Because the toxicity of arsenic depends on its chemical form, risk assessments of arsenic exposure must consider the type of arsenic compound, and hence they require sensitive and robust methods for their determination. Furthermore, the assessment should include studies on the most vulnerable people within a population, such as newborns and infants, and thus there is a need to quantify arsenic species in human milk. Herein we report a method for the determination of arsenic species at low concentrations in human milk by HPLC/ICPMS. Comparison of single and triple quadrupole mass analysers showed comparable performance, although the triple quadrupole instrument more efficiently overcame the problem of ArCl+ interference, from the natural chloride present in milk, without the need for gradient elution HPLC conditions. The method incorporates a protein precipitation step with trifluoroacetic acid followed by addition of dichloromethane or dibromomethane to remove the lipids. The aqueous phase was subjected to anion-exchange and cation-exchange/mixed mode chromatography with aqueous ammonium bicarbonate and pyridine buffer solutions as mobile phases, respectively. For method validation, a human milk sample was spiked with defined amounts of dimethylarsinate, arsenobetaine, and arsenate. The method showed good recoveries (99-103%) with detection limits (in milk) in the range of 10 ng As kg-1. The method was further tested by analyzing two Norwegian human milk samples where arsenobetaine, dimethylarsinate, and a currently unknown As species were found, but iAs was not detected.