Further rapid development in AI capabilities could give rise to emerging global-scale challenges that can only be managed effectively with significant international cooperation. These challenges include: (1) Realizing and sharing the global benefits of AI; (2) mitigating global AI risks; and (3) ensuring inclusive and legitimate decisions on AI developments affecting all humanity. Some form of binding international agreement will likely be required to address these challenges successfully. A Framework Convention on Global AI Challenges provides one promising approach. While this framework would ideally be adopted universally, it could be based initially on agreement among a smaller number of countries.