Switzerland releases AI regulation report with three governance approaches

Switzerland published a comprehensive analysis of artificial intelligence regulation options on February 12, 2025, marking a significant development in the country's approach to emerging technology governance. The Federal Department of the Environment, Transport, Energy and Communications DETEC and Federal Office of Communications OFCOM released the report after the Federal Council mandated an examination of regulatory approaches in November 2023.

The 37-page document presents three distinct regulatory pathways. The first approach would continue sector-specific regulatory activities without overarching coordination. The second approach involves ratifying the Council of Europe's AI Convention through either minimum or extensive implementation. The third approach combines AI Convention ratification with comprehensive product regulation aligned with the EU AI Act.

According to the report, Switzerland currently has no overarching AI-specific legislation despite AI applications permeating multiple economic sectors. The analysis examined 20 countries across six continents between spring and autumn 2024, revealing diverse international approaches to AI governance. Apart from the 27 EU Member States where the AI Act entered into force on August 1, 2024, few countries had implemented legally binding AI-specific instruments as of autumn 2024.

The Council of Europe's AI Convention represents the world's first intergovernmental agreement on AI binding between contracting parties. The Committee of Ministers adopted the framework on May 17, 2024, with Switzerland chairing negotiations involving 57 countries from almost all regions. When opened for signature on September 5, 2024, the EU, Andorra, Georgia, Iceland, Israel, Moldova, Norway, San Marino, the UK and the US already signed the convention.

The convention establishes general obligations for states to guarantee protection of human rights, integrity of democratic processes and respect for rule of law throughout AI system lifecycles. It defines principles including human dignity, individual autonomy, transparency, oversight, accountability, responsibility, equality, non-discrimination, privacy, personal data protection, reliability and safe innovation.

Switzerland's baseline legal analysis identified specific implementation requirements. For transparency and oversight provisions under Article 8, Swiss law does not go far enough compared to convention obligations. Adjustments would also be necessary for safe innovation provisions under Article 13, remedies under Article 14 and procedural safeguards under Article 15.

For risk and impact management frameworks required by Article 16, Swiss law provides no corresponding standards. The legal order has only regulated selective aspects, such as impact assessment requirements in the Data Protection Act. New legal bases would need creation to meet convention obligations.

The analysis shows that for certain convention provisions, Swiss law offers sufficient protection without requiring adjustments. This includes integrity of democratic processes and respect for rule of law under Article 5, and public consultation requirements on important AI questions under Article 19.

The Federal Office of Justice conducted baseline legal analysis covering developments through August 31, 2024. The examination concluded that ratification would require legislative adjustments affecting several laws due to breadth of legal areas concerned. Coordination with the Data Protection Act appears particularly necessary.

Switzerland's regulatory principles emphasize that policymakers should provide optimal frameworks conducive to innovation for business and research while market and society preferences determine which technologies prevail. Fundamental rights must be respected throughout the legal system aiming for full realization. Switzerland should pursue principle-based and technology-neutral legislative approaches but allow exceptions if necessary.

The report establishes three overarching regulatory objectives. First, strengthening Switzerland as an innovation location by ensuring business and research have access to relevant markets and research areas with barrier-free imports. Second, safeguarding protection of fundamental rights including economic freedom by ensuring protection keeps pace with technological developments. Third, strengthening public trust in AI through appropriate transparency, traceability and explainability of processes and decisions.

According to survey results from 41 responding federal agencies, 80% believe an overarching approach does not seem out of place. The majority stated that purely sectoral regulatory approaches were insufficient. Cross-sectoral approaches could potentially resolve broadly based challenges using common logic while sector-specific issues would still require individual solutions.

The Federal Data Protection and Information Commissioner stated on November 9, 2023, that current data protection legislation is directly applicable to AI. The principles of the Data Protection Act apply when AI is used. However, according to survey feedback, challenges are amplified when AI is deployed. The increasing ability of AI to link different datasets and compare information types makes distinguishing between personal and factual data more difficult.

Market access considerations prove particularly relevant for Swiss companies. The EU represents Switzerland's main sales market for AI products and services given existing economic relations and geographical proximity. With an estimated 2024 market size of USD 31.6 billion, the EU already ranks among the world's largest AI sales markets with expected annual growth rate of 28.43% leading to market volume of USD 141.8 billion by 2030.

The agreement between Switzerland and the EU on mutual recognition in relation to conformity assessment (MRA) carries significant importance in connection with AI Act impact on Swiss economy. The agreement covers approximately two-thirds of trade in industrial products between Switzerland and the EU by value. More than half—12 out of 20—product sectors listed in the MRA are affected by AI Act provisions if products contain AI components.

These 12 affected product categories include machinery, lifts, pressure vessels, gas appliances, personal protective equipment, toys, medical devices, telecommunications terminal equipment, equipment for potentially explosive atmospheres, motor vehicles, agricultural or forestry tractors and cableway installations. The AI Act classifies products in these sectors as high-risk AI systems if they undergo conformity assessment procedures by third-party conformity assessment bodies according to existing EU harmonisation legislation.

From August 2027, AI Act requirements for high-risk AI systems will apply to products in these 12 MRA sectors containing high-risk AI components. These requirements are not currently covered by the MRA. If products contain AI components, additional conformity assessment must be carried out by conformity assessment bodies in the EU according to AI Act requirements beyond existing conformity assessments currently applicable for Swiss and EU internal markets.

Double conformity assessment will lead to additional work and costs for affected companies. Access to EU internal markets will entail additional difficulties for Swiss products with AI components starting August 2027. To avoid new technical barriers to trade in these areas, Switzerland would need to adapt product regulations in these sectors to those in the AI Act. Furthermore, the MRA would require expansion to include product-related AI legislation.

The AI Act's implementation schedule creates graduated compliance requirements. According to Commission guidance, enforcement becomes applicable one year later for new models and two years later for existing models placed on market before August 2025. The European Commission opened consultation for AI transparency guidelines in September 2025, with transparency obligations under Article 50 becoming applicable from August 2, 2026.

International technical standards development occurs at multiple levels. The International Electrotechnical Commission (IEC), International Organization for Standardization (ISO) and International Telecommunication Union (ITU) are developing various standards addressing society's concerns about AI through the World Standards Cooperation. Standards aim to provide suitable guidelines for responsible, secure and trustworthy AI development.

The European Commission adopted an implementing decision on May 22, 2023, on a standardisation request directed to the European Committee for Standardisation and European Committee for Electrotechnical Standardisation in support of Union policy on artificial intelligence. This request would prepare necessary technical environment for AI Act implementation. Work started on European standards and European standardisation documentation addressing specifications for design and development of high-risk AI systems.

Standards development encountered delays affecting high-risk AI systems. The initial European standardization request deadline of April 1, 2025, expired without completion, creating uncertainty about timeline expectations. The Netherlands published its fifth AI and Algorithms Report on July 15, 2025, outlining that regulatory sandbox initiative will become operational by August 2026.

Switzerland's regulatory approach would determine institutional structures and financial implications. Continuation of sector-specific activities would require no new institutions, only expanded scope of existing ones. AI Convention ratification would necessitate coordinated approach at federal level ensuring coherent handling of cross-sectoral challenges such as transparency or risk assessment of AI systems.

For minimum AI Convention implementation, existing federal agency could be tasked with coordination function. Coordination could occur within competence network framework for regulatory AI issues where several existing bodies join forces. Under this approach, no new institutions would be created, only additional tasks assigned.

More extensive AI Convention implementation could examine creation of new coordination office ensuring coherent approach to cross-sectoral issues. Supervisory bodies required for monitoring compliance could receive extensive powers including decision-making authority or sanction imposition capabilities.

Comprehensive product regulation aligned with the AI Act would require identification of institutions needed for creation or adaptation. These institutions would perform tasks including supervision of legal requirements. Analysis would determine whether supervisory bodies specified by AI Convention should expand to include product-centric tasks or whether separate performance makes more sense.

The report acknowledges that each regulatory approach carries cost implications. Because regulatory approaches can be presented only in generic terms at this stage, serious assessment of specific financial impact remains impossible. However, well-crafted regulation combined with AI can have positive financial impact by increasing Switzerland's competitiveness.

The EU's regulatory impact assessment of the AI Act provides frame of reference. Total compliance cost is estimated between EUR 100 and 500 million by 2025, representing up to 4-5% of investment in high-risk AI systems. Compliance assessment costs could account for additional 2-5% of investment in high-risk AI systems. Companies or public authorities developing or using AI applications not considered high-risk would not incur such costs.

Federal agencies whose sectors involve particularly relevant AI applications are addressing regulation from regulatory perspective or plan examination in near future. The Swiss Federal Office of Energy is conducting screening of AI activities including new standards and international best practices in AI promotion and regulation. Potential regulatory measures and recommendations for energy sector are being identified on that basis.

The Federal Office of Police is procuring new automated fingerprint identification system. Given widespread AI use in such systems, framework agreement with provider contains explicit provisions on responsible AI use. Accreditation under ISO 17025 of the service using the system operationally sets guidelines regarding traceability and validation.

As part of overview of legal framework for AI applications in financial sector, the State Secretariat for International Finance is reviewing aspects intended to result in secure legal framework for AI use. FINMA already addresses risks associated with AI applications as part of supervisory practice.

The Federal Statistical Office addresses AI applications and their handling as part of complete revision of statistical ordinances into single Federal Statistical Ordinance. AI applications offer great potential for public statistics as public statistics per se involve only minor intrusions into privacy of people concerned.

Monitoring of federal AI guidelines adopted in 2020 shows mixed results. The 2024 survey demonstrated that AI guidelines are well known in Federal Administration and are used in specific cases as general frame of reference. However, just over one-third of respondents see need for adjustment. Survey results show need for practical AI instructions and directives in Federal Administration.

The sectoral survey showed AI poses new challenges for all sectors in principle. Focus is particularly on issues of transparency and traceability as well as protection of privacy and data protection. At issue is recognizable, traceable and comprehensible use of AI applications and respect for privacy and personality protection when training and using AI systems.

Issues relating to discrimination and fairness as well as cybersecurity are also relevant for sectors. Use of AI applications should not treat people differently on basis of protected characteristics, and AI applications should be robust against attacks and errors. However, all issues are weighted differently depending on sector, as each focuses on different aspects within its field of activity involving different risks.

Germany expressed concerns over AI Act implementation in October 2025, with the Bundesverband Digitale Wirtschaft highlighting challenges in national implementation approach. The organization identifies significant challenges in draft legislation that could affect Germany's competitiveness as AI development hub.

Switzerland's approach emphasizes stakeholder engagement. Over course of 2024, OFCOM provided regular information on overview via Plateforme Tripartite open to all interested parties, offering opportunity for exchange and input at several meetings. Interested stakeholders were able to present positions on AI regulation.

In May 2024, exchange on AI regulation took place in cooperation with State Secretariat for Economic Affairs within extraparliamentary Economic Policy Commission. Potential regulation of AI was dominant topic at Swiss Internet Governance Forum on June 5, 2024. On September 11, 2024, Federal Councillor Albert Rösti invited experts from business sector, academia and civil society to advisory committee meeting as part of Digital Switzerland Strategy.

The report emphasizes that existing legal framework already addresses many core challenges in Switzerland today. Apart from protection of fundamental rights and empowerment of market participants, public trust in security of useful and efficient applications is crucial. Without public trust, no basis can be created for social dialogue and further development of new emerging technologies.

Switzerland has been considered very strong innovation location internationally for years. In the Global Innovation Index 2024, Switzerland ranked first out of 132 economies worldwide, as in previous year. The Global Innovation Index is ranking presenting innovative capacity of individual countries published annually by French business school INSEAD, Cornell University and World Intellectual Property Organization.

The baseline analyses consistently show need for action to regulate AI in Switzerland. The various regulatory approaches build on each other. Each approach goes further than previous one in terms of implementation design. Depending on Federal Council's decision, regulatory approaches will need examination in greater depth in follow-up work.

A fundamental decision by Federal Council as to which regulatory approach it intends to pursue is now needed. Depending on regulatory approach chosen, in-depth clarifications will be required. These clarifications concern precise design of particular approach and its institutional and financial impacts.

Timeline

• August 2023: Personal Information Protection Commission of South Korea announces policy direction for safe usage of personal data in AI age
• November 22, 2023: Federal Council mandates DETEC and Federal Office of Communications to prepare overview of regulatory approaches for AI by end of 2024
• May 17, 2024: Committee of Ministers of Council of Europe adopts AI Convention
• June 5, 2024: Swiss Internet Governance Forum focuses on potential AI regulation as dominant topic
• August 1, 2024: EU AI Act enters into force, establishing comprehensive regulatory framework
• September 5, 2024: Council of Europe's AI Convention opens for signature with EU, Andorra, Georgia, Iceland, Israel, Moldova, Norway, San Marino, UK and US signing
• September 11, 2024: Federal Councillor Albert Rösti invites experts to advisory committee meeting on AI regulation
• July 2025: Commission releases AI Act guidelines and Meta announces refusal to sign code of practice
• August 2025: US Attorneys General target AI companies for child safety failures in bipartisan letter
• October 2025: German digital association expresses concerns over AI Act implementation approach
• November 2025: European Commission proposes major GDPR changes through Digital Omnibus initiative affecting AI development
• February 12, 2025: Switzerland publishes comprehensive overview of artificial intelligence regulation report

Summary

Who: Switzerland's Federal Department of the Environment, Transport, Energy and Communications DETEC and Federal Office of Communications OFCOM prepared the report following Federal Council mandate from November 22, 2023. The Interdepartmental Coordination Group on EU Digital Policy conducted the work involving approximately 25 federal agencies under leadership of OFCOM and Federal Department of Foreign Affairs Europe Division. A core team comprised Federal Office of Justice, Directorate of International Law, European Division of Federal Department of Foreign Affairs and OFCOM.

What: Switzerland published a 37-page overview analyzing three possible regulatory approaches for artificial intelligence on February 12, 2025. The first approach continues sector-specific regulatory activities without overarching coordination. The second approach involves ratifying Council of Europe's AI Convention through minimum or extensive implementation requiring new standards for transparency, risk and impact assessment, and oversight mechanisms. The third approach combines AI Convention ratification with comprehensive product regulation aligned with EU AI Act, introducing risk-based classification systems and obligations for AI system providers and deployers.

When: The Federal Council mandated the overview on November 22, 2023, with publication occurring February 12, 2025. The analysis examined developments through autumn 2024, with baseline legal analysis covering through August 31, 2024. The Council of Europe adopted the AI Convention on May 17, 2024, opening for signature September 5, 2024. The EU AI Act entered force August 1, 2024, with most obligations beginning August 2, 2025, and transparency requirements applicable from August 2, 2026.

Where: The regulatory approaches affect Switzerland's entire federal system including all sectors where AI applications are used, from financial services and energy to transportation and public statistics. The analysis examined 20 countries across six continents. Twelve of 20 product sectors in Switzerland's Mutual Recognition Agreement with EU are affected by AI Act provisions if products contain AI components, creating market access implications for Swiss companies exporting to EU internal market.

Why: Switzerland currently has no overarching AI-specific legislation despite AI applications permeating multiple economic sectors. The Federal Council defined AI as priority topic in 2023-2027 legislative plan calling for fundamental decision on future Swiss regulatory approach in first half of legislative period. The report identifies three overarching regulatory objectives: strengthening Switzerland as innovation location, safeguarding protection of fundamental rights including economic freedom, and strengthening public trust in AI through appropriate transparency, traceability and explainability of processes and decisions using AI systems.