The second proposal reduces dependence but may be prohibitively expensive. Choosing either proposal entails trade-offs. A few miles away, Samuel, a resident of the capital city, walks past a new array of AI-powered cameras. At first, he welcomed them; they promised to reduce traffic congestion and speed up emergency response. But during a recent peaceful protest over rising living costs, the cameras were used to identify and track participants, turning a tool for public safety into an instrument of surveillance. These scenes illustrate two of the four axes along which AI is changing power dynamics: between countries, between governments and corporations, between corporations and individuals, and between citizens and states. Although AI presents developing countries with an opportunity to transform their economies, power and politics will determine whether this opportunity is realized. The opportunity can be lost in three ways. Countries may fail to access layers of the AI stack because of geopolitics.1 They may waste the opportunity because of misaligned incentives between governments and corporations, as well as between corporations and individuals. And governments may misuse the opportunity in their relationship with citizens. What can go wrong along each axis depends on where power lies. Between countries, power stems from control over key inputs in the AI value chain: semiconductors, data centers, cloud networks, data, and critical minerals. Between corporations and governments, power arises from firms’ control over frontier technologies and governments’ authority over markets, regulation, and procurement. Between corporations and individuals, power reflects market structures that produce concentration, leaving users and workers with limited ability to push back on the terms companies set. Between states and citizens, power stems from the state’s capacity to surveil, influence, and shape public discourse. Across all four axes, structural incentives can lead outcomes to diverge from the public interest. 202 AI does not determine political outcomes on its own. It expands the range of possibilities for surveillance as well as transparency, for manipulation as well as mobilization, and for concentration as well as diffusion of power. The outcomes depend on who controls the technology, who sets the rules for its use, and who has the capacity to contest those rules. Where institutions are weak and civic space is constrained, AI is more likely to amplify authoritarian tendencies; where the rule of law is robust and civil society is active, it can strengthen democratic accountability. This challenge places many developing countries in a difficult situation. They need AI to improve public services and accelerate economic growth but lack the resources to build domestic AI capacity and often have a weak regulatory environment to govern its use. This leaves them vulnerable to both state overreach and unfavorable terms of engagement with global technology firms, while limiting their influence over the rules being written. This chapter examines how AI is reshaping these four axes of power, with particular attention to developing-country contexts. These four dimensions are neither exhaustive nor independent. They intersect and reinforce one another in ways that shape development outcomes. The chapter draws on evidence where such evidence exists and identifies gaps where the evidence is limited (or nonexistent). Dynamics between countries: A technology with geopolitical stakes States have long competed for control over transformative technologies. Naval power shaped imperial rivalry in the early modern period. Nuclear weapons and space technology became critical arenas of competition during the Cold War. Today, governments increasingly view AI as a foundational technology that shapes economic activity, social and political structures, national World Development Report 2026