Global, selective, or both? The case for differentiated cooperation in AI governance

Current debates about international cooperation in AI governance remain both simplistic and muddled, pitting global forms of collaboration against selective alliances among "like-minded countries". We propose a more nuanced and systematic approach to cooperative AI governance based on three considerations. First, different kinds of governance issues lend themselves to different kinds of cooperation. Second, not all AI is created equally: different kinds of AI raise different governance challenges, thus requiring varied forms of cooperation. Third, the same is true for the development and deployment phases of AI systems. Integrating these three factors, we explain why some challenges can only be addressed through global cooperation, while for others selective cooperation is an equally effective, or even more effective, solution. In consequence, governments should not insist on either global or selective cooperation. Instead, they should opt for the scope of collaboration that is most effective for solving any particular governance issue at hand.

The online version contains supplementary material available at 10.1007/s11077-026-09606-y.