It is a collective work: nearly twenty authors, mostly legal and consulting practitioners, walking through the regulation in parts — the risk-tier approach, the prohibitions, high-risk systems and their requirements, the fundamental rights impact assessment, conformity marking, transparency, general-purpose models — and then taking it down to sectors: finance, insurance, education, healthcare, advertising, the public sector.
What I took from it
The role distinction, which is the most transferable part outside Europe. The regulation does not ask what the system does but what part each party plays: whoever provides it does not carry the same obligations as whoever deploys it in their operation. Translated into my work, when a client switches on a model-backed feature inside their management system, the question that opens everything is who answers for the outcome, and the answer is rarely the one the client assumes.
The idea that technical documentation and post-market monitoring are part of the system rather than paperwork. Recording which version answered what, and reviewing after go-live, is what lets you explain one specific case six months later.
And human oversight understood as design: not putting someone there to watch, but defining which decisions the system may not take alone and what the person taking them needs in front of them.
What I left out
Almost everything specifically European, which binds nobody here, and the sector chapters that do not touch my work. Two more warnings: as a many-handed book it repeats itself and the level is uneven; and the book itself states that the English version was machine-translated and then revised, which shows in several passages. Read it for the structure, not the prose.