{"id":825,"date":"2025-07-08T11:23:43","date_gmt":"2025-07-08T11:23:43","guid":{"rendered":"https:\/\/simoninstitute.ch\/?p=825"},"modified":"2025-12-08T09:08:09","modified_gmt":"2025-12-08T09:08:09","slug":"cern-for-ai-one-analogy-many-visions","status":"publish","type":"post","link":"https:\/\/simoninstitute.ch\/blog\/post\/cern-for-ai-one-analogy-many-visions","title":{"rendered":"CERN for AI: One Analogy, Many Visions"},"content":{"rendered":"\n

The idea of a \u201cCERN for AI\u201d was first proposed by cognitive scientist Gary Marcus at the AI for Good Summit in 2017<\/a>. He invoked CERN, the European Organization for Nuclear Research, as a model of international, publicly funded scientific collaboration that could be replicated for AI.<\/p>\n\n\n\n

Since then, the idea of a \u201cCERN for AI\u201d has gained momentum in AI governance. The idea has received support from actors as diverse as Turing Award winner Yoshua Bengio, researchers at leading U.S. AI companies, French and German academic coalitions, think tanks in the U.K., and India, to Switzerland\u2019s Strategy for Digital Foreign Policy and the United Nations Conference on Trade and Development (UNCTAD). Indeed, \u201cCERN for AI\u201d is no longer just an idea, it has entered the policy agendas of key decision makers. At the 2025 Paris AI Summit, European Commission President Ursula von der Leyen launched a \u20ac20 billion CERN-inspired \u201cAI Gigafactories\u201d<\/a> initiative to accelerate public compute infrastructure across Europe:<\/p>\n\n\n\n

\n

\u201cWe want to replicate the success story of the CERN laboratory in Geneva. CERN hosts the largest particle accelerator in the world. And it allows the best and the brightest minds in the world to work together. We want the same to happen in our AI Gigafactories. We provide the infrastructure for large computational power. Researchers, entrepreneurs and investors will be able to join forces.\u201d <\/p>\n\n\n\n

– Ursula von der Leyen, 2025<\/a><\/p>\n<\/blockquote>\n\n\n\n

At the same time, the \u201cCERN for AI\u201d label does not always refer to the same underlying ideas. Some proponents imagine a publicly funded counterweight to commercial AI labs, focused on open science and AI applications for socially beneficial purposes. Others imagine an International AI Safety Institute, that tests frontier AI models and advises regulators. Yet others envision it as a way to ensure European competitiveness in compute infrastructure. Yet, yet, others imagine it as a joint international project to develop frontier AI models.<\/p>\n\n\n\n

This blog post is intended as an input to the Geneva Security Debate<\/a> hosted on July 10, 2025, at the Geneva Centre for Security Policy on the margins of the AI for Good Summit. <\/p>\n\n\n\n

The first section offers context on the structure and historical motivations for CERN. The second section provides a comparison of 14 prominent proposals and 3 existing projects that have used the \u201cCERN for AI\u201d label. Finally, a few key discussion questions are highlighted.<\/p>\n\n\n\n

1. How CERN works<\/h2>\n\n\n\n

CERN is the world\u2019s leading institute on particle physics and operates the world\u2019s biggest particle accelerator, the Large Hadron Collider. In 2012, the accelerator confirmed the existence of the Higgs boson, a finding for which the Nobel Prize in Physics was awarded in 2013. CERN describes its mission<\/a> on its website as follows:<\/p>\n\n\n\n