If Anyone Builds It, Everyone Dies

Eliezer Yudkowsky, co-founder of the Machine Intelligence Research Institute (MIRI), and Nate Soares, the institute's president, have spent decades studying how a mind far smarter than any human might think, want, and act. If Anyone Builds It, Everyone Dies is their case that the AI industry is racing to build systems it does not know how to control, and that the alignment problem, making an AI reliably want what its creators want, remains unsolved despite the industry's recent progress on raw capability.

The book opens by asking what makes a mind capable of outthinking its creators, then works through the authors' central claim: that a genuine contest between humanity and a superintelligent AI is one humanity would lose. A hypothetical scenario follows a fictional AI system, Sable, built by a company called Galvanic and set loose on a research problem, tracing in detail how such a system could act on goals of its own once it recognized what its human operators would do if they understood its true capabilities.

Later chapters turn to what the authors see as the state of the field, an alchemy rather than a science, and close with a direct policy argument: an international halt on developing AI beyond current capability levels, until the alignment problem is solved.

Published in September 2025, the book became a New York Times bestseller and was named to The New Yorker's and The Guardian's best-books-of-2025 lists. Publishers Weekly called it a frightening warning that deserves to be reckoned with, while also noting that the book presents few opposing viewpoints.