Compliance with the established exposure limits for the electric field (E-field) induced in the human brain due to low-frequency magnetic field (B-field) induction is demonstrated by numerical dosimetry. The objective of this study is to investigate the dependency of dosimetric compliance assessments on the applied methodology and segmentations. The dependency of the discretization uncertainty (i.e., staircasing and field singularity) on the spatially averaged peak E-field values is first determined using canonical and anatomical models. Because spatial averaging with a grid size of 0.5 mm or smaller sufficiently reduces the impact of artifacts regardless of tissue size, it is a superior approach to other proposed methods such as the 99th percentile or smearing of conductivity contrast. Through a canonical model, it is demonstrated that under the same uniform B-field exposure condition, the peak spatially averaged E-fields in a heterogeneous model can be significantly underestimated by a homogeneous model. The frequency scaling technique is found to introduce substantial error if the relative change in tissue conductivity is significant in the investigated frequency range. Lastly, the peak induced E-fields in the brain tissues of five high-resolution anatomically realistic models exposed to a uniform B-field at ICNIRP and IEEE reference levels in the frequency range of 10 Hz to 100 kHz show that the reference levels are not always compliant with the basic restrictions. Based on the results of this study, a revision is recommended for the guidelines/standards to achieve technically sound exposure limits that can be applied without ambiguity. Bioelectromagnetics 34:375–384, 2013. © 2012 Wiley Periodicals, Inc.