This critical review examines Eliezer Yudkowsky and Nate Soares’s If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All, focusing on its treatment of artificial intelligence risk, evidentiary certainty, precautionary regulation, and global governance. The book presents a compelling warning that superhuman AI developed through current machine-learning approaches could create irreversible and potentially existential harm. Its principal strengths lie in making complex alignment challenges accessible, distinguishing interpretability from control, and placing the burden of demonstrating safety on developers of high-risk systems. However, this review argues that the book’s claim of near-certain human extinction exceeds the evidence it presents. Its hypothetical scenarios establish plausibility and urgency but do not adequately model defensive capabilities, detection mechanisms, coordinated shutdowns, or alternative outcomes. The proposed regulation of advanced computing resources also raises concerns relating to institutional legitimacy, concentrated authority, equitable participation, and the technological dependence of developing countries. The review concludes that the book is most persuasive as an argument for precautionary governance rather than as a definitive prediction of extinction. Its central contribution is the proposition that uncertainty surrounding catastrophic harm should strengthen not weaken the requirement for rigorous safety evidence before advanced AI systems are deployed.