To say that it is challenging to write a book on a topic that is continuously in flux is an understatement. Certainly, this work focuses on the larger picture painted by the impact of algorithmic regulation on the rule of law, thus ensuring the relevance of its insights regardless of technological evolutions – or so I hope. Nevertheless, both the technology and the laws I examined incessantly shifted shape. This meant that, if I cherished the hope of ever publishing this book, I had to find a way to deal with its ever-changing scope, and I had to let go of the idea that I could eternally integrate the latest legal and technological developments into its pages.

When I first started this research project, the General Data Protection Regulation had just entered into force, and there was nary a soul at the European Commission who thought a new regulation, let alone an AI-specific regulation, was needed. The High-Level Expert Group on AI, which I coordinated at the time, had only just commenced its mandate to write non-binding Ethics Guidelines for Trustworthy AI, and most citizens in Europe and elsewhere barely had an idea of the technology’s existence (and of the fact that they were unknowingly already using it or subjected to it on a daily basis).

By the time I was halfway through my research, the European Commission published a proposal for a regulation on AI, thus radically changing the stakes. I eagerly started analysing its aptness to deal with the rule of law challenges I had already identified in the meantime, only to be sorely disappointed. Those challenges were not the object of the Commission’s attention (nor that of anyone else for that matter, unfortunately). Two days before I defended my doctoral thesis on which this book is largely based, the Council of the European Union formally adopted its comments on the Commission’s proposal, introducing a number of substantive changes to the original text. These changes also included some language on ‘general-purpose AI’, to correct the Commission’s wholly understandable omission of tackling the concerns arising from applications like ChatGPT, still inexistent when the proposed AI Act was published in 2021.

A few months after my doctoral defence, the European Parliament finally got its (AI) act together too, and published its own amendments to the Commission’s proposal, introducing yet another set of (arguably more sweeping) changes. Most importantly, it included a broad number of provisions that I, together with others, had fiercely advocated for in a stern critique on the original proposal. While this somewhat assuaged my initial disappointment, it soon remained evident that many of the problems I initially outlined remained unsolved. The finalisation of the AI Act and the noticeable sense of relief that followed its adoption – despite the many flaws it contains – thus only made me more determined to relay the results of my research in this book, and to caution against the mistaken optimism that EU legislation is now sufficiently comprehensive to tackle the threat of algorithmic rule by law. Evidently, both the law and the technology continue to evolve, and this book’s cut-off point is July 2024. Only time will tell what the future of AI governance will bring; yet it is my hope that this book will contribute to its improvement.