Safe Artificial Intelligence in Basel: Practical Experiences of Critical Infrastructures at the Cantonal Level

Pour Demain conducted the project "Safe Artificial Intelligence for Basel-Stadt" in 2025 with support from the Christoph Merian Foundation. The project aimed to help critical infrastructure (CRITIS) operators in the safe and secure deployment of AI and to incorporate insights gained into standardisation and legislative processes.

Illustration generated by Ideogram AI on AI in critical infrastructure such as energy service providers

Core Mandate and Participants

Seven CRITIS organisations actively participated in this cantonal project, including representatives from mobility, administration, security, health, finance, as well as electricity and water supply.

The project comprised an analysis of legal frameworks and best practices as of early 2025, two webinars, a full-day workshop in spring 2025, and bilateral in-depth discussions with individual organisations.

Thematic Findings

The participating CRITIS organisations are predominantly in an early AI adoption phase. Central challenges include:

1. Governance: slow transition from bottom-up to top-down approaches, extensive internal coordination procedures

2. Data Protection, Safety and  Security: strong preference for on-premise solutions, scepticism towards cloud providers, demand for "human in the loop" (no automated decisions in safety-critical processes)

3. Resources: limited technical expertise, difficult budget processes for rapidly obsolescing AI infrastructure

4. Regulation: orientation towards the EU AI Act and sector-specific requirements (e.g. FINMA), unresolved liability questions

   
   

Priority governance instruments identified were guidance for large language model usage, a systematic AI inventory with risk classification, and staff training. All participants signalled concrete interest in further cross-organisational exchange.

Impact and Sustainability

As an immediate result of the workshop, bilateral exchange between individual CRITIS organisations on AI topics was strengthened. Recommendations for an inventory of deployed AI systems including their risk classification were directly adopted. Throughout the year, the following topics in particular were pursued further: deploying AI assistants on in-house AI infrastructure, reviewing or deploying cloud AI for official use, and completing AI governance to be applied to these systems.

For February 2026, it is planned to invite Basel's CRITIS organisations to the cantonal format "Data Science and AI Exchange" of the city administration to sustain knowledge transfer. Additional CRITIS organisations from Baselland are to be included. Furthermore, there is existing exchange with other cantons. This was supplemented with the city administration of Zurich, where significant synergy potential with Basel was identified.

The project results were presented to the AI working group of the Swiss Association for Standardisation (SNV) in autumn 2025 and shared with experts on CEN-CENELEC as well as ISO/IEC standardisation work. Further input into national legislative processes was also provided in autumn 2025.

Lessons from the Project

The project confirmed the high interest of CRITIS organisations in practice-oriented exchange formats on AI governance. The combination of expert inputs (legal and technical perspectives), peer learning among CRITIS professionals, and bilateral in-depth discussions proved valuable.

Challenges arose from the time availability of individual CRITIS professionals, leading to short-notice cancellations. For future similar formats, the following is recommended:

1. Modular structure: combination of group exchange for peer learning and bilateral discussions for sector-specific deepening

2. Connection to existing cantonal or sectoral exchange formats for sustainability

   
   

The often resource-constrained non-participation of CRITIS professionals in national and international standardisation processes remains a major challenge for standardisation organisations. This requires structural solutions beyond individual projects (e.g. resource allocation, leave arrangements, external financial support), particularly when state actors wish to rely on robust standards that should be the result of diverse expert participation.

We are happy to be available for questions and exchanges about the project at info@pourdemain.ch. We thank the Christoph Merian Foundation for generously enabling this project.