Empire of AI


Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI is a book by Karen Hao released on May 20, 2025. It focuses on the history of OpenAI and its culture of secrecy and devotion to the promise of artificial general intelligence.
The book includes interviews with around 260 people, correspondence, and relevant documents. The title makes reference to colonial empires of the 1800s.

Origins

Hao visited OpenAI's offices and covered the company for the MIT Technology Review two years before ChatGPT was released. Her experience there and reporting on topics of AI and "AI colonialism" for seven years led her to develop the book's thesis and title, Empire of AI, and ultimately became published as a book''.''

Contents

  • Prologue: A Run for the Throne
  • Part I
  1. Divine Right
  2. A Civilizing Mission
  3. Nerve Center
  4. Dreams of Modernity
  5. Scale of Ambition
  1. Ascension
  2. Science in Captivity
  3. Dawn of Commerce
  4. Disaster Capitalism
  1. Gods and Demons
  2. Apex
  3. Plundered Earth
  4. The Two Prophets
  5. Deliverance
  • Part IV
  1. The Gambit
  2. Cloak-and-Dagger
  3. Reckoning
  4. A Formula for Empire

Reception

This book is one of two written about OpenAI that were released around the same time, the other being Keach Hagey's The Optimist: Sam Altman, OpenAI, and the Race to Invent the Future. Both Hao and Hagey rely on similar sources to write their respective books. Hao's book is noted to be "broader and more critical of the two" and "darker", and "dispels any doubt that OpenAI’s belief in ushering in AGI to benefit all of humanity had messianic undertones". Some have compared Hao's book to historian William Dalrymple's The Anarchy: The Relentless Rise of the East India Company, which recounts the rise of another corporate empire, The East India Company.
OpenAI declined to cooperate with Hao on her book and CEO Sam Altman has publicly criticised Hao's book on social media.

Factual Errors and Retraction

In November 2025, after by Andy Masley and others, Karen Hao that the published version of the book mistakenly overestimates the water usage of a datacenter in Chile by a factor of 1,000x due to a unit conversion error. The discovery of this error was covered in media outlets such as Wired.