A systematic, tiered approach to assess the safety of engineered nanomaterials (ENMs) in foods is presented. The ENM is first compared to its non-nano form counterpart to determine if ENM-specific assessment is required. Of highest concern from a toxicological perspective are ENMs which have potential for systemic translocation, are insoluble or only partially soluble over time or are particulate and bio-persistent. Where ENM-specific assessment is triggered. Tier 1 screening considers the potential for translocation across biological barriers, cytotoxicity, generation of reactive oxygen species, inflammatory response, genotoxicity and general toxicity. In silica and in vitro studies, together with a sub-acute repeat-dose rodent study, could be considered for this phase. Tier 2 hazard characterisation is based on a sentinel 90-day rodent study with an extended range of endpoints, additional parameters being investigated case-by-case. Physicochemical characterisation should be performed in a range of food and biological matrices. A default assumption of 100% bioavailability of the ENM provides a 'worst case' exposure scenario, which could be refined as additional data become available. The safety testing strategy is considered applicable to variations in ENM size within the nanoscale and to new generations of ENM.