AI regulation in Switzerland (2026)

FINMA published results of a survey of ~400 licensed institutions showing around half already deploy AI, with 91% of those using generative AI. The survey found governance structures lagging adoption, with many institutions under-managing model risk, bias, and explainability gaps.

Switzerland formally signed CETS No. 225 — the world's first legally binding international treaty on AI — in Strasbourg, as announced by the Federal Council on 12 February 2025. Ratification requires parliamentary approval of sector-specific legislative amendments expected in a consultation draft by end-2026.

After reviewing three governance models, the Federal Council chose not to enact a comprehensive cross-sectoral AI law. Instead, Switzerland will ratify the CoE AI Convention, make targeted sector-specific legislative amendments, and rely on non-binding measures (self-disclosure, industry codes) — explicitly rejecting unilateral adoption of the EU AI Act.

The Swiss Financial Market Supervisory Authority published its first dedicated AI supervisory guidance, clarifying that existing technology-neutral rules on governance, risk management, and outsourcing fully apply to AI use in banking and insurance. It identified model risk, third-party dependency on Big Tech, and lack of explainability as priority supervisory concerns.

The first binding international AI treaty — covering AI systems' alignment with human rights, democracy, and rule of law — opened for signature. As a Council of Europe member, Switzerland's subsequent decisions on AI governance were directly shaped by this instrument.

The Federal Council instructed DETEC (OFCOM) and the Federal Department of Foreign Affairs to prepare an overview of possible AI regulatory approaches for Switzerland by end-2024. This mandate directly triggered the February 2025 policy decision on Switzerland's governance path.

The Federal Data Protection and Information Commissioner (FDPIC/EDÖB) issued a formal update affirming that the newly in-force revFADP — drafted in technology-neutral terms — covers AI-based personal data processing. Providers and users of AI must ensure transparency, purpose limitation, and data-subject rights.

Switzerland's modernised data-protection law — aligned with GDPR and Convention 108+ — became enforceable, giving the FDPIC expanded investigative and enforcement powers and imposing criminal fines up to CHF 250,000. Because the Act is technology-neutral, it immediately became the primary legal instrument governing AI data-processing in Switzerland.

The Federal Council published binding orientation guidelines for federal agencies and bodies using AI, establishing principles of human-centredness, transparency/explainability, fairness, safety, accountability, data protection, and sustainability. These remain the internal governance baseline for AI use within the Confederation.

The Federal Council's IDAG AI (under SERI/SBFI) released a 37-priority-action-area report concluding that Switzerland's existing legal framework was broadly adequate for AI but required targeted clarification on transparency, accountability, bias, and sector-specific gaps. This report set the baseline for all subsequent Swiss AI policy deliberation.

The Federal Council's Digital Switzerland Strategy created an interdepartmental working group on artificial intelligence (IDAG AI) under SERI to coordinate federal AI policy, identify regulatory gaps, and ensure Switzerland's global competitiveness. This body authored the 2019 Challenges report and the 2020 AI Guidelines, anchoring Switzerland's foundational AI governance architecture.