In the federal administration, the use of artificial intelligence (AI) is coordinated by the Federal Chancellery. Its AI competence network serves as a "single point of contact" for the Confederation's services and relies on seven thematic hubs. In parallel, the Digital Switzerland action plan lists an AI strategy, its implementation plan and projects such as AI in office automation.

This page deals solely with the use of AI by federal services. The general framework, which also applies to companies, is presented on our page on AI regulation in Switzerland.

Key figures

  • 7 thematic competence hubs surround the Federal Chancellery's contact point.
  • 4 months: the interval at which the contact point and the hubs set their work packages.
  • 12 December 2025: date of the concept on the coordination of AI in the federal administration.
  • 130 measures appear in the Digital Switzerland action plan table, consulted on 28 October 2026.
  • 2030: the year associated with the implementation plan of the federal administration's AI strategy.

Who coordinates AI in the federal administration?

Coordination lies with the Digital Transformation and ICT Governance (DTI) sector of the Federal Chancellery. According to the network page published by the Chancellery, this sector houses the central contact point of the AI competence network.

The Chancellery describes this network as the "hub" for AI in the federal administration. It is aimed at employees and administrative units looking for knowledge, guidance, exchanges or support.

According to the same page, the contact point performs five tasks:

  • coordinating the implementation of the AI strategy;
  • answering internal and external enquiries on the use of AI in the federal administration;
  • managing the network and ensuring coordination among its members;
  • providing information and documentation on the use of AI systems, and developing the project database;
  • reporting on the network's activities, organising joint events and spaces for exchange.

How does the AI competence network work?

Specialist knowledge comes from seven competence hubs, each attached to an administrative unit. The Chancellery divides them as follows:

  • Strategy and governance: DTI sector of the Federal Chancellery;
  • Tools and platforms: DTI sector of the Federal Chancellery;
  • Procurement: Federal Office for Buildings and Logistics (FOBL);
  • Continuing education and culture: Federal Office of Personnel (FOPER);
  • Law and regulation: the Chancellery's legal service, with the Federal Office of Justice (FOJ) and the Federal Office of Communications (OFCOM);
  • Information security and cybersecurity: State Secretariat for Security Policy (SEPOS), with the Federal Office for Cybersecurity and the Federal Data Protection and Information Commissioner (FDPIC);
  • Data science: the Data Science Competence Centre of the Federal Statistical Office (FSO).

Each hub has a single point of contact and leads a working group on its topic. Every four months, during joint planning, the contact point and the hubs define concrete work packages. The Chancellery sees this as the way to keep solutions practice-oriented.

All federal offices and administrative units are invited to join the network. Their employees can join the Confederation's AI community, an internal knowledge-sharing space, or a thematic working group.

What rules govern the use of AI by federal services?

The network page refers to several reference texts, without reproducing their content:

  • the ordinance on digital services and digital transformation in the federal administration;
  • the concept aimed at developing coordination for the use of AI in the federal administration, dated 12 December 2025;
  • a press release entitled "Use of artificial intelligence in the federal administration: the Confederation plans targeted measures and strengthens coordination".

Data protection law also applies. The FDPIC ensures that federal bodies do not process personal data unlawfully. In its note of 24 September 2025, it points out that this law applies directly to processing that uses AI.

Security forms the second strand. On 16 October 2026, the Federal Office for Cybersecurity pointed out that carelessly entering confidential information into an AI can expose that data. This office is involved in the network's security hub.

Our analysis: the sources consulted publish neither a list of authorised AI tools nor a ban on any specific service for federal staff. The "Tools and platforms" hub exists, but its choices are not detailed publicly.

What AI projects is the Confederation running?

The Digital Switzerland action plan, published by the Confederation on digital.admin.ch, lists 130 measures, each with a responsible department or actor and a year. Around ten of them concern AI directly:

MeasureResponsibleYear indicated
Strategy for the use of AI systems in the federal administrationFederal Chancellery, DTI sector2025
Implementation plan for this strategyFederal Chancellery, DTI sector2030
Use of AI in office automationFederal Chancellery, DTI sector2026
Proposal to strengthen the coordination of AI use in the federal administrationFederal Statistical Office2025
"CFC-Copilot" for employeesSwiss Federal Audit Office2024
AI stocktakingFederal Office of Communications2025
Monitoring of the AI guidelinesFederal Office of Communications2024
Legislative project implementing the Council of Europe Framework Convention on AIFederal Office of Justice2026
Non-binding measures related to this conventionFederal Office of Communications2026
Deployment of SwissGPTAlpineAI, an actor outside the administration2030

The first four rows concern the internal functioning of the administration. The title "Use of AI in office automation" refers to staff's office software, but the page does not detail its content.

The convention in question is the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law. The table assigns the corresponding legislative project to the FOJ, without specifying its scope.

What does the Swedish example of 28 October 2026 show?

On 28 October 2026, TechRadar Pro, in its French edition, reported that Sweden is banning ChatGPT, OpenAI's assistant, and Claude, Anthropic's, for government work. The site relies on Cybernews. The two tools remain permitted in approved environments.

Microsoft Copilot, Microsoft's assistant, remains usable, with reinforced security controls. According to the government quoted by TechRadar Pro, a limited number of employees will keep ChatGPT and Claude via a "special platform for AI assistants".

The measure follows a review, conducted earlier in 2026, of the AI tools used at the Swedish Transport Administration. It identified uses deemed "problematic" in official tasks. Mikael Wettercrantz, presented as chief information officer, mentions "consequences for information security in Sweden".

The risk described in the article is specific. Without an enterprise contract, confidential data entered into a chatbot can be transmitted to the provider and then retained. Depending on the applicable terms, it can even be used to train the models. The absence of an agreement also complicates audits.

Sweden had already banned DeepSeek, the Chinese AI assistant, for government use, like several other countries, on grounds of data confidentiality.

Our analysis: the Swedish approach designates authorised tools and excludes others. The Swiss sources consulted mainly describe an organisation: a contact point, hubs, a strategy and projects. The risk of data leakage is addressed there by a dedicated hub, which involves the Federal Office for Cybersecurity and the FDPIC.

What this changes for a company or a citizen in Switzerland

The contact point also answers external enquiries. A company, a researcher or a citizen can write to ia@bk.admin.ch. The postal address is that of the DTI sector, Monbijoustrasse 91, 3003 Bern.

For a small or medium-sized enterprise (SME) that sells AI solutions, the Procurement hub matters. It is attached to the FOBL. Our analysis: it is a logical point of contact on the Confederation's AI purchases, but the page does not publish any purchasing criteria.

Organisations drafting their own usage rules can find other models. On 12 October 2026, five Swiss churches and Christian organisations published ten guidelines on AI. The other texts of the legal framework for AI are grouped in our section.

What we don't know yet

  • The content of the federal administration's AI strategy and its implementation plan: the pages consulted give only their titles.
  • The AI tools that federal staff may use, and those that would be prohibited to them.
  • The date and details of the press release announcing "targeted measures".
  • The exact meaning of the year associated with each measure in the Digital Switzerland action plan, which the page does not specify.
  • The number of projects entered in the database managed by the contact point.
  • On the Swedish side, the exact scope of the ban: TechRadar Pro attributes it to the government, while citing a review conducted at the Swedish Transport Administration.

In this series

Sources