Switzerland is not in the EU, and the EU AI Act is not Swiss law. For a lot of small businesses in Basel, Geneva or St. Gallen, that is where the thinking stops. It is worth going one step further, because the Act follows the product and its output, not the postcode of the company that built it. If your customers, users or partners are in the EU, part of the Act may already apply to you.

This article sets out where things stand after the EU's 2026 amendments, and what a Swiss SME can sensibly do now. It is general information, not legal advice.

Why a Swiss company can be in scope

The scope article of the AI Act covers providers that place AI systems on the EU market whether they are based inside the Union or not. It also reaches providers and deployers located in a third country where the output of the AI system is used in the EU, as the Commission's AI Act Service Desk sets out.

In practice, that gives two common Swiss situations:

  • You are a provider. You sell software, a SaaS product or a device with an AI feature to customers in Germany, France, Italy or Austria.
  • You are a deployer whose output lands in the EU. You use an AI system in your own operations, and what it produces (a chatbot reply, a generated document, a decision) reaches people in the EU.

This mirrors something many Swiss firms already know from data protection. The GDPR applies to companies outside the Union when they offer goods or services to people in the EU, according to Article 3(2) of the GDPR. If you went through that exercise in 2018, you already have a head start on mapping where your AI touches EU customers.

You will not be alone in paying attention. A Thomson Reuters Foundation analysis of nearly 3,000 companies, reported by Euronews, found that 47% of firms referencing the AI Act in their governance disclosures were headquartered outside the EU.

The timeline as it now stands

In May 2026, EU governments and Parliament agreed an "AI omnibus", a package of targeted amendments to the Act, as Euronews reported. It was published as Regulation (EU) 2026/1744 on 24 July 2026 and, according to the European Commission, entered into force on 27 July 2026. The resulting dates, as the Commission lists them:

  • 2 February 2025. The prohibited practices and the AI literacy obligation began to apply.
  • 2 August 2025. Rules for general-purpose AI models applied.
  • 2 August 2026. Transparency obligations, including telling people they are dealing with a chatbot, began to apply, and national authorities and the AI Office took up enforcement.
  • December 2026. A new prohibition on AI systems that generate non-consensual sexually explicit content takes effect.
  • 2 December 2027. High-risk obligations apply to the sensitive use cases in Annex III, such as employment and education.
  • 2 August 2028. High-risk obligations apply to AI built into regulated products under Annex I.

The headline is that the heaviest obligations, for high-risk systems, have moved back by more than a year. Most of what touches an ordinary SME has not moved at all.

What changed for smaller firms

The omnibus included some relief aimed at smaller businesses. According to Euronews, it brings simplified technical documentation and broader access to regulatory sandboxes for SMEs and small mid-caps, and narrows the high-risk category so that tools which merely assist users are no longer automatically caught. The Commission confirms that some simplified requirements once reserved for SMEs now extend to small mid-cap companies.

The AI literacy duty was also reworded. The amended text asks providers and deployers to take measures to support the AI literacy of their staff, and states that this does not require them to guarantee any specific level of literacy for any individual, according to the published regulation. That is a more workable standard for a twenty-person firm, but it is still an obligation.

What has not moved

Three things did not get postponed:

  1. The prohibitions. The Commission's list of banned practices includes social scoring and emotion recognition in workplaces and education.
  2. Transparency. Chatbot disclosure and labelling duties have applied since August 2026.
  3. Penalties. The most serious breaches can reach €35 million or 7% of global annual turnover, as Euronews notes. Fines are not the only reason to comply, but they show how seriously the EU takes the Act.

A checklist for the next quarter

  • List every AI system that touches EU customers. Website chat, email drafting, lead scoring, product features. Note whether you built it, bought it or embedded it.
  • Decide your role for each. Provider (you put it on the market) or deployer (you use it). The obligations differ.
  • Rule out prohibited uses. Pay particular attention to anything that infers emotions from staff.
  • Fix chatbot disclosure now. It is already in force; see our note on chatbot transparency.
  • Record your AI literacy measures. A short internal guideline and a training log are a reasonable start.
  • Flag Annex III areas early. If you use AI to screen CVs or manage staff, plan for December 2027 now rather than in 2027.
  • Ask suppliers for their AI Act position. Most SMEs are deployers of someone else's model, and that supplier's documentation becomes part of your file.

The sensible posture

For most Swiss SMEs, the AI Act is less about heavy compliance programmes and more about knowing what you run, where its output goes and whether people are told when they are talking to a machine. The extra time on high-risk rules is useful, but it is time to prepare, not a reason to wait.

Not sure which of your tools reach EU customers? The free AI Readiness Audit maps your AI use and the data flows behind it.

Sources

  • European Commission AI Act Service Desk, "Article 2: Scope", accessed 7 November 2026. link
  • European Commission, "AI Act" (regulatory framework page), last updated 19 October 2026. link
  • EUR-Lex, "Regulation (EU) 2026/1744 (Digital Omnibus on AI)", 24 July 2026. link
  • EUR-Lex, "Regulation (EU) 2016/679 (General Data Protection Regulation)", 27 April 2016. link
  • Euronews, "The EU simplified its toughest AI law: what changed and why it matters", 22 May 2026. link
  • Euronews, "How the EU AI Act is reshaping company rules from Washington to Tokyo", 30 July 2026. link