The Core Argument: AI Companies as Modern Empires
Journalist Karen Hao, author of Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI, argues that the major AI companies are best understood through the lens of empire. She identifies four pillars: claiming resources not their own (data, intellectual property), exploiting labor (global contractor networks and automating away the very jobs they train on), monopolizing knowledge production (funding most AI research, suppressing inconvenient findings), and using existential mythology to justify anti-democratic consolidation of power.
The Mythology of AGI and Existential Risk
Hao traces the term 'artificial intelligence' back to 1956 and John McCarthy, noting there has never been scientific consensus on what human intelligence even is. She argues that this definitional vacuum allows companies like OpenAI to redefine 'AGI' for whatever audience they're addressing—cancer cure for Congress, revenue engine for Microsoft, digital assistant for consumers. The existential risk narrative (e.g., Dario Amodei's claim of a 10–25% chance of civilizational catastrophe) is described as a dual-purpose tool: it scares people into deferring to these companies as the only responsible stewards, while also justifying the exclusion of democratic participation in AI development.
Inside OpenAI: Altman, Musk, and the Power Struggle
Based on over 300 interviews including 90+ OpenAI insiders, Hao reconstructs the founding and early years of OpenAI. She details how Altman mirrored Elon Musk's language about existential AI risk to recruit him as a co-founder and major donor—and how Altman later maneuvered Greg Brockman and Ilya Sutskever to back him over Musk for CEO of the new for-profit entity. Musk's departure and subsequent legal battle stem from this. Altman's 2023 firing and rapid reinstatement is also reconstructed: Ilya Sutskever and Mira Murati brought documented concerns about Altman's chaotic management—pitting teams against each other, opacity around finances like the OpenAI Startup Fund (which was legally Altman's personal fund)—to the independent board members, who concluded the stakes of building AGI were too high to tolerate such leadership instability.
Labor, Jobs, and the Career Ladder Problem
Hao pushes back on binary narratives about AI and employment. She acknowledges real displacement—citing Klarna's reduction from ~6,000 to under 3,000 employees and a 40% decline in entry-level hiring in white-collar sectors—but stresses that executive decisions to downsize (sometimes using AI as convenient cover) are as responsible as model capabilities. Most troublingly, she describes a 'broken career ladder': entry-level and mid-tier roles are being eliminated, forcing displaced workers (including award-winning Hollywood directors) into low-paid data annotation work—training the very models that replaced them—with no clear pathway back up.
What Should Be Done
Hao argues the problem isn't the technology itself but the governance structure: a small group of unelected, ideologically homogenous billionaires are making decisions that affect billions of people worldwide with no democratic accountability. She calls for breaking up the AI empires and rebuilding an innovation ecosystem that serves the public interest. She is skeptical of autonomous vehicles reaching mass adoption soon, challenges the hypothesis that scaling statistical models equals general intelligence, and insists that research already shows AI tools work best as human-augmenting instruments rather than replacements—but that this path is less profitable and thus deprioritized.