Switzerland to ratify Europe's landmark AI Framework Convention – What does it mean?

Currently, Switzerland does not have overarching AI-specific legislation. Instead, regulatory adjustments are made as needed in individual sectors where AI is likely to have a considerable impact and legislative changes could provide more clarity and safety. Given AI's rapid development and use, calls for the clarification of the legal situation of AI in more detail have been made. In November 2023, the Federal Council instructed the Federal Department of Environment, Transport, Energy and Communications (DETEC) and the Federal Department of Foreign Affairs (FDFA) to analyse possible regulatory approaches to AI in Switzerland, as AI is a priority in the Federal Council's 2023-2027 legislative plan.

On 12 February 2025, the DETEC presented an overview of possible AI regulatory approaches to the Federal Council. This overview considered existing Swiss legislation, sector-specific regulations, international regulatory approaches to AI, and relevant economic and European policy implications. One suggested approach was for Switzerland to ratify the Council of Europe Framework Convention on Artificial Intelligence (AI Framework Convention) and adapt Swiss law accordingly.

Also, on 12 February 2025 the Federal Council confirmed this approach and announced Switzerland's intention to ratify the AI Framework Convention. The Federal Council instructed the Federal Office of Justice (FOJ) to work with the DETEC to draft an implementation bill of the AI Framework Convention and draw plans for further non-binding measures. Results are expected by the end of 2026.

Further regulatory work on AI in specific sectors such as healthcare and transportation are also planned.

Switzerland's approach to AI regulation

Pursuant to the AI-Report, AI regulation should aim to achieve three key objectives:

• enhance Switzerland's position as an innovation hub;
• ensure the protection of fundamental rights, including economic freedom; and
• promote public trust in AI.

To this end, three possible regulatory approaches were considered and presented to the Federal Council:

• Sector-specific regulation: Swiss AI regulation would continue to be developed on a sector-specific basis, addressing key areas such as data protection where necessary. No overarching AI regulation would be planned, and cross-sectoral challenges would be addressed through existing consultation or self-regulatory processes.
• Ratification of the AI Framework Convention: Switzerland could ratify the AI Framework Convention with either a minimal or more comprehensive implementation. A minimal approach would impose stronger obligations on the state than on private actors, while a more comprehensive implementation would extend similar obligations to both sectors, particularly on key issues such as transparency and risk assessments.
• Alignment with Regulation (EU) 2024/1689 laying down harmonized rules on artificial intelligence (EU AI Act): Switzerland could adopt a risk-based approach in line with the EU AI Act, which could facilitate access to the EU market, but would also result in a higher regulatory burden.

These three regulatory approaches build on and complement each other. Additional measures could also be taken, notably in relation to the use of regulatory sandboxes (i.e. environments where companies can develop and test AI under the supervision of the regulator) and an agreement between Switzerland and the EU on mutual recognition for conformity assessments under the EU AI Act and/or Swiss law and vice versa.

The Federal Council opted for the second approach and decided that the AI Framework Convention will be incorporated into Swiss law. This will primarily apply to state actors at the federal level. Cantons would be responsible for implementing the AI Framework Convention within their cantonal jurisdictions. As said earlier, legislative changes should occur in a sector-specific manner, with cross-sectoral regulation limited to key areas relevant to fundamental rights, such as data protection, employment law or Swiss liability laws (e.g. Swiss code of obligations or Swiss Federal Act on Product Liability). For instance, the AI-Report highlights that due to transparency and non-discrimination challenges in employment law, certain adjustments might be necessary over time. Also, while the Swiss liability regime is considered sufficiently flexible, select adjustments might be necessary to modernise the laws for AI-related liability cases. In particular, the Swiss Federal Act on Product Liability and the Swiss Federal Act on Product Security should ideally be updated and aligned with more recent EU legislation, such as Directive 85/374/EEC on liability for defective products of 2024 and EU Regulation 2023/988 on product security. In addition, the implementation of the AI Framework Convention will also be supported by non-binding measures, such as self-disclosure agreements or industry solutions. Combining binding and non-binding measures will create a flexible and robust legal framework that can adapt to the rapid development of AI.

Key provisions of the AI Framework Convention

The AI Framework Convention – which Switzerland envisages ratifying – is a framework convention that aims to ensure compliance with international standards on human rights, democracy, and the rule of law throughout the life cycle of AI systems. While it does not impose specific prohibitions, it requires states to consider the need for moratoria and applies directly to the AI-related activities of public authorities. States are required to assess and address risks and impact in a manner consistent with the AI Framework Convention vis-à-vis the activities of private actors, particularly when these actors engage in activities that would affect fundamental rights.

While Swiss law already provides a sufficient level of protection regarding the integrity of democratic processes, respect for the rule of law and the requirement that there be public consultation on important issues related to AI, adjustments are needed in following areas:

• Transparency and oversight: typically for the identification of content generated by AI systems;
• Safe innovation: in particular, the establishment of controlled environments for the development, experimentation and testing of AI systems under the supervision of competent authorities;
• Ensuring the availability of accessible and effective remedies: in case of human rights violations resulting from the activities of AI systems;
• Ensuring the availability of effective procedural guarantees: safeguards and rights when an AI system has a significant impact on human rights, including informing persons interacting with AI systems that they are interacting with an AI bot and not a human being;
• Risk and impact management: taking measures to identify, assess, prevent and mitigate risks posed by AI systems on human rights; and
• Effective oversight mechanisms: to oversee compliance with the obligations under the AI Framework Convention.

These standards should be incorporated into Swiss legislation. An important part of this implementation is the principle of proportionality. Hence, an assessment of the severity and likelihood of harm resulting from specific AI applications must be considered.

AI Framework Convention vs. EU AI Act: what's the difference?

The AI Framework Convention sets out the above core standards to be incorporated into Swiss law, while allowing the Swiss legislator flexibility in their implementation. As mentioned above, these principles are primarily aimed at public authorities.

In contrast, the EU AI Act, which entered into force on 1 August 2024, is primarily applicable to EU-based companies and public authorities. It contains detailed and prescriptive provisions built on a risk-based classification of AI systems. Even if the EU AI Act is not directly applicable in Switzerland, it may have an impact on companies based in Switzerland as it applies to providers and deployers located in third countries. This means that Swiss companies will fall within the scope of the EU AI act if they do the following:

• make AI systems or general-purpose AI models available in the EU;
• supply AI systems for use in the EU; or
• have AI systems that generate output used in the EU.

In addition, Swiss-based companies that use AI systems as deployers to generate content used in the EU also fall within the scope of the EU AI Act.

In a nutshell, the AI Framework Convention aims to ensure that fundamental rights and the rule of law are safeguarded by the public authorities of signatory states. It does not contain a detailed and risk-based set of provisions to regulate AI. The EU AI Act, however, targets both public authorities and private entities operating in the EU and directly imposes requirements that must be met by such actors regarding AI. If the EU AI Act is violated, penalties and enforcement measures could apply. Furthermore, the EU AI Act contains detailed provisions and prescriptions, which are directly applicable in EU member states and can have an overarching and extraterritorial effect on third countries, such as Switzerland.

Impact of the AI Framework Convention on Swiss private actors

Switzerland's envisaged ratification of the AI Framework Convention does not appear to directly affect the Swiss private sector, except in limited cases where fundamental rights, such as the right to privacy, could have a "horizontal effect", requiring private parties to respect these rights within their own contractual relationships. In such cases, concerned Swiss private entities may need to abide by the core principles set out in the AI Framework Convention. It is also expected that governmental entities contracting with Swiss vendors in the AI sector will impose obligations of the AI Framework Convention on these vendors in their contracts. Companies frequently contracting with governmental entities in Switzerland should consider this in their procurement strategy and management.

Sector-specific legislative changes, however, are expected along with cross-sectoral regulation for fundamental rights issues such as data protection. This would potentially affect laws such as the Federal Copyright Act (for generative AI systems), the Federal Product Liability Act and the Federal Product Security Act (which will need to be modernised due to the technical development of products in the context of AI), employment and labour laws (regarding non-discrimination, transparency and data protection, such when using tools that influence the hiring process), and sectors impacted by AI such as healthcare and transport, as announced by the Federal Council.

Finally, financial regulations in Switzerland may also be impacted by these legislative adjustments. Currently, the relevant framework is embodied in the guidance of the Swiss Financial Market Supervisory Authority (FINMA) on governance and risk management when using artificial intelligence, which was released in December 2024. This guidance and other FINMA Circulars could be subject to changes over time due to Switzerland's envisaged ratification of the AI Framework Convention.

Outlook

With its decision to ratify the AI Framework Convention, the Federal Council has confirmed Switzerland's middle-ground approach to AI of:

• making select adjustments where necessary to safeguard fundamental rights and meet specific sectoral needs; and
• avoiding an overarching, heavy regulatory burden that a full alignment with EU AI Act would entail, which will strengthen Switzerland's position as an attractive hub for innovation and business.

The FOJ and DETEC are expected to have an implementation bill of the AI Framework Convention drafted and plans for further non-binding measures drawn up by late 2026.

For more information on AI regulations in Switzerland, contact your CMS client partner or local CMS experts.