The National Institute of Health’s (NIH) responsible conduct of research (RCR) training requirement focuses on avoiding fraud, ensuring data transparency, avoiding financial conflicts of interest, and equal treatment of peers and mentees regardless of race, gender, or professional status. This view of scientist’s ethical duties is often justified by appeal to an “idealized” scientific practice, that is: (i) co-operative, (ii) universal, (iii) disinterested, and (iv) skeptical. Yet actual scientific practice is necessarily laden with non-epistemic values, biased towards gendered and racialized hypotheses, inextricably linked to money and professionalization, and prone to confirmation bias. In this paper, I introduce the tools of “non-ideal theory”, developed within political philosophy, to the dialogue over RCR policy and education in the health sciences. I argue that by failing to attend to the actual practice of science, the current understanding of RCR may inadvertently undermine the epistemic and ethical aims of science. Instead, RCR rules and education should place greater emphasis on the determinants of non-ideal justice, including: (i) robust differences in basic capacities and needs amongst citizens, (ii) presence of historical and ongoing social hierarchies supported by the products of science, and (iii) the presence of bad actors who may misuse the products of science. By ignoring these aspects of our society, a theory of just science may inadvertently act as an ideology which entrenches existing injustice.