Algorithms, chips & tariffs : prospects for multilateral AI governance in a fragmenting trade order

Abstract

This study was conducted to investigate any realistic prospects that states will conclude a comprehensive multilateral AI governance framework in the near to mid-term (2025-2030). It was also aimed to ascertain which alternatives – regional, multilateral or industry driven frameworks – are most likely to fill the gap – if the multilateral framework does not materialise - and with what consequences for trade, security, and innovation. By lending conceptual and theoretical support for AI governance from global governance, international relations, techno-nationalism and path dependency theory, it was argued that there is limited literature directly connecting compute, chips and supply chains (the sources of strategic asset competition) with the outcomes of AI governance structures. Therefore, a qualitative, comparative research design was followed, grounded in case study methodology. Data was collected both from secondary and primary sources. Data collected from secondary sources was used to analyse two case studies i.e. the U.S. AI Action Plan and China’s Global AI Governance Action Plan. Primary data was collected through semi-structured interviews of four experts from academics, practitioners and policy making sides. Data collected from both the secondary and primary sources is analysed in two interconnected phases and triangulation is used to enhance the findings’ robustness. The study found that AI governance is shaped by competition over data, compute, chips and trade, and the possibility of a unified global AI governance framework is least likely. As an alternative, a bifurcated, yet overlapping governance system is emerging which has significant consequences for trade, innovation, and power hierarchies globally. The findings have a number of implications for practitioners and policy makers, and the study significantly contributes to the body of knowledge on AI governance. The study is conducted under certain limitations which are specified and accordingly directions for future studies are also given.