Hazard characterisation of low molecular weight chemicals in food and diet generally use a no-observed-adverse-effect level (NOAEL) or a benchmark dose as the starting point. For hazards that are considered not to have thresholds for their mode of action, low-dose extrapolation and other modelling approaches may be applied. The default position is that rodents are good models for humans. However, some chemicals cause species-specific toxicity syndromes. Information on quantitative species differences is used to modify the default uncertainty factors applied to extrapolate from experimental animals to humans. A central theme for extrapolation is unravelling the mode of action for the critical effects observed. Food can be considered as an extremely complex and variable chemical mixture. Interactions among low molecular weight chemicals are expected to be rare given that the exposure levels generally are far below their NOAELs. Hazard characterisation of micronutrients must consider that adverse effects may arise from intakes that are too low (deficiency) as well as too high (toxicity). Interactions between different nutrients may complicate such hazard characterisations. The principle of substantial equivalence can be applied to guide the hazard identification and hazard characterisation of macronutrients and whole foods. Macronutrients and whole foods must be evaluated on a case-by-case basis and cannot follow a routine assessment protocol.