Does Your Swiss Company Really Need to Comply with the EU AI Act?

Table of Contents

1. The EU AI Act at a Glance: What Already Applies
2. 2 August 2026: What Actually Changed
3. Swiss Companies: Covered in Four Scenarios
4. AI Literacy: the Obligation Almost Everyone Overlooks
5. What About Switzerland? A Deliberately Sectoral Approach
6. Your Five-Step Action Plan
7. Conclusion
8. FAQ

Many executives in French-speaking Switzerland still believe the EU AI Act is not their problem. After all, Switzerland is not part of the European Union, so why worry about a regulation written in Brussels? That reasoning is risky. In fact, the text reaches far beyond EU borders, exactly as the GDPR did before it. In practice, your chatbot, your CV-screening tool or your AI-generated content may well touch the European market. If so, you fall within its scope. Moreover, the 2 August 2026 deadline has just activated new transparency obligations, while the penalty regime is already fully operational. So, does your company really need to comply? The answer depends on four specific scenarios, which we break down in this article. You will also find a concrete action plan to get compliant without panicking.

The EU AI Act at a Glance: What Already Applies

The EU AI Act is the world’s first comprehensive legal framework dedicated to artificial intelligence. It entered into force on 1 August 2024 and applies in successive stages. In other words, part of the regulation has already been producing effects for quite some time.

A Timeline in Successive Waves

Since February 2025, eight practices deemed unacceptable have been banned. They include social scoring, emotion recognition in the workplace and harmful AI-based manipulation. In addition, this first wave already requires employers to ensure their staff have a sufficient understanding of AI. Then, since August 2025, providers of general-purpose AI models have had to document their training data. On top of that, the general penalty regime is active. The full timeline, published by the European Commission, now stretches to 2028.

As for penalties, the regulation does not pull its punches. Indeed, banned practices expose offenders to fines of up to 35 million euros, around 33 million Swiss francs. The fine can also climb to 7 percent of worldwide turnover. That level even exceeds the GDPR.

2 August 2026: What Actually Changed

On 2 August 2026, the transparency obligations set out in Article 50 of the regulation became applicable. In practice, three major changes now affect the daily life of businesses.

Transparency for Chatbots and AI Content

First, your chatbots, conversational agents and virtual avatars must now clearly tell users they are talking to an AI. The only exception covers cases where this is already obvious. Second, content generated or heavily modified by AI must be identifiable as such. Deepfakes, images, audio and text published without human editorial control require technical marking or a visible notice. However, systems placed on the market before that date benefit from a transition period until 2 December 2026.

High Risk Will Wait Until 2027

There is good news, on the other hand, for so-called high-risk systems. Indeed, the Digital Omnibus regulation, published in the Official Journal of the EU in July 2026, postponed the heaviest obligations. Standalone systems for recruitment, credit scoring or education will only face their full requirements on 2 December 2027. As a result, you have a little more time to prepare technical documentation, human oversight and risk management. Still, this delay should not become an excuse to do nothing, because the scale of the work remains considerable.

Swiss Companies: Covered in Four Scenarios

Now for the central question: are you affected from Switzerland? The answer lies in the extraterritorial reach of the text. Under Article 2 of the regulation, providers and deployers established in a third country are also covered. In other words, you do not need an office in Paris or Berlin to fall under the EU AI Act.

Your company falls within the scope of the regulation in the following situations:

• You place an AI system on the EU market, for instance software sold to European customers;
• The output produced by your AI system is used in the EU, even if the system runs in Switzerland;
• You deploy an AI system in a subsidiary or branch established in the EU;
• You act as an importer or distributor of AI solutions intended for the European market.

In practice, a Geneva-based SME whose chatbot answers French customers is therefore covered. Likewise, a fiduciary in Vaud that screens European job applications with an AI tool falls within the scope of the text. Conversely, a company that strictly limits its AI use to the Swiss market escapes the European regulation for now. Nevertheless, that border remains fragile: a website accessible from the EU or a single cross-border client is sometimes enough to tip the balance.

AI Literacy: the Obligation Almost Everyone Overlooks

Among all the obligations of the EU AI Act, one flies strangely under the radar: AI literacy. Yet it has applied since February 2025, well before the deadlines everyone talks about. Specifically, Article 4 of the regulation covers every employee who uses AI systems. They must sufficiently understand how these systems work, along with their limits and risks.

In other words, training your teams is no longer just good practice: it is a regulatory requirement for any company covered by the text. This obligation spans several dimensions. First, understanding what an AI model can do and, above all, what it cannot do. Second, identifying concrete risks: hallucinations, bias, leaks of confidential data. Finally, knowing the company’s internal rules on permitted and forbidden uses.

Moreover, this requirement matches a well-documented reality on the ground. Indeed, studies show that a large majority of employees already use AI at work, often without any framework set by their employer. The result: customer data pasted into consumer tools and AI answers published without verification. Nobody steers that legal exposure. Training therefore remains the fastest lever to turn this diffuse risk into a controlled advantage. Our 90-day method to get your teams AI-ready covers that work step by step.

What About Switzerland? A Deliberately Sectoral Approach

Faced with the European steamroller, Bern has chosen a different path. On 12 February 2025, the Federal Council announced its intention to ratify the Council of Europe Convention on AI. It also favours legal amendments that are as sectoral as possible. In other words, there will be no all-encompassing Swiss AI Act, but targeted adjustments in healthcare, transport and data protection.

The Swiss timetable is moving at its own pace. The Federal Department of Justice and Police is preparing a consultation draft expected by the end of 2026. This draft will define the measures needed on transparency, data protection, non-discrimination and supervision. In parallel, legally non-binding measures, such as self-declaration agreements, will complete the framework.

For companies in French-speaking Switzerland, this approach has two practical consequences. On the one hand, current Swiss law already applies to your AI use. The revised Data Protection Act indeed governs every processing of personal data, AI models included. On the other hand, the gap between Swiss and European requirements will progressively narrow. Consequently, aligning with the core principles of the EU AI Act today is a lasting investment, not wasted effort.

Your Five-Step Action Plan

You do not need an army of lawyers to get started. In practice, a structured five-step approach covers most of the road to compliance.

• Map your AI usage. List every system used in the company, including the consumer tools your teams rely on without formal approval.
• Qualify your exposure. For each use case, determine whether your systems or their output touch the EU market. Then classify them against the regulation’s risk categories.
• Put transparency in place. Label your chatbots as such, mark your AI-generated content and document your editorial choices.
• Train your staff. The AI literacy obligation already applies. Besides, a trained team makes fewer costly mistakes and gets more out of the tools.
• Write a usage policy. A clear document sets out the approved tools, the forbidden data and the required human checks. Finally, appoint someone to track regulatory developments.

This work also fits a broader trend. According to an EY study published in May 2026, artificial intelligence has become an established part of daily business life in Switzerland. However, many organisations are still in the early stages of scaling it. Those who structure their governance now will therefore gain a head start. We already made that point in our article on AI agents in business.

Conclusion

So, does your Swiss company really need to comply with the EU AI Act? Yes, as soon as your AI systems or their output touch the European market. That situation concerns far more companies in French-speaking Switzerland than most people imagine. Since 2 August 2026, transparency for chatbots and AI-generated content is no longer optional. As for the AI literacy obligation, it has applied since 2025. Furthermore, even purely local companies have an interest in structuring their governance, because Switzerland is preparing its own rules for the end of 2026. The right strategy fits in three words: map, train, govern. Start with an honest inventory of your AI usage, teach your teams the fundamentals and give yourselves a clear usage policy. You will turn a regulatory constraint into a lasting competitive edge. Explore our artificial intelligence articles and training courses to equip your teams today.

FAQ

Does the EU AI Act apply to Swiss companies?
Yes, in four scenarios: placing a system on the EU market, output used in the EU, deployment in a European subsidiary, or acting as an importer. A registered office in Switzerland offers no protection.

What exactly changed on 2 August 2026?
Chatbots must disclose that they are AI, and content generated or heavily modified by AI must be identifiable. The European Commission also gained stronger supervision powers over large model providers.

Are the high-risk obligations already in force?
Not yet. The Digital Omnibus of July 2026 postponed them to 2 December 2027 for standalone systems and 2 August 2028 for systems embedded in regulated products.

What is the AI literacy obligation?
Since February 2025, employees who use AI systems must understand how they work, their limits and their risks. Training is the most direct way to meet it.

Will Switzerland adopt its own AI Act?
No, not in that form. The Federal Council favours ratifying the Council of Europe Convention on AI along with sector-specific rules. A consultation draft is expected by the end of 2026.

What penalties does a non-compliant company face?
Banned practices carry fines of up to 35 million euros, around 33 million Swiss francs, or 7 percent of worldwide turnover. Breaches of other obligations follow lower scales but remain dissuasive.