From Democratic Control to Generative Risk

A Review of Maximilian Kasy's The Means of Prediction

Abstract

This review examines Maximilian Kasy’s The Means of Prediction as a critical intervention in debates on AI governance. Kasy argues that the objectives of AI systems are shaped by control over the resources required to build them: data, compute, expertise and energy. By shifting attention from human-machine conflict to conflicts among social groups and interests, the book challenges narratives that present AI development as technically inevitable. This review highlights the strength of Kasy’s framework in explaining how AI objectives, prediction and power are connected. It also identifies two limitations. First, the book may understate the importance of technical dimension such as explainability in complex AI systems. Second, its focus on predictive AI leaves open questions about generative and agentic AI, which may reshape the social conditions under which democratic control becomes possible.