Skip to content
Article

Interpretative organisation for AI regulation: Federal Council wants to ratify the Council of Europe’s AI Convention

Data protection law

On 12 February 2025, the Federal Council published an overview of possible regulatory approaches to artificial intelligence (AI), which it had commissioned DETEC and the FDFA to draw up in November 2023. On the basis of this overview, the Federal Council decided in favour of a regulatory approach for AI: ratification of the Council of Europe’s AI Convention with sectoral amendments to existing laws.

The Federal Council’s regulatory approach is based on three objectives:

– Strengthening Switzerland as a centre of innovation,
– safeguarding the protection of fundamental rights, including economic freedom, and
– Strengthening the public’s trust in AI.

On this basis, the FDJP, together with DETEC, is to draw up a consultation draft and an implementation plan by the end of 2026 (!). In this article, we reflect on the impact of sectoral regulation.

I. Interpretation

The analysis is based on three basic analyses: the legal analysis, the sectoral analysis and the country analysis. The results of the basic analyses show that there is a need for action with regard to the regulation of AI in Switzerland.

  1. The basic legal analysis

The basic legal analysis analyses the effects and requirements of the AI Convention of the Council of Europe, the AI Act of the European Union (EU AI Act) and the current status of selected areas of Swiss law.

  1. The sectoral analysis

The sectoral analysis provides an overview of the existing and planned changes to federal law in various sectors.

  • Specific need for action: There is a specific need for adaptation in certain areas of law. Overall, however, the existing standards already provide answers to many questions relating to the regulation of AI.
  • Adjustments already made in technology-driven sectors: Some sectors have already made initial legal adjustments, particularly in connection with the new data protection provisions, which also affect the use of AI applications.
  1. The country analysis

The country analysis presents the regulatory developments in 20 selected countries. The following findings were made:

  • Different international regulatory approaches: Countries around the world are taking different approaches to regulating AI. Many are in a testing phase and, apart from the EU member states, only a few countries have adopted and implemented specific AI regulations.
  • Common awareness of the need for action: There is a global understanding of the need for measures to regulate AI, with the risk-based approach of the EU AI Act often serving as a model.

In summary, the analyses show that Switzerland needs to become active in AI regulation, both from a legal and sector-specific perspective. At the same time, it is important to take international developments into account in order to maintain the competitiveness of the Swiss economy in a global context and to ensure the protection of the fundamental rights of people affected by the use of AI. Based on this, 3 possible regulatory approaches for Switzerland were developed.

II Regulatory approaches for AI in Switzerland

To date, Switzerland has pursued a technology-open and innovation-friendly approach to AI regulation. Nevertheless, an analysis of the current legal situation shows that there is a need for action in certain areas, particularly with regard to transparency, liability and the protection of fundamental rights. There are essentially three conceivable regulatory approaches:

  1. Continuation of topic- and sector-specific regulations

This approach would provide for adjustments in relevant areas of law such as data protection or product safety and financial market regulation. There would be no overarching, specific AI regulation, but rather adjustments within existing laws. This strategy would be the least invasive and would correspond most closely to Switzerland’s existing regulatory principles. Nevertheless, inconsistencies between sectors could arise and new risks could be insufficiently covered.

  1. Ratification of the Council of Europe’s AI Convention

In addition to data protection, the Council of Europe’s international AI Convention also addresses ethical and human rights aspects. Switzerland could ratify this convention and implement it nationally, whereby various degrees of implementation are conceivable. Within the minimum implementation, only basic principles of the convention would be adopted, with more extensive implementation leading to the addition of further national regulations. The advantage of this would be international compatibility and a high degree of flexibility. Nevertheless, the less specific rules, such as the EU AI Act, would have to be kept in mind.

  1. Adaptation to the EU AI Act

Switzerland could also adapt its legislation to the requirements of the EU AI Act in order to facilitate access to the EU internal market. The advantages of this would be legal certainty for companies, the preservation of EU market access and the avoidance of trade barriers. However, this would limit Switzerland’s regulatory autonomy and the possibility of high adjustment costs for Swiss companies cannot be ruled out.

A central point of the regulatory considerations in favour of this approach is economic compatibility with international trade standards. This relates in particular to the agreement on the mutual recognition of conformity assessments (MRA) between Switzerland and the EU. As 12 out of 20 product sectors in the MRA are affected by the requirements of the EU AI Act, Switzerland would have to adapt its product regulations in order to avoid trade barriers with the EU. This would require an update or expansion of the MRA to include the AI sector. However, such an adjustment would not be possible until 2028 at the earliest as part of the stabilisation of the bilateral path between Switzerland and the EU.

III Appraisal and outlook

The Federal Council’s decision to actively tackle the issue of AI regulation is to be expressly welcomed. However, Switzerland’s decision to ratify the Council of Europe’s AI Convention instead of its own law on the regulation of AI and thus to regulate artificial intelligence through sector-specific adaptations could also have detrimental effects in the long term:

  • Fragmentation of regulation: a sector-specific approach could lead to inconsistent regulations in different industries that are difficult for companies and users alike to understand. This also increases the complexity for companies operating in several sectors. Furthermore, this could impair the efficiency and coherence of regulation. Furthermore, this inevitably leads to foreign technology providers being the last to take Switzerland into account, as adapting to Swiss-specific regulation is costly and time-consuming.
  • Risk of regulatory gaps: As AI technologies are often used across sectors, there is a risk that certain applications cannot be clearly assigned to a sector and therefore remain unregulated, creating a regulatory grey area.
  • Coordination effort: The development and implementation of sectoral regulations require close coordination between different authorities and sectors, which can lead to increased administrative effort for those involved. Potential conflicts of interest cannot be ruled out. In terms of company compliance, this also means that different AI rules may apply to one product in the absence of generally applicable rules.
  • Competitive disadvantages: Compared to countries with comprehensive, horizontal AI regulations, Switzerland could be perceived as less stringent, which could lead to tensions in international trade relations and delay the product launch for Switzerland.
  • Delayed responsiveness: The rapid development of AI requires flexible and rapid adjustments to the legal framework. Experience has shown that a sector-specific approach is sometimes associated with considerable time expenditure and could impair the responsiveness of regulation, as adjustments would be necessary in several sectors at the same time.
  • Loss of public confidence: Inconsistent or incomplete regulation could undermine public confidence in the safe use of AI, especially if incidents occur that are attributable to regulatory gaps and offer those affected no or insufficient options for protecting their rights.

The Federal Council’s decision shows that Switzerland is pursuing a pragmatic approach to AI regulation. The combination of binding and non-binding measures is intended to create a balance between legal certainty and technological openness. While Switzerland refrains from comprehensive cross-sector regulation, the ratification of the Council of Europe’s AI Convention and targeted adjustments in important sectors will ensure international compatibility. It is now up to the institutions involved to shape the regulatory framework in such a way that both Switzerland’s innovative strength is maintained and social and legal risks are minimised.

Sources