The EU AI Act is here. Does it affect my medical practice?

Five practical questions for healthcare businesses in the EU and Switzerland considering or using AI.

The short answer: The EU AI Act introduces obligations for medical practices using AI within its scope in the EU. It may also apply to certain activities of Swiss organisations, including where an AI system's output is used in the EU. Practices operating exclusively in Switzerland are not automatically covered, but Swiss data-protection and professional obligations remain relevant.

AI offers opportunities to simplify everyday healthcare operations. It can help practices respond to routine patient enquiries, coordinate appointments and reduce manual administrative work.

But whether you're exploring AI for the first time, comparing providers or reviewing a system you already use, it's important to understand which requirements may apply and what responsible adoption involves.

The EU AI Act provides a useful starting point. Here are five questions to help you understand what matters for your practice.

1. Does the EU AI Act apply to my practice?

That depends on where your practice operates, its role and how its AI systems are used.

EU only. The Act generally applies when you use AI systems within its scope. Your specific obligations depend on your role and the systems involved.

EU and Switzerland. Assess your EU activities under the Act and consider the relevant legislation in each country where you operate.

Switzerland only. The Act does not automatically apply, but it may be relevant if, for example, the output of an AI system you use is used in the EU.

Note for practices in Switzerland

Switzerland does not currently have an overarching AI law equivalent to the EU AI Act. Instead, existing legislation, including the Swiss Federal Act on Data Protection (FADP), professional obligations and relevant healthcare regulations, continues to apply. Switzerland is developing its own AI regulatory framework, taking the EU AI Act into account while following a different, more sector-specific approach. For Swiss practices considering AI, the EU AI Act can therefore provide a useful reference for understanding risks, transparency and appropriate safeguards, even where it does not directly apply. However, it does not replace existing Swiss legal obligations or guarantee compliance with future Swiss regulations.

Understanding which rules apply to your activities is the starting point for assessing individual AI systems.

2. Does the Act treat every use of AI in the same way?

No. The EU AI Act takes a risk-based approach, with different requirements depending on the system and its intended purpose.

For example, AI that answers questions about opening hours serves a different purpose from AI that assesses how urgently a patient needs medical attention.

Simply using AI in healthcare does not automatically make a system high risk.

When considering a potential application, establish what the technology is designed to do, how its output will be used and whether it falls into a category defined by the Act.

These questions are also useful when comparing providers. A product that includes AI functionality is not necessarily suitable for every healthcare task.

3. Which EU AI Act requirements already matter?

The EU AI Act is being introduced in stages. As of 15 October 2026, the following milestones are particularly relevant to practices:

  • 2 February 2025: AI-literacy requirements and prohibitions on certain AI practices began to apply.
  • 2 August 2026: Most remaining rules began to apply, including requirements to inform people when they interact with certain AI systems.
  • 2 December 2027: Requirements for certain high-risk applications, such as AI used for emergency healthcare patient triage, are scheduled to apply (Annex III).
  • 2 August 2028: Requirements for certain AI-enabled regulated products, such as qualifying diagnostic medical devices, are scheduled to apply (Article 6(1), Annex I).


Specific transitional provisions may affect individual systems. Not every requirement applies to every practice. An AI assistant handling routine administrative enquiries, for example, is not automatically subject to the same requirements as a system used for emergency healthcare patient triage.

Your obligations depend on your role, the system's intended purpose and its legal classification.

4. Where could AI be useful in our practice?

Start with a familiar workflow rather than a list of available AI features.

Consider how patients contact your practice, request appointments, ask for information and receive follow-up communication.

Are there routine enquiries that take up significant staff time? Does your team transfer information manually between communication channels and practice software? Do patients have difficulty reaching the practice outside opening hours?

These situations may reveal opportunities for appropriately designed AI.

If you're already using AI, include your existing systems in the review. Consider what each one does, which information it processes and how it fits into the wider workflow.

The objective is to identify practical needs before deciding which technology to introduce or expand.

5. What should we consider before introducing AI?

Once you've identified a potential use, establish what the system needs to do and what should remain the responsibility of your team.

Consider the information it will process, the output it will produce and what happens after that output is generated.

Identify any uses that interact directly with patients, involve sensitive information or influence consequential decisions. These may require closer legal, clinical or data-protection assessment.

Ask potential providers to explain the system's intended purpose, limitations and relevant safeguards. If you already use AI, review those same questions for your existing systems.

You don't need to become an AI expert to begin exploring the technology. You do need to understand the intended workflow, obtain appropriate advice where necessary and ensure that applicable requirements are met before introducing a system.

What this means for your practice

Choose one routine patient workflow, such as answering enquiries or coordinating appointments.

Identify where AI could make a meaningful difference. If you're already using AI, include the relevant systems in your review.

For each potential or existing use, establish what the technology should do, which information it would process and when a person should be involved.

This gives you a practical starting point for understanding the relevant requirements, assessing potential risks and selecting technology that fits your practice.

Responsible AI adoption begins with a clear understanding of the work you want the technology to perform.

What comes next?

Once you've identified a potential use, the next step is to understand its risks and assess whether the technology is appropriate for the task.

Next article: Is healthcare AI high risk under the EU AI Act?

References


This article provides general information and is not legal advice.

Back