On 2 October 2026, National Council members from all six parliamentary groups, including Gerhard Andrey (Greens/FR) and Isabelle Chappuis (Centre/VD), filed a joint motion on the 2027 AI Summit in Geneva. Reporting by blue News says it asks the Federal Council to raise its ambition, give the summit more resources and explicit goals, and aim for verifiable results. It also asks that the summit prioritise global AI risks, including extreme ones. One author is quoted as saying that "simply acting as a good host is not enough."
The motion lands at an awkward moment. The summit is a foreign-policy showcase, but Switzerland's domestic AI rulebook is still unwritten. The strongest case for the motion deserves a fair hearing before we say where we think it goes wrong.
The case for more ambition
The motion's logic is hard to dismiss. The summit series has produced a lot of declarations and few commitments anyone can check. Switzerland has real convening credibility: neutrality, Geneva's standing in multilateral diplomacy, and a strong research base. If a host with that position does not push for measurable outcomes, nobody else is likely to. Concern about extreme risks is also not fringe. Legislators from the Greens to the centre-right sharing a position suggests the worry is mainstream, and a summit that only celebrates opportunity would leave that concern unaddressed.
What the government has actually set up
According to the BAKOM summit page, the fifth international AI summit will take place at Palexpo in Geneva on 21 and 22 June 2027 under the theme "Bridging Innovation and Trust." It has two complementary priorities: using AI for prosperity and progress, and promoting trustworthy, responsible and safe use. Fabiola Gianotti, CERN's Director-General from 2016 to 2025, has been appointed as the Federal Council's delegate. A series of "Road to Geneva" pre-events will gather input from governments, academia, business and civil society.
When the Federal Council announced the summit on 25 June 2025, it named three aims: a role for Switzerland in shaping AI use and international governance, a stronger position as a research and innovation hub, and attention to international law and fundamental rights. It expected no additional cost to the Confederation, with funding from the existing budget, third-party contributions and sponsorship. The BAKOM page does not state a budget, so we cannot say how the motion's call for "more resources" would translate into francs.
The government's stance on regulation, as reported by blue News, is that "regulation should be kept to a minimum so as not to hinder innovation." Federal Councillor Albert Rösti wants to present Switzerland as a hub for research and innovation. That is a coherent position. It is also one that sits uneasily with a summit whose headline task becomes the catalogue of extreme risks.
The domestic gap
The more consequential Swiss AI decision is not in Geneva. On 12 February 2025, the Federal Council decided that the Council of Europe's AI Convention "will be incorporated into Swiss law." It chose a sector-specific route: legal changes sector by sector where possible, with broad cross-sector rules only for core fundamental-rights areas such as data protection, and non-binding tools such as self-declaration agreements and industry solutions as a supplement. The Justice and Police department, with DETEC and the Foreign Affairs department, was to prepare a consultation draft covering transparency, data protection, non-discrimination and oversight by the end of 2026. A separate implementation plan, which is to check compatibility with Switzerland's main trading partners, is due on the same timetable.
As of this writing the consultation draft has not been published. That matters for the motion. The problem with demanding "verifiable results" from an international summit is that verification needs a baseline. A host that cannot yet show its own binding commitments will find it hard to persuade 190-odd governments to accept checkable ones. Credibility on measurable outcomes starts at home.
The Greens criticised the February 2025 approach as "toothless and hesitant", while the union Syndicom wanted coverage of private firms and mandatory impact assessments. Those objections are the other half of the cross-party picture: the same parliament that wants a tougher summit is split on how tough domestic law should be.
Where proportionality points
We think the sector-specific approach is the right default. Switzerland's economy runs on pharmaceuticals, finance, machinery and precision engineering, and each has an existing regulator that understands its risks. Amending healthcare or transport rules is more targeted than a horizontal AI act, and it avoids regulating harmless uses with rules written for dangerous ones. It also keeps Swiss firms close to the rules of their largest trading partners, which the implementation plan is meant to test.
But proportionate does not mean vague. If the summit is to produce results anyone can check, three things would help, and none requires heavy regulation:
- Publish the draft on time. A consultation draft before the end of 2026 would give the summit's host a concrete domestic position.
- Define "verifiable" narrowly. Reporting commitments, shared evaluation methods and incident-sharing are checkable. Open-ended pledges about extreme risks are not.
- State the budget. If the motion wants more resources, Parliament should see what the current budget-neutral plan can and cannot deliver.
The risk in the motion is mission creep. A summit that tries to settle extreme-risk governance in two days will either produce another declaration or alienate the governments, companies and researchers whose participation is the point of holding it in Geneva. A narrower set of verifiable deliverables would be both more ambitious in practice and easier to deliver.
The Federal Council now has to answer the motion. Its best reply would pair a firmer summit agenda with a published domestic draft, so that Switzerland can ask others for what it has already done itself.