Switzerland - AI Systems Strategy

Overview

The Strategy: Use of AI systems in the Federal Administration (SB021) is a DTI part‑strategy prepared and published by the Federal Chancellery to set the foundational rules and principles for all AI systems used by the Swiss Federal Administration. It defines a lifecycle approach to AI governance, establishes three priority action fields — competencies, trust and efficiency — and presents a vision of AI as a tool to improve public services while safeguarding democracy, fundamental rights and the rule of law. The strategy is published by the Federal Chancellery; the official PDF and web page are available via the Federal Chancellery portals (Bundeskanzlei: Strategy PDF) and the DTI guidance pages (BK: Strategy page). The strategy is intended to be implemented via a separate, detailed measures plan and through existing operational regulations.

Definitions

SB021 supplies operational definitions to create a common understanding across departments. A KI‑system (AI system) is defined as a machine‑based system that infers, from inputs, outputs such as predictions, content, recommendations or decisions that may affect physical or virtual environments and can have varying autonomy levels. The document frames a 'life cycle' including plan & design; data collection and processing; model building/adaptation; testing/validation; deployment; operation/monitoring; decommissioning. It references CNAI terminology and international sources (OECD AI principles) to harmonise definitions and foster interoperability of governance approaches.

Governance and Institutional Framework

Governance responsibility is allocated to the lead DTI function within the Federal Chancellery, with cross‑departmental coordination shaped by CNAI (the federal Competence Network for AI) and active engagement from affected departments (e.g., EJPD, UVEK, WBF). The strategy is issued as a partstrategy under Article 17 of the VDTI and is integrated into the broader Strategy Digital Federal Administration and the national 'Digitale Schweiz' programme. Governance elements include central guidance and blueprints, service‑level standard components, department‑level responsibilities, and a mandate to create department‑spanning coordination structures and knowledge exchanges. Technical governance requires integration with information security (ISG/ISV), data protection obligations (DSG), and procurement/contracting rules. For authoritative documents and contact points, see the Federal Chancellery DTI pages (DTI – Federal Chancellery) and CNAI (CNAI terminology).

Key Focus Areas

The strategy sets out three interdependent handlungsfelder: (1) Competencies to ensure staff, leaders and service providers hold the necessary technical, legal and operational skills; this includes role‑based training, service support and reuseable architecture blueprints. (2) Trust mechanisms to ensure legal and ethical compliance, transparency, documentation and privacy safeguards; this includes life‑cycle documentation, processes to assess generative models (language models), and tools to handle sensitive data securely. (3) Efficiency improvements centred on planning, cost/benefit analyses, pragmatic automation of process steps, standard services, and reuseable AI components. Across these focus areas the strategy advances concrete objectives: e.g., by 2030 prioritise cost/time savings in defined use cases; create standard services for recurring administrative tasks; ensure modular, interoperable solutions that respect digital sovereignty. The strategy emphasises human‑centred design, sustainability (economic, ecological, social), robustness and resilience of AI systems, and the need to maintain transparency and explainability appropriate to the context of use.

Implementation Framework

Implementation is assigned to the Bundeskanzlei (DTI) in cooperation with CNAI and the departments; a separate implementation plan with concrete measures and timelines is to be published. Implementation tools include: department‑level planning documents with objective and KPI statements, cost/benefit and risk assessments, mandatory documentation templates covering lifecycle and data provenance, standard service catalogues (central model hosting, APIs, secure data services), and training curricula. The strategy instructs departments to integrate AI governance into procurement, security, and personnel policies. It requires that deployments of generative or high‑impact models be subject to specific review for traceability, training data provenance and documentation to enable downstream oversight. Technical integration must align with existing ISG/ISV information‑security rules and with the forthcoming DigiV provisions now in force for DTI guidance.

Monitoring and Evaluation

Monitoring is to be realised by reporting mechanisms coordinated by the DTI and CNAI, with periodic public transparency on the types of AI systems used and aggregated risk profiles. Evaluation criteria include achievement of stated efficiency gains, compliance with legal/ethical principles, robustness of security measures, and uptake of training/competency measures. The strategy calls for an implementation plan including monitoring indicators, milestone reviews and mid‑term evaluations; departmental self‑assessments and central audits are anticipated as part of oversight. Public transparency is balanced with security and privacy requirements; aggregated reporting and cataloguing of systems will be used to inform stakeholders and policy makers.

Penalties, Liability, and Appeals

SB021 itself is an internal part‑strategy and does not create new criminal or civil penalties; rather it establishes mandatory internal governance obligations for administration units. Legal liability and enforcement remain anchored in existing law (e.g., liability rules in Swiss civil law, data protection obligations under the Federal Act on Data Protection) and relevant administrative disciplinary regimes. The document foresees remedial measures, administrative withdrawal or suspension of AI system approvals, and internal accountability mechanisms where non‑compliance poses legal or operational risk. Appeal pathways for affected persons remain under established legal procedures; where statutory obligations arise from later implementing legislation, legal remedies under those statutes would apply.

Relationship to Other Instruments

The strategy is explicitly situated within the Strategy Digitale Bundesverwaltung and the national 'Digitale Schweiz' priorities. It is a partstrategy pursuant to the VDTI (Article 17) and must be implemented in alignment with the Digitalisation Ordinance (DigiV), the Federal Data Protection Act (DSG), the Information Security Act (ISG) and related ordinances and procurement rules. The document references previous federal AI guidance (e.g., 2020 Leitlinien) as complementary inputs and indicates cooperation with EJPD, UVEK, EDA and WBF on legal/regulatory and sectoral matters. It also anticipates and coordinates with the federal work programme to implement the Council of Europe AI Convention and proposed national AI regulatory steps to be prepared through 2026.

International Alignment

The strategy emphasises alignment with international principles and norms (OECD AI Principles, Council of Europe AI Convention) and cooperation with international partners on interoperability, norms and procurement best practice. It instructs CNAI and DTI to consider international standards, certification schemes and sectoral regulatory work when developing technical and governance templates. The approach aims to preserve Swiss competitiveness and to facilitate cross‑border data flows under appropriate safeguards while observing international human rights and non‑discrimination commitments; coordination with federal foreign affairs and justice departments is specifically foreseen for international legal alignment.

Implementation Timeline

Milestone	Target date	Notes
Bundesrat commission to BK to develop strategy	2024-09-13	Political mandate to develop partstrategy
Strategy enters into force / public rollout	2025-03-21	Public reporting and press coverage
PDF published on BK website	2025-04-11	Official downloadable SB021 PDF
DTI to publish implementation plan	2025-12-31	Plan to detail measures, services and KPIs (timeline per strategy text)
Mid‑term review and monitoring report	2027-06-30	Central review of KPIs and risk management
Full evaluation / alignment with national AI regulatory developments	2026-12-31	Alignment with national AI regulatory roadmap/Europarat convention implementation

Compliance Checklist

Requirement	Compliant (Y/N)
Unit‑level AI use plan with objectives and KPI	Yes / planned
Documentation of life‑cycle and data provenance	Yes / mandatory
Role‑based staff training implemented	Planned / in progress
Risk assessment & DPIA for sensitive uses	Required
Technical security controls aligned with ISG	Required
Review of generative models before deployment	Required
Use of central standard services where applicable	Encouraged / mandated when available

Sources and References

Source	Type
Strategie Einsatz von KI‑Systemen in der Bundesverwaltung (SB021) — PDF	Primary Source
BK: SB021 web page	Primary Source
Digitale Schweiz — Measure page	Primary Source