Global AI governance: barriers and pathways forward

Abstract Recent advances in artificial intelligence (AI), combined with a proliferation in use, have led to a newfound emphasis on strengthening the global governance of AI. In this article, we assess the prospects for stronger global AI governance and consider potential pathways forward. We map the nascent landscape of international regimes focused on AI governance and conclude that a governance deficit remains due to the inadequacy of existing initiatives, gaps in the landscape and difficulties reaching agreement over more appropriate mechanisms. First-order cooperation problems stemming from interstate competition and second-order cooperation problems from dysfunctional international institutions problematize overcoming current deficiencies in global AI governance. In light of these cooperation problems, we evaluate two pathways for strengthening global AI governance: (a) developing new centralized international AI institution(s); and (b) strengthening coordination between, and capacities of, existing institutions. We argue that strengthening the existing weak ‘regime complex’ of international institutions is the more politically legitimate and viable path forward. Improving coordination between, and capacities of, existing international institutions governing AI would support inclusive and mutually reinforcing policy change that can mitigate a range of risks associated with these technologies. As first steps for strengthening the weak AI regime complex, we recommend that the OECD is foregrounded as a centre of expert AI knowledge so that it can facilitate peer pressure among states and harmonization of policies. We also encourage scrutiny of different nodes in the regime complex to ensure that they are fulfilling appropriate functions based on their democratic mandates.