Empire of AI by Karen Hao

Published in May 2025, Empire of AI is a masterful piece of investigative reporting by Karen Hao, a journalist who has been following the AI tech scene in the US for the last 15 years or so, with a special focus on OpenAI. The subtitle in my edition gives away the thesis defended in the book: Inside the reckless race for total domination.

TL;DR

I fully recommend this book. It is an example of rigorous investigative journalism. It documents several dimensions of the development of generative AI:

• the exploitation of digital labor workers in Venezuela and Kenya, who help post-train AI models by labeling offensive / disturbing / frankly psychologically damaging content
• the exploitation of communities in the Global South for their resources in fresh (potable) water to cool data centers
• the birth of OpenAI, the creation of GPT, and the split that led to the creation of Anthropic
• the character of Sam Altman

The thesis

The thesis defended by Karen Hao is that Sam Altman is building an empire with OpenAI in a ruthless way: extractive in nature, leading to a massive concentration of a new kind of wealth, and driven by an unlimited thirst for power.

To give some perspective, Hao draws on Daron Acemoglu and Simon Johnson’s Power and Progress (PublicAffairs, 2023): generative AI is one of those fundamental technologies (like the cotton gin) that can increase productivity to new levels and create immense new wealth in a given industry, but at tremendous social, political, and environmental costs. In the case of the cotton gin, it expanded and entrenched slavery-based cotton production in the US.

Through meticulous, decade-long reporting, Karen Hao documents the multiple dimensions of this empire-building activity. There are 40 pages of detailed endnotes that provide sources for all the assertions she makes in the book, including interviews with hundreds of actors who played a role in the development of OpenAI or were impacted by it.

I am not convinced by…

Karen Hao chooses to include and document Annie Altman’s life story, in an effort to fully document Sam Altman’s personality and worldview. No doubt it was a long-matured decision, and there are emerging directions in journalism that can justify this approach.

I was reminded, for example, of Elon Musk’s conflictual relationship with his father. At Elon Musk’s request, his first biographer chose not to address this issue; in return, the author gained elevated access to Musk.

His second biographer chose the opposite. Delving into* Musk’s relationship with his father became a central key to the psyche of his subject, and my impression is that the second biography provides deeper, more valuable insights into Musk’s trajectory.

* NB : yes I am in control of this sentence and this particular verb

But in the case of Sam Altman’s relationship with his sister, my personal impression is that it does not provide a missing key to his psyche, decisions, or trajectory. In consequence, it left me ill at ease. Discussing how individuals live and cope with close relatives who suffer mental health issues and then making broad, fragile inferences from it: it is a tough call.

Second, and more fundamentally, I am left uncertain by the overall framing. A critical viewpoint is necessary, and this 482-page critical history of OpenAI and generative AI is absolutely worth reading. “This book is not a corporate book,” the author states clearly in a preliminary note. But as it stands, the story could create the impression that generative AI only causes harm. Its benefits are left unmentioned, beyond mandatory and cursory nods.

I suppose there is a legitimate division of labor in reporting: one book is not supposed to be a kaleidoscope of all viewpoints on a given subject. Still, the imbalance is noticeable. I encourage everyone to read Ethan Mollick’s Co-Intelligence (2024) alongside Karen Hao’s book to develop that balance.

Last impression

The creation and evolution of OpenAI; Sam Altman, Greg Brockman, Ilya Sutskever, and Mira Murati and their relationships; the watershed moment of ChatGPT on November 30, 2022; the ousting and reinstatement of Sam Altman; the split that led some OpenAI employees to create Anthropic; the electricity and water costs of new data centers built to train AI models; the theft of copyrighted material through unfiltered, blanket scraping of the web; and the exploitation of workers to fine-tune these models: this book provides fact-checked insider detail on all these crucial issues, with rigor and strong independence. For all these reasons, I strongly recommend reading it.

About Me

I’m an academic and independent web app developer. I created nocode functions, a free, point-and-click tool for exploring texts and networks. It’s fully open source. Try it out and let me know what you think. I’d love your feedback!

• Email: analysis@exploreyourdata.com
• Bluesky: @seinecle
• Blog: Read more articles on app development and data exploration.