When mothers of preterm infants are unable to produce sufficient volumes of breastmilk, neonatologists in many Western countries prescribe pasteurized donor breastmilk. Breastmilk has a paradoxical presence in the neonatal intensive care unit while it has therapeutic properties, it also has the potential to transmit disease. National health authorities and local neonatal intensive care unit policies each delimit the safety of donor milk by focusing on the presence or absence of pathogens. It is in this light that breastmilk from the human milk bank is both sought and legitimated to minimize safety concerns. This research uses data arising from an ethnographic study of two human milk banks and two neonatal intensive care units in the United States, and 73 interviews with milk donors, neonatal intensive care unit parents and clinicians. The primary research question framing the study was 'What are the underlying processes and practices that have enabled donor milk to be endorsed as a safe and legitimate feeding option in neonatal intensive care units?' This study is framed using three key principles of Latour's 'new critique', namely, adding to reality rather than debunking it, getting closer to data rather than turning away from fact and creating arenas in which to assemble. As a result, conceptions of donor milk's safety are expanded. This case study of donor milk demonstrates how Latour's new critique can inform science and technology studies approaches to the study of safety in health care.