Middle powers like India and Switzerland are shaping global AI governance. Comment.

There is still no binding global treaty on Artificial Intelligence (AI), so its rules are being set mainly through summits and shared norms. Middle powers are neither AI superpowers nor passive rule-takers, and they increasingly set the agenda. India hosted the AI Impact Summit in February 2026 [1], and Switzerland will host the Geneva AI Summit in 2027 [2]. Both show this shift, but their influence so far shapes norms more than hard rules.

How middle powers are shaping AI governance

• Convening power: The PM welcomed Switzerland's offer to host the next AI Summit, to be followed by the UAE [1]. Agenda-setting stays outside a US–China duopoly.
• Shaping norms: In the 5 October 2026 joint statement, both countries committed to responsible, inclusive and human-centric AI [2]. This brings development and access to the centre, alongside safety.
• Credible bridge-builders: India's non-alignment and Switzerland's neutrality fit together well, and Geneva is a hub of multilateral diplomacy [3]. Both are seen as honest brokers.
• Reforming institutions: Both agreed that global institutions need reform [2]. AI governance is part of this wider demand for fairer representation.
• Cooperation that sets an example: The two countries plan joint research in health, clean energy and space, and an agreement with the University of Zurich will build the capacity of Indian science start-ups [2]. This follows a February 2026 commitment to joint research and capacity building [1]. It shows a cooperative, capacity-sharing model that other countries can copy.
• Voice of the Global South: By hosting the summit, India can bring developing countries' concerns to the table: access to computing power, data and skills.

Limits

• Summits produce declarations, not rules that can be enforced.
• Frontier AI models, advanced chips and computing power are concentrated in a few large economies and firms. Middle powers can influence norms but not control the technology.
• Parallel summit tracks risk fragmentation and "forum-shopping", where countries choose whichever forum suits them.
• India's own capacity in computing power and AI regulation is still developing, which limits how much it can lead by example.

Overall, India and Switzerland are shaping how AI governance is framed and continued, rather than enforcing it. The Delhi–Geneva summit sequence should now produce common standards, shared computing resources and capacity-building for developing countries, in line with SDG 9 (innovation) and SDG 17 (partnerships). This would turn "human-centric AI" from a slogan into a working global framework, with middle powers as its main architects.