Digital sovereignty: migrating to Swiss and European tools, where to start

On 7 July 2026 the canton of Geneva published its digital sovereignty strategy, led by the Department of Institutions and Digital Affairs. It addresses the cantonal administration, but the questions it asks apply to any business: which providers do we depend on, what happens if one of them changes its terms, and where is our data really?

This series offers a practical answer. I drew up an inventory of 17 services commonly used in a French-speaking Swiss SME, from WhatsApp to AWS via ChatGPT, and for each one the Swiss or European equivalents available in September 2026. This article presents the method and the overview; the six following articles detail each family of tools.

What the Geneva strategy says

The text rests on four principles: adapt requirements to data sensitivity, keep real control over the systems used, favour scalable solutions, and ensure technical, economic and environmental sustainability. Concretely, the canton wants to assess technological risks better, avoid irreversible dependencies, limit the number of critical providers and favour diversified solutions. The chosen wording: the public service must “remain free in its digital choices and bear their consequences over time”. The document is under consultation with the communes and autonomous public institutions.

None of this is specific to the public sector. A clinic, a practice, a services company have the same dependencies, often without having listed them.

Sovereign does not mean Swiss at any price

In the inventory, each solution carries an origin. Switzerland: Swiss company and hosting, the priority. EU: company from a European Union country. Europe outside the EU: United Kingdom and Norway, which benefit from a GDPR adequacy decision. International (open source): publisher outside Europe, but free software you can host yourself in Europe. The classification is a decision tool, not a dogma: free software of American origin hosted in Zurich is often more sovereign than a closed Swiss service whose data cannot be exported.

The inventory: 17 services, 59 equivalents

Two observations. First, Switzerland is well placed on messaging, e-mail, office software, storage and cloud, with Infomaniak, Proton, Threema, Exoscale and, for language models, Apertus from EPFL and ETH Zurich. Second, on technical AI tools (orchestration, vector databases, gateways) there is no Swiss equivalent: sovereignty comes from free software, often German or Dutch, hosted yourself.

The method: by sensitivity, in stages

• Inventory. List the services used, who uses them, what data passes through them. One sheet is enough; it is the one used here.
• Rank by sensitivity. Patient or client data, contracts, finances on one side; drafts, public communication, monitoring on the other. The first Geneva principle: requirements follow sensitivity.
• Choose one solution per service. Only one, to stay readable, with its origin, its pricing model and the ability to export.
• Migrate in stages. Start with what is simple and reversible (browser, search, office software), finish with what structures the work (e-mail, cloud, AI tools in production).
• Operate and measure. A migration is judged six months later: costs, incidents, adoption by the teams.

What it changes for AI

This is the most moving part. The best-performing models remain American, but European and Swiss APIs (Infomaniak AI Tools, Mistral, Scaleway, OVHcloud, IONOS) serve good open models, and a gateway such as LiteLLM lets you switch provider without rewriting the application. In the solutions I put in place, that reversibility is what I aim for: the model is a replaceable component, not a foundation.

The series

• Messaging, e-mail and video calls
• Office software and file storage
• Cloud and hosting
• AI: APIs, assistants, models and images
• AI tools for technical teams
• Browser, search and design