This article reviews the procedures for the assessment of safety/risk of chemicals to human health. Because the nature and severity of toxicity and the extent of the database vary from chemical to chemical, the assessment is done on a case by case basis. Essentially 5 steps are involved in the assessment: (a) identification of hazards based on appropriate human and animal data; (b) determination of the dose-response relationship of the adverse effects of the chemical; (c) extrapolation of the dose-response data from test subjects to human populations; (d) estimation of the exposure; and (e) assessment of the safety/risk of the chemical under a specified exposure. Emphasis in this article, however, is placed on the extrapolation of the dose-response data to the human situation. The extrapolation is done by the identification of a no-observed-adverse-effect level (NOAEL) and the application of a safety factor, thereby arriving at an acceptable daily intake (ADI). The safety factor is selected on the basis of, inter alia, the severity of the adverse effect and the adequacy of the database. On the other hand, with genotoxic carcinogens, mathematical modeling is used for extrapolation. This is because the effects of genotoxic carcinogens are generally believed to have no threshold. The ADI approach, which involves the identification of a NOAEL, is therefore not applicable. A number of mathematical models have been developed to assess, from the dose-response data, either the risks that may be associated with a specified dose, or the 'virtually safe dose' at a specified risk level. The evolution, application and shortcomings of these procedures and the potential improvements in the ADI approach and in the dose-response characterization based on these mathematical models are also discussed.