Swiss FADP vs EU GDPR AI Services Key Provisions

Swiss FADP Vs EU GDPR AI Services Key Provisions

If your AI service has users in Switzerland and the European Union, you are not dealing with one privacy law. You are dealing with two. Switzerland’s revised Federal Act on Data Protection (nFADP, in force since September 1, 2023) and the EU’s GDPR look similar from a distance, but their key provisions work differently in practice, and those differences matter a lot when the product you run is an AI service that collects prompts, trains on user data, and makes automated decisions.

This guide goes provision by provision. For the broader comparison, start with our overview of the Swiss Federal Data Protection Act vs GDPR for AI — here we focus on the exact rules that change how you design, document, and run an AI product.

Provision 1 – Do You Need a Legal Basis for Processing?

Under the GDPR, every processing activity needs one of six legal bases (consent, contract, legal obligation, vital interests, public task, legitimate interests) before you touch personal data. No basis, no processing.

The nFADP works the other way around. Processing personal data is allowed in principle, as long as you respect the law’s principles — lawfulness, good faith, proportionality, purpose limitation, accuracy, and transparency. You do not need to name a legal basis for ordinary processing. Justification is only needed when processing goes against a data subject’s will in specific situations, such as overriding privacy interests.

For AI services, this changes onboarding design. An EU-facing product usually needs consent flows or a documented legitimate-interests assessment before training or analytics. A Swiss-only product has more room, but the moment the same service targets EU users, the GDPR basis requirement applies in full.

Compliance team reviewing data protection documents for an AI service

Provision 2 – Consent and High-Risk Profiling

Both laws raise the bar for risky processing, but in different places.

The nFADP requires express consent when you process sensitive personal data or carry out high-risk profiling. High-risk profiling means automated evaluation that creates a serious risk for the person’s rights — think an AI that scores creditworthiness, health risk, or job suitability.

The GDPR requires explicit consent for special categories of data and gives people a right not to be subject to purely automated decisions with legal or similarly significant effect (Article 22). The nFADP has a matching rule in Article 21 — the person must be informed about an automated individual decision, can state their position, and can request a human review.

If your AI service scores, ranks, or filters people, both laws effectively force the same design — tell users, get proper consent where required, and keep a human in the loop. Tools built on intelligent automation hit this provision fastest, because scoring pipelines rarely stay “low risk” for long.

Provision 3 – Data Protection Impact Assessments

The GDPR (Article 35) requires a DPIA when processing is likely to create high risk — large-scale profiling, sensitive data, or systematic monitoring. The nFADP (Article 22) requires one in similar situations, especially large-scale processing of sensitive data or systematic surveillance of public areas.

One practical difference — under the nFADP, if the assessment still shows high risk after your safeguards, you must consult the Federal Data Protection and Information Commissioner (FDPIC) unless you have appointed a data protection advisor who has reviewed it. Under the GDPR, you consult the supervisory authority in the same situation. Either way, an AI service launching a new model feature that profiles users should treat a DPIA as a launch checklist item, not paperwork for later.

Provision 4 – Breach Notification Deadlines

This is where the two laws feel most different day to day.

The GDPR gives you 72 hours from discovering a breach to notify the supervisory authority, unless the breach is unlikely to risk people’s rights. The nFADP says notify the FDPIC “as soon as possible”, and only when the breach is likely to create a high risk for the people affected. There is no fixed hour count in Swiss law, but “as soon as possible” is not a loophole — you still need detection, logging, and an escalation path that works on a weekend.

For an AI company, the safest setup is one incident process built to the stricter GDPR clock, applied everywhere. Running two different breach playbooks for the same product is how deadlines get missed.

Provision 5 – Fines and Who Gets Punished

The GDPR fines companies — up to 20 million euros or 4% of worldwide annual turnover, whichever is higher. The nFADP takes a strikingly different approach — its criminal fines (up to CHF 250,000) target the responsible individuals, not the company. A company can be fined up to CHF 50,000 only in narrow cases, such as when identifying the individual would require disproportionate effort.

That flips the internal conversation. Under the GDPR, privacy risk is a balance-sheet problem. Under the nFADP, it is a personal liability problem for the managers and employees who ignored the rules. AI startups selling into both markets need founders to understand that Swiss compliance is not the “light” version — it is personal.

Provision 6 – Data Protection Officers and Representatives

Under the GDPR, many companies must appoint a Data Protection Officer, and non-EU companies targeting EU users usually need an EU representative (Article 27). Under the nFADP, appointing a data protection advisor is voluntary for private companies — but it earns real benefits, including relief from some consultation duties after a DPIA. Foreign controllers who process data of people in Switzerland on a large scale, as part of regular monitoring or offering goods and services, must appoint a representative in Switzerland (Article 14).

So a US or UK AI service can easily end up needing both — an EU representative for the GDPR side and a Swiss representative for the FADP side. Budget for both before you launch, not after a regulator writes to you.

Provision 7 – Cross-Border Transfers and Training Data

The GDPR restricts transfers of personal data to countries without an EU adequacy decision; you then need safeguards like Standard Contractual Clauses and, in practice, a transfer risk assessment. Switzerland benefits from an EU adequacy decision, so EU–Swiss transfers are straightforward. Under the nFADP, transfers abroad are allowed to countries on the Federal Council’s adequacy list (which covers the EU/EEA); for other countries you need contractual safeguards or another exception.

AI services feel this provision in their training pipelines. Prompts and logs sent to a US-based model provider are a “transfer” in both systems. Map where your training and inference data actually flows — the answer decides which clauses and assessments you owe. Our breakdown of AI companies compliance in Switzerland under GDPR walks through the company-side setup in more depth.

And One Layer Above Both – the EU AI Act vs the Swiss Approach

The GDPR and nFADP govern personal data. AI-specific duties sit on top. The EU AI Act adds risk-based obligations for AI systems themselves, phasing in from 2024 to 2027 — prohibited practices, duties for high-risk systems, and transparency rules for chatbots and generative models. Switzerland has chosen a different path — no single AI act. The Federal Council decided in February 2025 to regulate AI through existing sectoral laws plus targeted amendments, and Switzerland signed the Council of Europe’s Framework Convention on AI in March 2025.

Practical result — an AI service in the EU carries three layers (GDPR + AI Act + member-state rules). The same service in Switzerland carries two (nFADP + sectoral rules). The Swiss stack is lighter, but it is not empty, and it is still being written.

Secure data center storing AI service data in Europe

A Working Checklist for AI Services

  • Document your GDPR legal basis for every processing purpose; for Switzerland, document how you meet the nFADP principles instead.
  • Get express consent for sensitive data and high-risk profiling; build human review into automated decisions.
  • Run a DPIA before launching profiling or large-scale sensitive-data features — in both markets.
  • Build one breach process on the GDPR’s 72-hour clock.
  • Appoint representatives where required — EU for GDPR, Switzerland for nFADP — and decide whether a voluntary Swiss data protection advisor is worth it (it usually is).
  • Map every cross-border data flow, including model providers, before you sign the contract.

Frequently asked questions

What is the Swiss FADP in simple words?

The FADP (revised version called nFADP) is Switzerland’s federal data protection law, in force since September 1, 2023. It sets the principles for processing personal data, gives people rights over their data, and is enforced by the FDPIC. It is Switzerland’s counterpart to the GDPR, but it is a separate law with its own rules.

Does the GDPR apply to a Swiss AI company?

Yes, if the company offers services to people in the EU or monitors their behaviour. The GDPR applies based on where your users are, not where your company sits. A Swiss AI service with EU customers usually has to comply with both the GDPR and the nFADP at the same time.

What is the biggest difference between the FADP and GDPR for AI services?

Two stand out. First, the GDPR requires a legal basis before processing; the nFADP allows processing in principle if you follow its principles. Second, GDPR fines hit companies (up to 4% of global turnover), while nFADP fines of up to CHF 250,000 target the responsible individuals personally.

Do AI services need a data protection impact assessment under both laws?

In high-risk situations, yes. The GDPR (Article 35) and the nFADP (Article 22) both require a DPIA for things like large-scale profiling or sensitive-data processing. An AI feature that scores or monitors users should have one before launch in either market.

How fast must an AI company report a data breach?

Under the GDPR, within 72 hours of discovery to the supervisory authority, unless the risk is unlikely. Under the nFADP, you notify the FDPIC as soon as possible, but only when the breach likely creates a high risk for the affected people. Most dual-market companies simply run the 72-hour process everywhere.

Does Switzerland have its own AI law like the EU AI Act?

Not a single one. Switzerland regulates AI through existing sectoral laws with targeted amendments, following a February 2025 Federal Council decision, and it signed the Council of Europe Framework Convention on AI in March 2025. The EU AI Act, by contrast, is one dedicated law with risk-based duties phasing in through 2027.

Leave a Reply

Your email address will not be published. Required fields are marked *