Switzerland - AI Regulatory Approaches

Overview

The Federal Office of Communications (OFCOM) prepared an "Overview of artificial intelligence regulation" as a report to the Federal Council, presented on 12 February 2025. The document responds to a Federal Council mandate from 22 November 2023 to map possible regulatory approaches for AI in Switzerland and to propose follow-up steps. The overview identifies three nested approaches: (1) continuation of topic- and sector-specific regulation; (2) ratification of the Council of Europe AI Convention with minimum or more extensive implementation; and (3) ratification combined with implementation aligned with the EU AI Act's risk-based model. The full report and related baseline analyses are available from OFCOM: OFCOM – Artificial Intelligence and the report PDF: Overview of artificial intelligence regulation (OFCOM, 12 Feb 2025).

Definitions

The overview adopts a functional definition of AI consistent with international practice: systems that use data-driven or knowledge-based models to perform tasks that ordinarily require human intelligence, including machine learning, deep learning and certain rule-based systems. It distinguishes general-purpose (GPAI) models from application-specific systems and recognises context-dependent risk: the same model may present different risks depending on use. The report cross-references terminology used in the Council of Europe AI Convention and the EU AI Act to ensure conceptual interoperability.

Governance and Institutional Framework

The report analyses institutional options: (i) keep existing sectoral regulators responsible for AI within their domains; (ii) strengthen a coordinating federal function (DETEC/OFCOM acting as coordinator) and mandate lead authorities for cross-cutting duties; or (iii) create a specialized oversight authority or AI safety institute. OFCOM led the preparatory work and will continue coordination with DETEC, the FDFA (foreign policy implications) and the FDJP (legal implementation). The report also examines the role of independent regulators, data protection authorities and conformity-assessment bodies. Links to the coordinating bodies and departments: DETEC (Federal Department), FDJP, FDFA.

Key Focus Areas

The overview identifies core thematic areas that merit federal attention regardless of the chosen approach: transparency and explainability (disclosure of AI role and provenance of outputs); risk and impact assessment (pre-deployment assessments for systems affecting rights and safety); data protection and data quality; non-discrimination and fairness; robustness, cybersecurity and model security; conformity assessment and market surveillance for AI components embedded in products; post-market monitoring and incident reporting; documentation and record-keeping for auditability; and measures to preserve innovation (regulatory sandboxes, innovation hubs, standards uptake). It draws particular attention to categories that carry elevated risk — e.g. biometric identification, law-enforcement profiling, critical infrastructure control, medical devices and transport safety systems — and to general-purpose AI (GPAI) which may span multiple risk domains. The report maps these focus areas to obligations under the Council of Europe AI Convention and the EU AI Act and highlights the trade-offs between legal certainty and preserving innovation space.

Implementation Framework

Depending on the Federal Council’s chosen approach, different implementation instruments are proposed: legislative amendments (targeted, sector-specific laws where necessary), transposition instruments for the Council of Europe AI Convention, administrative guidance and non-binding instruments (codes of conduct, self-declaration, sectoral agreements), and operational tools (regulatory sandboxes, conformity-assessment procedures and certification schemes). The report emphasises integration with existing Swiss regulatory instruments (e.g. data protection law) and the practical implications for market access to the EU via the MRA, including potential additional conformity assessments for AI components. The OFCOM overview suggests sequencing: ratification of the Convention, then targeted legal and non-legal measures while monitoring international developments.

Monitoring and Evaluation

OFCOM recommends an active monitoring and evaluation framework combining periodic reviews of implemented measures, post-market surveillance obligations for deployers, standard-setting participation, and stakeholder dialogue through the Plateforme Tripartite and CNAI. The report notes OFCOM’s prior monitoring of the Federal AI guidelines and proposes regular assessments to measure whether objectives (innovation, rights protection and trust) are met. Where alignment with the EU AI Act is pursued, the report highlights the need for data collection on compliance costs and market impacts, including the consequences for Swiss conformity-assessment bodies and exporters.

Penalties, Liability, and Appeals

The overview does not set final penalty levels but outlines liability and enforcement models observed internationally (administrative fines, corrective orders, product market removals, criminal liability in extreme cases). It highlights that specific sanctions and redress mechanisms will be decided in implementing legislation; where alignment with the EU AI Act is chosen, Switzerland would likely adopt similar enforcement tools and sanctions calibrated to the risk category. The report also stresses accessible remedies for affected persons (complaints procedures, judicial review) and the need for clear allocation of responsibilities among providers, deployers and regulators.

Relationship to Other Instruments

The report maps interactions with Swiss and international instruments: domestic data-protection law, consumer and product-safety regulations, sectoral health and transport rules, and international treaties. It analyses how ratifying the Council of Europe AI Convention and optional alignment with the EU AI Act would affect the Switzerland–EU MRA and conformity-assessment flows. The document also discusses technical standards and their role in demonstrating conformity, and the potential for mutual recognition arrangements to reduce duplicate assessments for exporters.

International Alignment

International alignment is a central theme: the overview stresses the benefits of convergence with the Council of Europe AI Convention (ratification decided by the Federal Council) and contemplates calibrated alignment with the EU AI Act to ease market access. It recommends active Swiss participation in international fora (Council of Europe, OECD, UNESCO, EU dialogues) and coordination to avoid a ‘‘Swiss finish’’ that would diverge materially from EU norms. The report also highlights the role of international standards and cross-border cooperation in supervision and conformity assessment.

Implementation Timeline

Event	Date / Target
Federal Council mandate to DETEC/FDFA	2023-11-22
OFCOM overview presented to Federal Council	2025-02-12
Publication on OFCOM website	2025-03-10 (OFCOM page published)
Consultation draft to be prepared (implementing AI Convention)	By 2026-12-31
Implementation plan for non-legislative measures	By 2026-12-31

Compliance Checklist

Checklist Item	Required action
Determine regulatory approach	Federal Council decision (Convention ratification / alignment level)
Legal gap analysis	Complete sectoral amendments where necessary (data protection, discrimination law)
Risk classification	Adopt risk-based categorisation if aligning with EU AI Act
Documentation & record-keeping	Establish obligations for providers/deployers for technical documentation
Conformity assessment	Plan for conformity/certification and MRA implications
Non-binding measures	Design self-declaration/codes and sandboxes

Sources and References

Source	Type
OFCOM – Artificial Intelligence (overview page)	Primary Source
Overview of artificial intelligence regulation (OFCOM report, 12 Feb 2025)	Primary Source
News service: AI regulation: Federal Council to ratify Council of Europe Convention (12 Feb 2025)	Primary Source