Benzene, a common industrial chemical and a component of gasoline, is radiomimetic and exposure may lead progressively to aplastic anaemia, leukaemia, and multiple myeloma. Although benzene has been shown to cause many types of genetic damage, it has consistently been classified as a non-mutagen in the Ames test, possibly because of the inadequacy of the S9 microsomal activation system. The metabolism of benzene is complex, yielding glucuronide and sulphate conjugates of phenol, quinol, and catechol, L-phenylmercapturic acid, and muconaldehyde and trans, trans-muconic acid by ring scission. Quinol is oxidised to p-benzoquinone, which binds to vital cellular components or undergoes redox cycling to generate oxygen radicals; muconaldehyde, like p-benzoquinone, is toxic through depletion of intracellular glutathione. Exposure to benzene may also induce the microsomal mixed function oxidase, cytochrome P450 IIE1, which is probably responsible for the oxygenation of benzene, but also has a propensity to generate oxygen radicals. The radiomimetic nature of benzene and its ability to induce different sites of neoplasia indicate that formation of oxygen radicals is a major cause of benzene toxicity, which involves multiple mechanisms including synergism between arylating and glutathione-depleting reactive metabolites and oxygen radicals. The occupational exposure limit in the United Kingdom (MEL) and the United States (PEL) was 10 ppm based on the association of benzene exposure with aplastic anaemia, but recently was lowered to 5 ppm and 1 ppm respectively, reflecting a concern for the risk of neoplasia. The American Conference of Governmental Industrial Hygienists (ACGIH) has even more recently recommended that, as benzene is considered an A1 carcinogen, the threshold limit value (TLV) should be decreased to 0.1 ppm. Only one study in man, based on nine cases of benzene associated fatal neoplasia, has been considered suitable for risk assessment. Recent re-evaluation of these data indicated that past assessments may have overestimated the risk, and different authors have considered that lifetime exposure to benzene at 1 ppm would result in an excess of leukaemia deaths of 9.5 to 1.0 per 1000. Although in this study, deaths at low levels of benzene exposure were associated with multiple myeloma and a long latency period, instead of leukaemia, which might justify further lowering of the exposure limit, the risk assessment model has been found to be non-significant for response at low levels of exposure. The paucity of data for man, the complexity of the metabolic activation of benzene, the interactive and synergistic mechanisms of benzene toxicity and carcinogenicity, the different disease endpoints (aplastic anaemia, leukaemia, and multiple myeloma), and different individual susceptibilities, all indicate that in such a complex scenario, regulators should proceed with caution before making further changes to the exposure limit for this chemical.