Which parts of the EU AI Act apply to a Swiss company, and when
The EU AI Act applies to a Swiss company that places AI systems on the EU market, or whose AI output is used in the EU, even without an EU office. After the AI Omnibus, published in the EU’s Official Journal on 24 July 2026, the key dates are: prohibited practices and AI literacy since 2 February 2025, general-purpose AI model rules since 2 August 2025, transparency duties from 2 August 2026, and high-risk rules from 2 December 2027, or 2 August 2028 for AI inside regulated products. Knowlytica maps each of your systems to its role, risk class and date, and builds the documentation; your counsel signs off.
Status: available now. Dates reflect the AI Omnibus as published on 24 July 2026.
- 2 Feb 2025
- Prohibited practices; AI literacy
- 2 Aug 2025
- General-purpose AI models; penalties
- 2 Aug 2026
- Transparency duties (Art. 50); general application
- 2 Dec 2026
- Content marking for systems already on the market; ban on generating non-consensual intimate content
- 2 Dec 2027
- High-risk systems in Annex III areas
- 2 Aug 2028
- High-risk AI inside regulated products (Annex I)
How to check what the EU AI Act requires of your company
Six steps per AI system. Most companies hold more than one role, so the answer is a table, not a single verdict.
- Step 01
Check whether you are in scope
Under Art. 2 the Act covers providers placing AI systems on the EU market wherever they are based, and providers and deployers outside the EU whose system’s output is used in the EU. A Swiss company with EU customers or EU staff using its AI is likely in scope.
- Step 02
Name your role for each system
Provider if you build it or put your name on it; deployer if you use it in your work; importer or distributor if you bring someone else’s system to the EU market.
- Step 03
Classify each system
Prohibited (Art. 5); high-risk (Annex III areas such as hiring, credit scoring and education, or Annex I products such as medical devices); transparency duties for chatbots and generated content (Art. 50); or minimal risk with no specific duties.
- Step 04
Put a date on each duty
Use the timeline above. Transparency duties apply from 2 August 2026; high-risk duties start on 2 December 2027, or 2 August 2028 for AI inside regulated products.
- Step 05
Build the file for high-risk systems
Providers need risk management, data governance, technical documentation, logging, human oversight and accuracy measures (Arts. 9–15), a conformity assessment and EU database registration. A provider outside the EU also appoints an authorised representative in the EU.
- Step 06
Keep AI literacy going
Art. 4, as amended by the Omnibus, asks providers and deployers to take measures that support their staff’s AI literacy. A short training record per team documents it.
Buying an AI tool does not hand over your duties
A vendor’s claim that its product is “AI Act compliant” covers the vendor’s duties as provider. If you use its CV screening tool, you are the deployer of a high-risk system, and Art. 26 gives you a list of your own.
As deployer you must use the system as instructed, assign human oversight to trained people, keep the logs it generates for at least six months, and inform workers’ representatives and affected workers before using it at work. None of that is in the vendor’s contract unless you put it there.
What Knowlytica adds to the EU AI Act’s requirements
The Act says what must exist. These are the registers, controls and files we put in place so it does.
A system register
Every AI system with its role, risk class and the date that applies.
Transparency built in
Chatbot disclosure and machine-readable marking of generated content.
A human oversight step
Agents wait for a trained person’s approval where Art. 14 or Art. 26 calls for it.
Logs kept six months or longer
Stored where you choose and exportable for authorities.
One file for two laws
The same inventory and logs serve the revFADP.
A vendor checklist
Questions for each supplier about the provider duties they carry.
Frequently asked questions
Does the EU AI Act apply if we only serve Swiss customers?
Not if none of your AI systems is placed on the EU market and none of their output is used in the EU. The revFADP still applies in Switzerland.
How much does it cost?
The first 30-minute call is free. The register, controls and files are quoted in a written proposal for the agreed scope.
What isn’t it?
It is not legal advice, not a notified body’s conformity assessment and not a certificate. We prepare the technical file; your counsel and, where required, a notified body decide.
Does it apply to our use of ChatGPT or Claude?
The rules for general-purpose AI models fall on their providers, OpenAI and Anthropic. Your duties depend on the use: a customer-facing chatbot needs a disclosure, and using a model to screen job applicants makes you the deployer of a high-risk system.
Our vendor says its HR screening tool is AI Act compliant. Are we covered?
Only for the vendor’s provider duties. Your deployer duties under Art. 26 remain: trained human oversight, logs kept for at least six months, and informing workers before the tool is used.
What are the fines?
Up to €35 million or 7% of worldwide annual turnover for prohibited practices, and up to €15 million or 3% for most other breaches, whichever is higher. For SMEs, the lower of the two amounts applies.
Also works with
For Switzerland’s own rules, see revFADP compliance for AI assistants. Oversight and logging are built into our custom MCP servers, our Claude rollouts and the systems on AI agents for marketing agencies. Our own controls are on the trust page, and every service is on the Knowlytica home page.