Send message to Author

Paper short abstract

This paper examines how power asymmetries shape AI governance, showing how formal digital trade rules and informal political economy dynamics—such as lobbying, capacity gaps, and geopolitics—drive development outcomes and constrain regulatory sovereignty over AI in the Global South.

Paper long abstract

As AI increasingly shapes global economic relations, asymmetries between rule-makers and rule-takers have become more pronounced, particularly in the Global South. Integrating political economy analysis into AI governance is therefore essential to understanding—and preventing—the reproduction of power imbalances and economic inequalities within digital economies (Banga and Hernandez 2021; Beyleveld and Sucker 2023).

This paper interrogates the evolving architecture of AI governance by examining how formal rules—such as trade agreements, digital trade chapters, and regulatory standards—interact with informal norms. It employs John Gaventa’s ‘powercube’ framework to examine how ‘visible’, ‘hidden’, and ‘invisible’ forms of digital power are opening or closing spaces of decision-making on digital trade and AI flows; the actors that are shaping the dominant narrative on digital trade; and their incentives and motivations. Adopting a political economy approach, it explains why particular relationships, alliances, and disagreements emerge around digital trade and AI rules.

Drawing on thematic analysis of policy documents, and in-depth elite interviews, including with high-level digital trade policy makers of developing economies, the analysis shows that beyond the text of formal agreements, underlying political incentives, strategic alignments, and informal “rules of the game”—including lobbying networks, capacity asymmetries, and geopolitical interests—play a decisive role in determining whose preferences prevail and how digital trade rules operate in practice. In exploring this, the paper offers a more grounded account of how formal and informal power interact in shaping AI outcomes.