Over recent years, human pharmaceuticals have been detected in the aquatic environment. This, combined with the fact that many are (by design) biologically active compounds, has raised concern about potential impacts in wildlife species. This concern was realized with two high-profile cases of unforeseen environmental impact (i.e., estrogens and diclofenac), which have led to a flurry of work addressing how best to predict such effects in the future. One area in which considerable research effort has been made, partially in response to regulatory requirements, has been on the potential use of preclinical and clinical pharmacological and toxicological data (generated during drug development from nonhuman mammals and humans) to predict possible effects in nontarget, environmentally relevant species: so-called read across. This approach is strengthened by the fact that many physiological systems are conserved between mammals and certain environmentally relevant species. Consequently, knowledge of how a pharmaceutical works (the “mode-of-action,” or MoA) in nonclinical species and humans could assist in the selection of appropriate test species, study designs, and endpoints, in an approach referred to as “intelligent testing.” Here we outline the data available from the human drug development process and suggest how this might be used to design a testing strategy best suited to the specific characteristics of the drug in question. In addition, we review published data that support this type of approach, discuss the potential pitfalls associated with read across, and identify knowledge gaps that require filling to ensure accuracy in the extrapolation of data from preclinical and clinical studies, for use in the environmental risk assessment of human pharmaceuticals.