Book Review: The Ethics of AI: Power, Critique, Responsibility

Rainer Mühlhoff’s The Ethics of AI: Power, Critique, Responsibility, published by Bristol University Press in 2025, is a timely and thought-provoking examination of artificial intelligence from a philosophical and social perspective. At 232 pages, the book moves beyond the common tendency to discuss AI ethics primarily in terms of abstract principles such as fairness, transparency, and safety. Instead, Mühlhoff asks a more fundamental question: who has power through AI, how is that power exercised, and who is responsible for its consequences? The book is openly available under a Creative Commons licence, making it particularly valuable for students and researchers.

The central argument is that AI should not be understood as a neutral technological tool. According to Mühlhoff, AI systems are embedded within social, economic, political, and technological structures that influence how they are designed and used. The author therefore proposes a “power-aware ethics of AI”, encouraging readers to examine the relationships between technology and power rather than concentrating only on the behaviour of individual users or developers. The book's introduction explicitly frames the project around what it means to do a power-aware ethics of AI. 

The book is organised into three major sections: The Power of AI, The Power of Prediction, and The Power of Control. This structure is one of its strongest features because it shows how AI can exercise influence at different levels. The first section examines the nature of AI itself, human involvement in supposedly automated systems, the influence of technological design, and the relationship between AI and human subjectivity. Particularly important is the chapter on Human-Aided AI, which challenges the assumption that AI systems are completely autonomous and demonstrates the continuing importance of human labour and participation. 

The second section focuses on prediction, one of the most important functions of contemporary AI. Mühlhoff examines AI systems as prediction machines and explores the consequences for privacy and knowledge. The concept of predictive privacy is particularly significant because modern AI can infer information about individuals even when they have not explicitly provided it. This raises questions that traditional concepts of privacy may not adequately address. The book therefore demonstrates that AI can create new forms of knowledge and influence simply by analysing patterns in large quantities of data.

The final section, The Power of Control, examines AI cybernetics, opacity, algorithmic bias, and collective responsibility. These chapters are particularly relevant to contemporary debates about AI governance. Mühlhoff argues that the difficulty of understanding how complex machine-learning systems reach particular outcomes can create serious problems of accountability. If an algorithm influences employment, financial opportunities, social services, or other important decisions, determining who is responsible for an unjust outcome becomes increasingly difficult. The chapter on collective responsibility therefore moves the discussion away from blaming a single programmer or user and toward examining the wider systems in which AI operates. 

One of the book's greatest strengths is its interdisciplinary approach. Mühlhoff combines philosophy, ethics, technology studies, political thought, and examples from contemporary digital systems. He is particularly interested in how AI can influence everyday behaviour through mechanisms such as prediction, optimisation, targeting, and design. Consequently, the book is not concerned only with hypothetical future scenarios in which machines become autonomous or superintelligent. Instead, it focuses on the forms of power that already exist in today's AI systems. This makes the book highly relevant to current discussions about AI regulation, social media, data protection, algorithmic discrimination, and corporate power.

There are nevertheless some limitations. Readers seeking a technical explanation of machine learning will find relatively little practical programming or mathematical material. The book is fundamentally a work of critical AI ethics, rather than an AI engineering textbook. In addition, its strong emphasis on power and critique means that questions about the potential benefits of AI sometimes receive less attention than questions concerning its risks and social consequences. Readers interested in developing a balanced framework for responsible AI may therefore wish to complement it with literature on innovation, human-centred AI, and the positive social applications of technology.

Overall, The Ethics of AI: Power, Critique, Responsibility is an important contribution to contemporary AI ethics. Its most significant contribution is its insistence that ethical AI cannot be achieved simply by adding principles or technical safeguards to existing systems. We must also examine the structures of power, responsibility, and control within which those systems operate. This makes the book especially valuable for research into AI governance, human dignity, privacy, algorithmic bias, social justice, technology and society. 

Full book: Download *The Ethics of AI* — JSTOR Open Access