Karen Hao · Bilingual online reading edition
Dreams and Nightmares in Sam Altman’s OpenAI.
An investigative account of how ideals became power—and technology became empire.
An award-winning journalist covering AI’s impact on society. Formerly a Wall Street Journal reporter and MIT Technology Review senior AI editor, she writes for The Atlantic and leads the Pulitzer Center’s AI Spotlight Series.
Maintains the isomorphic English and Simplified Chinese Markdown editions, including translation, chapter navigation, full-text search, whole-book print, and machine-readable outputs.
Author’s note
OpenAI’s inside story is a prism through which to see far beyond this one company.
— Karen Hao, Author’s Note to Empire of AI
The book draws on more than 300 interviews with roughly 260 people, alongside correspondence, documents, Slack messages, and reporting from communities around the world. More than 150 interviews involved over 90 current or former OpenAI executives and employees; another group covered roughly 40 people from Microsoft, Anthropic, Meta, Google, DeepMind, and Scale.
It records the internal history of a research lab, but asks a larger question: how scientific ambition became an ideological, money-fueled quest; how data, energy, land, and labor were organized beneath the models; and who ultimately gained the power to define the future.
The result is not an isolated judgment of one company. It is a material and political anatomy of the artificial-intelligence industry.
Chapter contents
Eighteen chapters follow OpenAI’s path to power: idealism shapes the institution, capital and scale rewrite the mission, global resources sustain expansion, and a governance crisis reveals the imperial logic underneath.
Copyright and use
Original copyright, the web edition, and the repository license are separate layers. Open access does not alter the claims of the author or publisher.
The English work is copyright Karen Hao and published by Penguin Press / Penguin Random House. This site is for learning, search, and research; please support the author and publisher through an authorized edition.
The copyright page states that no part of the book may be used or reproduced to train artificial-intelligence technologies or systems. This site’s accessibility does not waive that restriction.
The repository uses Creative Commons Attribution 4.0 International only for code, organization, and contributions their authors may license. It does not replace or expand rights to the English work.