In many countries, government regulations require environmental risk assessment prior to commercial sale and widespread planting of transgenic crops. Here I evaluate the design and statistical rigor of experiments used by industry to assess the safety of transgenic plants for nontarget organisms, as required under U.S. regulations. This review reveals that a few simple improvements in experimental design could greatly increase the rigor and information content of studies required under current regulations. For example, although most experiments were conducted for 1–4 wk, some of the tested species can live a year or more and could experience much longer periods of exposure. Moreover, the number of replicates used in these studies was generally quite small (usually 2–6 replicates per treatment), resulting in experiments that had little chance of detecting real effects. Clearly, sample sizes should be bolstered, and nonsignificant results should be accompanied by an analysis of statistical power. In addition, information readily available over the Internet is insufficient for a quantitative assessment of a transgenic crop's safety. Improved access to information regarding the details of risk assessment studies could greatly increase the public's ability to evaluate industry's claims of safety.