Empire of AI

Hao, who covered artificial intelligence for MIT Technology Review and the Wall Street Journal, wrote a reported history of OpenAI and an argument about the industry it created.

The institutional narrative traces OpenAI from its 2015 founding as a non-profit committed to developing artificial general intelligence safely and openly, through the creation of the capped-profit subsidiary in 2019, the Microsoft partnership, the shift away from publishing research, the release of ChatGPT, the board's removal of Sam Altman in November 2023 and his return days later. Hao had reported on the organisation since 2019 and had access to internal material and to current and former staff.

The argument in the title is that the leading AI companies operate on an extractive model comparable to earlier imperial enterprises: they take data produced by others without compensation, consume large quantities of energy and water in locations that bear the environmental cost, and rely on low-paid contract workers in Kenya, the Philippines and elsewhere to label data and to filter traumatic content. She reports on these workers directly. She also argues that the scaling approach, which pursues capability through ever-larger models and compute, was a choice rather than a technical necessity, and that it concentrates the field in the small number of organisations that can afford it.

Hao's reporting on labour conditions and on OpenAI's internal history is detailed and largely documented. The imperial framing is an interpretation and readers who reject it can still use the reporting. The book is critical throughout and does not present the strongest version of the case that scaling has produced genuine capability gains or that the resulting systems create value that offsets the costs she catalogues. Events in this sector also move faster than publication schedules.