The greatest global powers in AI right now are also powers at odds with one another. The resources for this, which could be the defining factor in training frontier AI, are concentrated in the United States and China, whose strategic rivalry could impede how the benefits of AI are expanded. These great powers have little desire to share their advancements with no promise of a guaranteed return, even if it means cutting out small and middle powers who might be their allies.
Calling upon the case studies of Cold War negotiations over satellite resources, Wim Howson Creutzberg claims that middle powers can successfully advance benefit-sharing agreements by leveraging the strategic interests of great powers, especially if they focus on gaining access to non-zero-sum resources, such as training data and software needed to deploy advanced AI systems.
