Human exposure to perchlorate is commonplace because it is a contaminant of drinking water, certain foods, and breast milk. The U.S. Environmental Protection Agency (EPA) conducted a perchlorate risk assessment in 2002 that yielded a reference dose (RfD) based on both the animal and human toxicology data. This assessment has been superceded by a recent National Academy of Science (NAS) review that derived a perchlorate RfD that is 20-fold greater (less stringent) than that derived by the U.S. EPA in 2002. The NAS-derived RfD was put on the U.S. EPA's Integrated Risk Information System (IRIS) database very quickly and with no further public review. In this commentary we raise concerns about the NAS approach to RfD development in three areas of toxicity assessment: the dose that the NAS described as a no observable adverse-effect level is actually associated with perchlorate-induced effects; consideration of uncertainties was insufficient; and the NAS considered the inhibition of iodine uptake to be a nonadverse effect. We conclude that risk assessors should carefully evaluate whether the IRIS RfD is the most appropriate value for assessing perchlorate risk.