for connectors and agents.

The last point in particular is underestimated. With connectors via the Model Context Protocol (MCP), a kind of USB-C plug for AI, Claude today accesses e-mails, drives or accounting. Agents such as Claude Cowork complete tasks on their own. As a result, far more personal data reaches Anthropic than employees consciously type in. Minimal permissions and approvals for critical actions are a must here.

Claude direct or via a Swiss platform?

Both routes can be compliant. The question is how much review work remains with you.

Claude Team / Enterpriseaibox (Swiss platform)
Contractual partnerAnthropic, USAAI now AG, Baar
Chats and documents stored inUSASwitzerland
ModelsClaude onlyClaude, ChatGPT, Gemini, Perplexity and Swiss models
Featuresfull Claude world, newest features firstchat, knowledge base, audio transcription, MCP-ready
Effort for disclosure abroadreview and document yourselflow

Claude direct fits if your team specifically and intensively uses Claude and its newest features such as Claude Code or agents – and you can properly document the US transfer. A Swiss platform such as aibox fits if you use several models, want to keep your data in Switzerland and want a contractual partner under Swiss law. For a broader market overview, see our AI platform comparison Switzerland.

Conclusion: it's not the licence that decides, it's the homework

Claude can be used in Swiss companies in a data-protection-compliant way – but only with the right plan, a clean contract, a reviewed disclosure abroad and clear rules for the team. If you don't want to build this yourself, choose a Swiss platform that already does much of this work.

The best time for this is not when the FDPIC comes asking, but now – before the private Pro accounts in your team multiply further.

My question to you: do you know how many of your employees already use Claude or ChatGPT today – and with which account?

Free guide + personal consulting — Download the guide "Using Claude in line with the revFADP" for free. Or have your AI usage reviewed directly by our AI Consulting team.

Download the guide

FAQ

Can Claude be used in Switzerland in compliance with data protection law?

Yes, but not automatically. For personal data in a company, only the commercial plans Claude Team, Claude Enterprise or the API are suitable, because only they include a data processing agreement (DPA). In addition, the disclosure to the USA must be safeguarded, the privacy policy updated and internal rules introduced.

May I use Claude Pro for customer data?

We advise against it. Claude Free, Pro and Max run under consumer terms without a data processing agreement. More performance does not mean more data protection here – for personal data from everyday business, at least Claude Team is required.

Where does Anthropic process Claude's data?

According to Anthropic, on infrastructure in the USA, in all plans. Anthropic currently does not offer data residency in Switzerland or the EU for claude.ai. This constitutes a disclosure abroad under Art. 16 and 17 FADP.

Who is liable for violations of the FADP?

Unlike under the GDPR, fines under Swiss law are imposed on the responsible natural person, not primarily on the company. For intentional violations, such as a disclosure abroad without adequate safeguards, fines of up to CHF 250,000 may be imposed.

Do I need a data protection impact assessment for Claude?

Only if the processing entails a high risk for the data subjects, for example in the AI-supported analysis of health or HR data. For everyday use with clear rules it is generally not mandatory, but a brief risk assessment of the US transfer is always advisable.

Note: This article provides general guidance and does not constitute legal advice. As of October 2026.

Read more

All articles