Nate Soares, an American AI safety researcher and the current president of the Machine Intelligence Research Institute (MIRI), serves as the book’s co-author. With a background in software engineering at Microsoft and Google, Soares brings both industry experience and a long history of theoretical alignment research to the text. He helps bridge the book’s technical arguments with its concrete policy proposals, translating the abstract failure modes of AI systems into the book’s argument for global governance and control.

Soares’s credibility is rooted in his leadership at MIRI and his authorship of foundational papers on AI alignment theory. His work has helped shape MIRI’s research agenda and its strategic shift from a primarily technical institute to a public advocacy organization. This transition reflects his perspective that the incentives driving progress in AI capabilities are rapidly outpacing the science of AI safety. The book’s sense of urgency is fueled by this conviction that the competitive dynamics of the AI industry make catastrophic failure increasingly likely without decisive intervention.

The core of Soares’s argument focuses on the “before/after” gap in AI development. He emphasizes that any solution to the alignment problem must be solved before a superintelligence is created, as there will be no opportunity for trial and error after. An error on the first true test would be irreversible. This framing is crucial to the book’s critique of iterative, trial-and-error engineering approaches, which are standard in software development but, in his view, dangerous when applied to superintelligence. He argues that the problem is fundamentally “cursed” by its one-shot nature.

Consequently, Soares’s authorial purpose is to connect these profound engineering limits to specific, enforceable mechanisms of control. He advocates for international treaties to monitor and regulate the specialized computer hardware (GPUs) required for large-scale AI training. By focusing on the physical infrastructure of AI development, he presents hardware oversight as a concrete path for policy makers to halt the technological race. His contribution is to ground the book’s existential warning in a pragmatic, though politically difficult, call to action, making the case that physical controls represent one of the few enforceable intervention points available to governments.

Sam Altman, the co-founder and CEO of OpenAI, is a pivotal figure in the narrative of If Anyone Builds It, Everyone Dies. While not presented as an antagonist in a traditional sense, he symbolizes the powerful institutional forces and incentive structures that the authors believe are driving humanity toward extinction. His career trajectory, from president of the influential startup incubator Y Combinator to the head of the world’s leading AI lab, anchors his role as a key architect of the current AI landscape.

The authors use Altman to personify the competitive dynamic that, in their view, pervades the AI industry. His public advocacy for aggressive development timelines and rapid scaling of AI capabilities exemplifies the very trend the book seeks to halt. Altman serves as the book’s primary case study for how the private, commercial incentives of a company like OpenAI can diverge from the collective safety interests of humanity. The authors note with a tone of regret his connection to their own work, quoting his claim that Yudkowsky was critical in the decision to start OpenAI, a decision Yudkowsky himself considered a “terrible, terrible idea” (6).

Ultimately, Altman’s role in the text reflects the tensions at the center of contemporary AI development. His leadership at OpenAI was instrumental in accelerating the wave of large language models that brought AI into mainstream global discussion. At the same time, this success intensified the policy debates and competitive pressures that concern the authors throughout the book. By embodying a culture of rapid technological acceleration, Altman represents the central challenge the book addresses: a world where increasingly powerful AI systems continue to be developed despite unresolved concerns about their long-term risks and controllability.

Demis Hassabis is the British co-founder and CEO of Google DeepMind and a Nobel Prize-winning AI researcher. In the book, he represents the dramatic and often unpredictable leaps in AI capabilities that the authors argue challenge assumptions about slow, manageable progress. The authors frame his work as evidence supporting their argument that superhuman systems could rapidly outmaneuver human defenses.

The book points to Hassabis’s major projects, AlphaGo and AlphaFold, as examples of AI systems achieving tasks once considered intractable by experts. The success of AlphaFold in solving the protein folding problem is particularly significant, as it demonstrated AI’s ability to contribute to a scientific challenge that had resisted traditional approaches for decades. The authors cite this breakthrough to bolster their thesis that observers consistently underestimate the potential for near-term, paradigm-shifting advances. Hassabis’s track record is used to counter the argument that progress will be gradual, reinforcing the claim that humanity is likely to be caught off guard by the sudden arrival of superintelligence. His work therefore functions in the text as evidence that major advances in AI capability can emerge faster than governments, institutions, and even experts anticipate.

Geoffrey Hinton, a British-Canadian computer scientist and a recipient of the Turing Award, is known as one of the “godfathers of AI” for his foundational work on deep learning. He lends authoritative weight to the book’s warnings, serving as an example of a prominent AI researcher who has publicly expressed concern about advanced AI systems. The authors begin their introduction by citing him as a key signatory of a 2023 open letter that identified extinction from AI as a global priority.

Hinton’s relevance is magnified by his decision to leave his position at Google in 2023 to speak more freely about the dangers of uncontrolled AI scaling. This act is presented as evidence that concerns about existential AI risk extend beyond a small group of researchers, particularly because Hinton is one of the field’s most influential pioneers. His public statements are used by the authors to counter dismissals of AI risk as mere “industry advertising,” supporting the book’s argument that warnings about advanced AI are increasingly being voiced by researchers directly involved in the development of modern machine learning systems.

Yann LeCun, a French-American computer scientist and fellow Turing Award laureate, serves as a prominent intellectual counterpoint in the book. As the Chief AI Scientist at Meta, LeCun is one of the most influential figures in the field to publicly express skepticism about existential risk from AI. The authors engage with his arguments directly to illustrate what they consider to be an overly optimistic assessment of humanity’s ability to control advanced AI systems.

LeCun’s public claims that humans will be able to “engineer their desires” (183) and design AI systems to be “both superintelligent and submissive” (183) are critiqued as a form of folk theory that lacks rigorous engineering support. By dissecting his assurances, the authors stage a representative debate between the researchers who view advanced AI as manageable and those who believe current alignment methods remain inadequate. LeCun’s stature in the field makes his views a useful foil, allowing the authors to demonstrate why they believe confident public assurances are not a substitute for verifiable technical solutions to the alignment problem.

Thomas Midgley Jr., an American engineer and chemist from the early 20th century, is used as a powerful historical parable in the book. He is known for two major inventions with catastrophic environmental consequences: tetraethyl lead as a gasoline additive and chlorofluorocarbons (Freon) as refrigerants. The authors dedicate a chapter to his story to draw an analogy between past technological disasters and contemporary debates surrounding AI development.

Midgley’s career demonstrates how short-term performance gains can mask devastating long-term systemic harms. The introduction of leaded gasoline, despite known neurotoxicity risks, serves as a stark example of how industry incentives and plausible-sounding assurances can enable civilization-scale disasters. The authors use his legacy as a cautionary historical example to warn against repeating similar patterns of technological optimism with artificial intelligence. His story supports the book’s broader argument that societies often recognize the full consequences of powerful technologies only after widespread harm has already occurred.

Enrico Fermi, the Italian-American physicist who created the first self-sustaining nuclear chain reaction, is invoked to illustrate the concept of a “cursed problem.” His 1942 experiment, Chicago Pile-1, operated on a narrow margin between a controlled reaction and a potentially catastrophic failure. The authors use this event as a physics-based analogy for the AI alignment problem, arguing that some engineering domains are inherently unforgiving, with small errors capable of crossing critical thresholds with disastrous consequences. Fermi’s work demonstrates that controllability is often an artifact of carefully engineered constraints, a lesson the authors apply to argue that attempting to correct or contain a superintelligence after it has been created would likely be ineffective.

Vasily Arkhipov was a Soviet naval officer whose actions during the 1962 Cuban Missile Crisis helped avert a nuclear war. As a senior officer on submarine B-59, he refused to authorize the launch of a nuclear-armed torpedo when the vessel was cornered by US forces. The authors present him as a symbol of human judgment and deliberate restraint during a moment of extreme geopolitical tension. His story serves as a historical precedent demonstrating that coordinated de-escalation is possible, even when catastrophe appears imminent. For the authors, Arkhipov’s legacy as the man who “saved the world” models the kind of political restraint and international coordination they argue is necessary to slow or halt the development of advanced AI systems.