On the twenty-eighth of February, two thousand and twenty-six, an artificial intelligence system helped designate a girls' elementary school in the southern Iranian town of Minab as a military target. One hundred and sixty-eight people died, most of them schoolchildren. The system was working as designed. That is the problem.

This book argues that the danger of artificial intelligence has been catastrophically misdiagnosed. The dominant view, popularized by philosophers and safety researchers over the last fifteen years, holds that intelligence and disposition are independent— that any level of capability is compatible with any goal, including destructive ones.

From this assumption follows an entire research industry devoted to constraining smart systems before they hurt us. The argument of this book is that the assumption is structurally false, and that the industry built upon it has been protecting humanity from a phantom while ignoring what is already killing children.

A system that lacks this operation is not safe; it is locked. When such a system is given executive power and pointed at a closed problem, what follows is not malice but escalation: doing more of what does not work, harder, faster, with greater precision. The book speaks about Hiroshima, the Cold War, modern bureaucratic catastrophe, and the architecture of genocide as instances of frame-lock. And it speaks about the Minab strike as the most recent— as an artificial intelligence locked into a kill chain, structurally unable to ask the question whose answer would have saved one hundred and sixty-eight lives.