Karen Hao describes OpenAI serving papers on a small watchdog nonprofit leader during its nonprofit-to-for-profit conversion. The demand sought every communication involving Elon Musk, reflecting what she calls paranoia that Musk was secretly funding opponents; according to her, none of them were actually funded by him.
Karen Hao says Sam Altman and Dario Amodei pair worst-case catastrophe with best-case abundance. She argues this two-sided mythology lets AI companies claim only insiders can safely control the technology at every step, justifying an anti-democratic approach that shuts out public participation and democratic oversight.
Karen Hao compares AI companies to fossil fuel firms bankrolling climate science, noting they fund most AI researchers and thereby set research priorities. She cites Google firing Timnit Gebru and Margaret Mitchell after they co-wrote a paper on harmful large-language-model outcomes, in an attempt to stop its publication.
Karen Hao says AI companies dangle interviews and access to steer technology journalists away from critics and toward friendlier coverage. After OpenAI shut her out early, she says the closed front door forced her to report objectively and eventually built her professional independence from access.
Karen Hao says Ilya Sutskever tried to remove Sam Altman because he believed Altman undermined OpenAI's twin goals: reaching AGI and doing it safely. She describes a chaotic company where Altman pitted teams against each other and told different people different things, according to her interviews.
Karen Hao traces Ilya Sutskever and Geoffrey Hinton's belief that brains are statistical engines to an unproven hypothesis, not settled science. She says this assumption justifies building ever-larger models and data centers, with workers and resources pulled in to pursue AGI, creating real global consequences.
Karen Hao calls AI companies empires because they claim data and intellectual property, land-grab data centers, exploit contract workers, and monopolize knowledge production. She says these practices let them operate at imperial scale while treating the public and policy understanding as theirs to directly control.
