After Diana Forsythe’s life was abruptly cut short in 1997, colleagues of the STS anthropologist posthumously published a collection of her ethnographic writings from medical informatics laboratories. (At that time, “medical informatics” was the reigning term for artificial intelligence applications in the health sector.) As a woman, a computer scientist, and a social scientist, Forsythe’s ethnography on the culture and production of artificial intelligence in health care remains a relevant source to guide how the law should govern the sector, particularly for AI’s potential to discriminate against women and people of color. Even over two decades on, artificial intelligence in health remains a field dominated by white men, while the population on whom such tools are used is increasingly diverse. Applications of AI in health can offer incredible scientific insights. But they also have been plagued by a reliance on unrepresentative and unreliable data sets (which fail to include women, people of color, and other marginalized communities, such as (in the U.S.) the uninsured; and which have historical inequities baked into them), a narrow focus on technical answers to health problems, and problem-solving approaches that ignore systemic inequality. This paper highlights how, in exposing the mechanisms of these processes, Forsythe’s and subsequent ethnographers’ work can shed light on regulating AI discrimination against women and intersecting identities in health care.