Arsenic contamination of groundwater occurs due to both geogenic and anthropogenic processes. Conventional arsenic remediation techniques require extraction of groundwater into pump-and-treat systems, which are expensive and require long operational times. Hence, there is a need for cost-effective remediation. In this study, we assessed and validated the in situ remediation of arsenic contamination in groundwater resources using permeable reactive barriers (PRBs) made of injectable, colloidal iron oxide nanoparticles in the laboratory and in field-scale pilot tests. Sand-packed, flow-through column studies were used in order to assess the sorption behavior of the iron oxide nanoparticles using field materials (sand, groundwater) in the laboratory. The breakthrough curves were analyzed using a reactive transport model considering linear and nonlinear adsorption isotherms and were fitted best with a chemical nonequilibrium consideration. The results were used to design a pilot-scale field test. The injected 28 m³ of nanoparticles (ca. 280 kg dry weight of iron oxide) were successfully delivered to the aquifer via an injection well. No mobile iron was detected downstream, confirming that a stable in situ barrier was formed that did not move with the groundwater flow. Arsenic concentrations in groundwater were reduced to the aimed 50% of the background value, despite the relatively short contact time between arsenic and the iron oxide in the barrier, due to the high flow velocity of 1.21 m/day. We compared the results of the laboratory and field tests and concluded that the single-parameter models based on retardation factor and/or adsorption capacity fail to predict the longevity of the barrier and the evolution of arsenic breakthrough with time, most likely because they do not consider the chemical nonequilibrium effects. Therefore, we propose that upscaling the laboratory findings to field design must be carried out with care and be coupled with detailed reactive transport models.