The Swiss approach to AI regulation

The Federal Council’s decision of 12 February 2025

For several years Switzerland watched the development of the EU AI Act without committing itself to a horizontal statute of its own. The Federal Council therefore commissioned an overview of possible regulatory approaches, prepared under the lead of the Federal Office of Communications. The resulting analyses covered the legal situation in Switzerland, the sectoral regulation already in place and a comparison with other countries, including the implications of the EU AI Act and of the Council of Europe’s Framework Convention on Artificial Intelligence. On 12 February 2025 the Federal Council took its decision. Switzerland would ratify the Council of Europe Convention and make the necessary amendments to Swiss law. The Federal Council preferred sector-specific amendments wherever possible, while general cross-sectoral rules were to be limited to areas relevant to fundamental rights, such as data protection. The approach combines legally binding measures with non-binding ones such as self-declarations and industry solutions.¹

The Federal Council named three objectives: to strengthen Switzerland as a location for innovation, to protect fundamental rights including economic freedom, and to increase public trust in AI.² It instructed the Federal Department of Justice and Police, together with the departments responsible for communications and foreign affairs, to prepare a consultation draft by the end of 2026. The draft is to implement the Convention in the areas of transparency, data protection, non-discrimination and supervision. A plan for the non-binding measures was to be developed by the same date. On 27 March 2025 Federal Councillor Albert Rösti signed the Convention on behalf of Switzerland in Strasbourg.³

The state of the work in September 2026

As at September 2026, the consultation draft has not yet been published. In its answer to a parliamentary question from the National Council in spring 2026, the Federal Council confirmed that the Federal Department of Justice and Police was preparing a draft for submission by the end of 2026. It explained that it relied on a combination of binding and non-binding measures for private actors. It also stated that it pursued three equal objectives, namely the protection of fundamental rights, public trust and the competitiveness of Swiss innovation, and that it sought a level of protection equivalent to that of the European Union.⁴ In February 2026 the Federal Council’s advisory board on Digital Switzerland discussed the implementation, with a focus on self-regulation in the media and energy sectors. Federal Councillor Rösti stressed transparency and traceability, while participants noted that legislative projects of this kind usually take several years.⁵

The Council of Europe Convention applies primarily to activities of public authorities. For the private sector, each party may decide how to address the risks, whether by applying the Convention’s obligations directly or by other appropriate measures.⁶ This flexibility is what allows Switzerland to pursue its sectoral approach. For financial institutions it means that the most important Swiss rules on AI are likely to remain those of financial market law, data protection law and FINMA practice, supplemented by any transparency or non-discrimination provisions that the forthcoming bill introduces.

The relevance of the EU AI Act for Swiss firms

The EU AI Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024. Its prohibitions of unacceptable practices have applied since 2 February 2025, and the obligations for providers of general-purpose AI models since 2 August 2025.⁷ Because Switzerland is a member of neither the EU nor the European Economic Area, the AI Act does not apply on Swiss territory as such. Its territorial scope nevertheless reaches Swiss firms in two ways. It applies to providers that place AI systems on the Union market or put them into service there, irrespective of where the provider is established. It also covers providers and deployers established in a third country where the output produced by the AI system is used in the Union.⁸

For a Zurich bank with clients in Germany, France or Italy, the second limb is the more important. A credit scoring model run in Zurich whose results are used to decide on a loan to a client resident in the Union may fall within the Act. The list of high-risk systems in Annex III includes AI intended to evaluate the creditworthiness of natural persons or establish their credit score, with an exception for fraud detection. It also includes AI used for risk assessment and pricing of life and health insurance for natural persons, and AI used in recruitment and the selection of candidates.⁹ Insurers in Zurich that price life or health cover for policyholders in the Union and banks that assess the creditworthiness of private clients there are therefore directly concerned.

The timetable for high-risk systems changed in 2026. The Digital Omnibus on AI, published in the Official Journal on 24 July 2026 and in force since 27 July 2026, postponed the application of the rules for high-risk systems under Annex III from 2 August 2026 to 2 December 2027. For AI embedded in products regulated under Annex I, the date moved to 2 August 2028. The remaining transparency obligations continue to apply from 2 August 2026, although systems placed on the market before that date have until 2 December 2026 to comply with the labelling of synthetic content.¹⁰ For Swiss institutions the postponement gives time, but it does not change the substance of what will be required. For many Zurich groups it is practical to apply the AI Act’s classification and documentation methods group-wide, because maintaining separate standards for EU and non-EU business would be costly and because FINMA’s own expectations on inventory, classification and testing point in the same direction.

The relationship between the two regimes also has a strategic dimension. The Federal Council’s aim of equivalent protection is meant to prevent the Swiss framework from becoming an obstacle to market access. For a financial centre whose institutions sell services into the Union mostly under national cross-border regimes, a Swiss AI law that is compatible with the AI Act reduces compliance costs. A Swiss law that goes further than the AI Act would, by contrast, weaken the argument that Switzerland offers a lighter regulatory environment. Geneva, the seat of the International Telecommunication Union and host since 2017 of the “AI for Good” Global Summit, which Switzerland co-convenes, provides the diplomatic stage on which Switzerland presents this approach internationally.¹¹ Zurich supplies the research and the industrial users that give the approach substance.

Notes

1. Federal Office of Communications (OFCOM/BAKOM), “Künstliche Intelligenz”, https://www.bakom.admin.ch/de/kuenstliche-intelligenz (accessed September 2026). ↩
2. Swiss Federal Council, “KI-Regulierung: Bundesrat will Konvention des Europarats ratifizieren”, press release of 12 February 2025, https://www.admin.ch/de/nsb?id=104110. ↩
3. Swiss Federal Council, “Switzerland signs Council of Europe Convention on Artificial Intelligence”, 27 March 2025, https://www.admin.ch/en/nsb?id=104646. ↩
4. Swiss Parliament, Curia Vista, interpellation 26.3414 “Umsetzung der KI-Konvention des Europarates in der Schweiz”, with the answer of the Federal Council, http://ws-old.parlament.ch/affairs/20263414 (accessed September 2026). ↩
5. Swiss Federal Council, “Beirat Digitale Schweiz: Massnahmen zur Umsetzung der KI-Konvention des Europarats”, 10 February 2026, https://www.admin.ch/de/newnsb/5oOec_8ZQEV1X-55waGOg. ↩
6. Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225), Article 3(1)(a) and (b), text at https://rm.coe.int/1680afae3c. Paragraph 1(a) covers activities of public authorities or private actors acting on their behalf, while paragraph 1(b) requires each party to address the risks arising from activities of private actors in a manner conforming with the object and purpose of the Convention. ↩
7. Gibson Dunn, “EU AI Act Omnibus Agreement — Postponed high-risk deadlines and other key changes”, May 2026, https://www.gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes/. ↩
8. EU Artificial Intelligence Act, Article 2(1)(a) and (c), https://artificialintelligenceact.eu/article/2/ (accessed September 2026). ↩
9. EU Artificial Intelligence Act, Annex III, points 4(a), 5(b) and 5(c), https://artificialintelligenceact.eu/annex/3/ (accessed September 2026). ↩
10. Lewis Silkin, “The Digital Omnibus on AI enters into force today”, 27 July 2026, https://www.lewissilkin.com/insights/2026/07/27/the-digital-omnibus-on-ai-enters-into-force-today-102nedo. ↩
11. Federal Department of Foreign Affairs, “AI for Good Global Summit 2026”, 8 July 2026, https://www.europa.eda.admin.ch/en/newnsb/CLgBwa65JwN9DjD02EOI9. ↩

For general information only. This article does not constitute legal or tax advice, does not replace advice on an individual case and creates no lawyer-client relationship. It reflects the legal position as of September 2025.