Introduction

A CERN (Conseil Européen pour la Recherche Nucléaire) analogue is one proposal for governing Artificial Intelligence (AI) systems, with recent calls for a CERN for AI. However, some proposals are vague, and it is unclear how they relate to each other or the extent to which they rely on the same mechanism. It is also unclear whether the goals are to promote AI safety or capabilities. While drawing a parallel with CERN may be useful to convey the idea of a new large collaboration for AI, there is a danger that we develop a false sense that we know how to implement an international organisation that adequately addresses the challenges of AI. This paper aims to clarify how the mechanism that makes CERN successful for “nuclear research of a pure scientific and fundamental character” [1] can work for AI by performing a mechanism mapping [2] from CERN to CERN for AI.