Abstract Background Early life exposure to inorganic arsenic may be related to adverse health effects in later life. However, there are few data on postnatal arsenic exposure via human milk. In this study, we aimed to determine arsenic levels in human milk and the correlation between arsenic in human milk and arsenic in mothers and infants urine. Methods Between March 2011 and March 2012, this prospective study identified a total of 120 new mother-baby pairs from Kashiani (subdistrict), Bangladesh. Of these, 30 mothers were randomly selected for human milk samples at 1, 6 and 9 months post-natally; the same mother baby pairs were selected for urine sampling at 1 and 6 months. Twelve urine samples from these 30 mother baby pairs were randomly selected for arsenic speciation. Results Arsenic concentration in human milk was low and non-normally distributed. The median arsenic concentration in human milk at all three time points remained at 0.5 μg/L. In the mixed model estimates, arsenic concentration in human milk was non-significantly reduced by -0.035 μg/L (95% CI: -0.09 to 0.02) between 1 and 6 months and between 6 and 9 months. With the progression of time, arsenic concentration in infant’s urine increased non-significantly by 0.13 μg/L (95% CI: -1.27 to 1.53). Arsenic in human milk at 1 and 6 months was not correlated with arsenic in the infant’s urine at the same time points (r = -0.13 at 1 month and r = -0.09 at 6 month). Arsenite (AsIII), arsenate (AsV), monomethyl arsonic acid (MMA), dimethyl arsinic acid (DMA) and arsenobetaine (AsB) were the constituents of total urinary arsenic; DMA was the predominant arsenic metabolite in infant urine. Conclusions We observed a low arsenic concentration in human milk. The concentration was lower than the World Health Organization’s maximum permissible limit (WHO Permissible Limit 15 μg/kg-bw/week). Our findings support the safety of breastfeeding even in arsenic contaminated areas.