Artificial Intelligence in Switzerland: Regulation and Administrative Use

Author: Federal Chancellery (Digital Transformation and ICT Governance Division)
Source: https://www.bk.admin.ch/bk/de/home/digitale-transformation-ikt-lenkung/bundesarchitektur/kuenstliche_intelligenz.html
Publication Date: February 12, 2025 (Federal Council Resolution)
Access Date: December 2025
Reading Time: approx. 8–10 minutes

Executive Summary

Switzerland is developing its first coherent AI regulatory framework, implementing the Council of Europe convention and to be finalized by end of 2026. In parallel, the federal administration is modernizing its AI governance through the strengthened competency network CNAI and clear deployment guidelines to leverage opportunities and minimize risks (data protection, discrimination, transparency). The focus is on managed innovation rather than prohibition policy – with self-commitments as a supplement to binding rules.

Critical Guiding Questions (AI Regulation & Administration)

1. Freedom & Innovation:
   Will the regulatory approach (transparency, data protection, non-discrimination) foster or hinder Swiss AI development through EU-like compliance hurdles?

2. Accountability:
   Who bears legal liability for AI-assisted administrative decisions – authorities, operators, or developers? Current guidelines mention "human responsibility," but legally binding mechanisms are still missing.

3. Evidence & Impact:
   What empirical foundations underpin the planned regulatory standards? How is over-regulation avoided?

4. Economic Consequences:
   Do established tech corporations benefit (higher compliance costs as market entry barriers) or do flexible self-commitments promote SME innovation?

5. International Competitiveness:
   How does Switzerland synchronize its AI regulation with the EU AI Act to avoid fragmentation and competitive disadvantages?

Scenario Analysis – AI Policy Perspectives

Core Topic & Context

Switzerland is currently acting on two parallel levels: (1) national AI regulation and (2) internal administrative reforms. While there has been no comprehensive AI legislation to date, the Federal Council is now pursuing a step-by-step, multi-level approach: binding minimum standards (regulation) combined with voluntary best-practice guidelines (governance and self-commitment). This reflects a liberal-pragmatic stance: enable innovation, but mitigate critical risks (data protection, non-discrimination, transparency) through clear rules.

Key Facts & Figures

Regulatory Milestones

• February 12, 2025: Federal Council decides on AI stocktaking and commissions three departments (EJPD, UVEK, EDA, WBF) to draft regulation.
• Timeline: Public consultation draft by end of 2026; parallel non-binding measures (industry solutions, self-commitments) by end of 2026.
• Reference Framework: Council of Europe AI Convention (not yet ratified, but serves as orientation framework).

Governance in Federal Administration

• CNAI (AI Competency Network):

  • Central coordination hub; transition to Federal Chancellery (BK-DTI) as of February 1, 2026.
  • Provides experience exchange, project database, AI terminology, tools.

• Eight Implementation Measures (federal strategy) for AI competency development in agencies.

Regulation Content (planned)

Four pillars according to Council of Europe convention:

1. Transparency: Disclosure of AI deployment and functionality.
2. Data Protection: Compliance with Federal Data Protection Act (DSG) and GDPR-like standards.
3. Non-discrimination: Control of bias and discriminatory outputs.
4. Oversight: Administrative monitoring and compliance verification.

Guidelines & Fact Sheets (already in effect)

• General AI Guidelines for the Federal Government: Establish orientation framework for responsible use.
• Fact Sheets on Generative AI Tools: Support staff in handling LLMs, ChatGPT, etc.
• Core Principle: Responsibility remains with humans; compliance with information security and data protection is mandatory.

Numbers & Scope

⚠️ Data Gap: The article does not provide concrete figures on:

• Number of ongoing AI projects in federal administration (project database mentioned but not linked).
• Budgets for AI development and regulation.
• Number of affected employees.

Stakeholders & Those Affected

Opportunities & Risks

Action Relevance for Decision-Makers

For Federal Administration:

1. Immediately: Stringently monitor implementation of existing AI guidelines; conduct compliance audits for ongoing projects.
2. By 2026: Early participation in public consultation process; provide feedback on regulation draft based on pilot project experiences.
3. Long-term: Build an AI ethics culture; train staff; establish transparency mechanisms for citizens.

For Private Sector:

1. Monitoring: Continuously track public consultation draft (expected 2026); early scenario planning for compliance.
2. Strategic Positioning: Companies already meeting high data protection and transparency standards gain competitive advantage.
3. International Coordination: Synchronize with EU AI Act to avoid double regulation.

For Regulators:

1. Avoiding Over-regulation: Evidence-based thresholds (e.g., strict regulation only for "high-risk" AI).
2. Maintaining Flexibility: Use self-commitments as supplements; account for technology pace.
3. Clarifying Liability: Establish legal clarity on responsibilities before AI controls critical infrastructure.

Quality Assurance & Evidence Review

• [x] Statements sourced from official Federal Council documents (high credibility).
• [x] Responsibility distribution clearly presented.
• [x] Timelines explicitly named.
• ⚠️ Data Gaps: No quantitative figures on AI projects, budgets, or risk extent; effectiveness measurements missing.
• ⚠️ Causality Assumption: Article implies regulation reduces risks without providing evidence.
• ⚠️ Conflicts of Interest: Federal Council has vested interest in AI expansion in agencies; independent evaluation pending.

Supplementary Research

1. EU AI Act (2024):
   European counterpart; Swiss rules should be compatible → https://eur-lex.europa.eu/eli/reg/2024/1689/oj

2. Council of Europe AI Convention:
   International legal framework Switzerland plans to ratify; particularly articles on transparency and human rights protection → Council of Europe AI Convention (CETS No. 236)

3. Swiss Data Protection Act (DSG 2024):
   Already in effect; forms basis for AI regulation → https://www.admin.ch/opc/de/classified-compilation/20202057/index.html

4. OECD AI Policy Observatory – Switzerland:
   International comparative data on AI regulation → https://oecd.ai

5. Critical Counter-Perspective:
   Tech industry associations warn against over-regulation (e.g., swico, digitalswitzerland); see their policy papers on AI compliance costs.

Bibliography

Primary Source:
Federal Chancellery (2025): Artificial Intelligence – https://www.bk.admin.ch/bk/de/home/digitale-transformation-ikt-lenkung/bundesarchitektur/kuenstliche_intelligenz.html

Federal Council Documents (cited):

• Federal Council: AI Regulation: Federal Council Wants to Ratify Council of Europe Convention (Press Release, 02.12.2025)
• Federal Council: Stocktaking on AI