Imagine you run an online store in Basel, and on your website, a chatbot answers questions from customers in Freiburg im Breisgau. Switzerland does not yet have an AI law. Nevertheless, under European law, starting August 2, 2026, your customers must be able to tell that they are speaking with an AI. The reason is the EU AI Act.

The AI Act is the EU’s AI Regulation, Regulation (EU) 2024/1689. It has been in force since August 1, 2024, and is being implemented in phases. In July 2026, the EU postponed several deadlines as part of the so-called Digital Omnibus. As a result, many articles online—including our own post from January 2025—are no longer up to date. Here is the current status as of fall 2026.

Does the EU AI Act apply to Swiss companies?

Yes, as soon as your use of AI affects the EU. The AI Act is not based on a company’s headquarters, but rather on where an AI system is offered or where its results are used. You’re probably familiar with this from data protection: The GDPR also applies to Swiss companies that offer goods or services to individuals in the EU.

A Swiss company must comply with the AI Act in particular when

This means that purely internal use of ChatGPT at a trust office in Aarau with no EU clients is not covered. However, a human resources tool that also evaluates job applications for the subsidiary in Milan is definitely covered.

Provider or operator: What is your role?

The AI Act assigns responsibilities based on roles. The two most important ones are:

  • Providers develop an AI system or have one developed and bring it to market under their own name. A Swiss software company that integrates an AI feature into its software for EU customers is a provider.
  • Operators are responsible for their own use of an AI system. A company that uses Copilot or a job application tool is considered an operator.

It is important to note that anyone who distributes a third-party high-risk system under their own name or makes significant changes to it may become a provider themselves. This role determines the scope of their obligations, so it is worth clarifying it early on.

Four Risk Levels

The AI Act regulates AI based on risk. The higher the risk, the stricter the requirements.

  • Prohibited practices: for example, social scoring or emotion recognition in the workplace and at school. Prohibited as of February 2, 2025. Starting December 2, 2026, bans will also apply to intimate deepfakes created without consent and to AI-generated abusive material.
  • High-Risk AI : AI used in sensitive areas such as personnel selection, education, creditworthiness, or critical infrastructure, as well as AI serving as a safety component in regulated products. Strict requirements apply to risk management, data, documentation, and human oversight.
  • Transparency Requirements: Chatbots, AI-generated content, and deepfakes must be identifiable.
  • Low risk: most applications, such as spam filters or text drafts. Aside from the AI-related obligations under Article 4, which apply to all providers and operators, there are no specific requirements here.

Here's an analogy: The AI Act works like the avalanche danger levels. At level one, you can just set out; at level four, you need equipment, planning, and experience, and some slopes are completely off-limits.

Timeline Following the Digital Omnibus

On July 27, 2026, Regulation (EU) 2026/1744—the Digital Omnibus on AI—entered into force. It primarily postpones the obligations for high-risk AI. The transparency obligations, however, remain on their original schedule. The European Commission summarizes the changes, and the AI Act Service Desk provides an overview of all deadlines.

  • August 1, 2024: The AI Act takes effect.
  • February 2, 2025: Prohibited practices and AI competence (Art. 4) take effect.
  • August 2, 2025: Obligations for providers of general-purpose AI models, as well as governance requirements and sanctions, take effect. Fines imposed by the European Commission on providers of such models will not be possible until August 2, 2026.
  • July 27, 2026: The Digital Omnibus Act takes effect.
  • August 2, 2026: The transparency requirements under Article 50 take effect: Label chatbots, disclose emotion recognition, and label deepfakes and certain AI-generated texts.
  • December 2, 2026: End of the transition period for the technical labeling of AI-generated content (Art. 50, para. 2) for generative systems that were on the market before August 2, 2026. On the same day, the new prohibitions on intimate deepfakes created without consent and AI-generated abusive material take effect.
  • December 2, 2027: Requirements for high-risk AI under Annex III—for example, in human resources, education, and critical infrastructure—instead of the originally planned August 2026.
  • August 2, 2028: Requirements for high-risk AI in regulated products listed in Annex I, instead of the originally scheduled August 2, 2027.

In a nutshell: It’s the substantial changes that have been postponed, not the visible ones. You’ll need to label chatbots and AI-generated content starting in August 2026.

What Has Changed in AI Competency

Article 4 originally required providers and operators to do their utmost to ensure that their staff possessed a sufficient level of AI competence. The Digital Omnibus Act has softened this requirement: As of July 27, 2026, they must take measures to support the development of AI competence. No one is required to guarantee a specific level of competence.

What does this mean for you, specifically? The requirement remains, but it’s a duty to make an effort. In most cases, this can be fulfilled through documented training, an AI policy, and designated points of contact within the team. You can find relevant offerings in the overview of AI training programs in Switzerland and among certified AI trainers.

Three Typical Cases from Switzerland

The chatbot on the website. A Zurich-based travel agency has set up its own chatbot using an AI service to answer inquiries from Germany. Because it operates the chatbot under its own name, it is considered a provider, and the chatbot is subject to transparency requirements: As of August 2, 2026, users must be able to recognize that they are speaking with an AI, unless this is already obvious.

The AI-generated images in the campaign. An agency in Lucerne is creating advertising images using AI for a client in Austria, including a realistic-looking image of a real person. If content appears to be a real photo or video of a person, it is considered a deepfake and must be disclosed as AI-generated.

The Application Tool. An industrial company in St. Gallen uses an AI tool that pre-screens job applications, including those for its subsidiary in Italy. This constitutes high-risk AI in human resources. The strict requirements take effect on December 2, 2027; preparations begin with selecting a provider that can demonstrate compliance with the requirements.

Here's how to check if you're affected

  1. Track AI usage. What AI systems do you use or offer?
  2. Clarify the EU connection. Are systems or their results used in the EU—for example, through customers, subsidiaries, or employees?
  3. Determine your role. Are you a supplier, operator, importer, or distributor?
  4. Assign a risk level. Prohibited, high-risk, subject to transparency requirements, or low-risk?
  5. Determine obligations and deadlines. Label immediately; prepare high-risk items by December 2027.
  6. Don't forget Swiss law. The Data Protection Act applies in parallel. Our overview of Switzerland's AI law explains what is currently in effect and what's coming, and our tracker details the consultation timeline.

Consequences of Violations

Fines are based on global annual revenue. For prohibited practices, fines can be up to 35 million euros or 7 percent; for other violations, up to 15 million euros or 3 percent. For SMEs and startups, the lower of the two amounts applies; since the Digital Omnibus, this also applies to small mid-caps, except in cases of prohibited practices. The market surveillance authorities of the member states are responsible, and for general-purpose AI models, the European Commission’s Office for Artificial Intelligence is responsible.

Frequently Asked Questions

Does the AI Act apply in Switzerland?

It is not Swiss law. However, it applies to Swiss companies that offer AI systems in the EU or whose results are used in the EU.

What has been in effect since August 2, 2026?

Transparency requirements under Article 50: Chatbots must identify themselves as AI, emotion recognition must be disclosed, and deepfakes and certain AI-generated texts must be labeled as such. The European Commission has published guidelines and a code of conduct on labeling. A transition period through December 2, 2026, applies to the technical labeling of existing generative systems.

Have the high-risk requirements been postponed?

Yes. Under the Digital Omnibus, these provisions apply to high-risk AI as defined in Annex III effective December 2, 2027, and to AI in regulated products as defined in Annex I effective August 2, 2028.

Do I have to train my employees?

Providers and operators must take steps to support their staff's AI skills. Since the Digital Omnibus Act, this has been a duty to make an effort, with no guaranteed level of proficiency.

Conclusion

For Swiss companies with ties to the EU, the EU AI Act is not just a distant prospect. The labeling requirements have been in effect since August 2026, and the high-risk requirements will be phased in by 2028. Those who take the time now to clarify their inventory, roles, and risk levels will have enough time to prepare without rushing.

In a nutshell:

  • The AI Act applies to Swiss companies if their AI systems or the results of those systems are used in the EU.
  • Transparency requirements for chatbots, deepfakes, and AI-generated content have been in effect since August 2, 2026.
  • The Digital Omnibus (effective July 27, 2026) postpones high-risk obligations until December 2027 and August 2028.
  • AI competence under Article 4 is now a duty to make reasonable efforts.
  • Fines of up to 35 million euros or 7 percent of revenue; for SMEs, the lower of the two amounts.
  • The Swiss Data Protection Act applies concurrently.

Not sure if your company is affected? Bring the question up at the Friday Weekly meeting or get started with the AI Strategy Package.

As of October 6, 2026. This article is intended as a general overview and does not constitute legal advice; the assessment depends on the specific circumstances of each case. It replaces our article dated January 27, 2025.

Sources: AI Act, consolidated version on EUR-Lex · Digital Omnibus, Regulation (EU) 2026/1744 on EUR-Lex · AI Act Service Desk, timeline · European Commission on the entry into force of the Omnibus · Lenz & Staehelin on the implications for Swiss companies