Malmberget is located in the northern Norrbotten ore province, Sweden, and hosts an iron oxide–apatite deposit of Kiruna type. The area is actively mined by the state-owned LKAB company, making it an electromagnetically noisy environment. New broadband magnetotelluric data have been collected and analysed, and the 3D electrical conductivity structure has been developed to improve the understanding of, and connection between, the local deposit and more regional structures. To mitigate the influence of noisy data on the inversion results, Huber’s loss function has been introduced and implemented based on an iteratively reweighted least squares scheme. The resulting robust 3D model shares major features such as the mine location and regional anomaly model acquired by standard L2 optimisation but shows differences on a local scale. Sites which received lower weights are generally identified as potentially noisy measurements due to the proximity to sources of anthropogenic noise, emphasising the ability of the scheme to quantitatively identify and deal with measurements contaminated by noise and improving overall data fit. Three main geological conclusions are gained from the 3D resistivity model. Firstly, a large east–west striking conductive anomaly at ~3−4 km depth is present in a large part of the model area. Secondly, three narrow, vertical, conductive structures connect the known mineralisation to the deep conductive anomaly. Thirdly, a second, deep conductive anomaly striking east–west in the south connects to the surface close to the Dundret ultramafic intrusion. Conductivity anomalies in the area are hypothesised to originate from a combination of magnetite mineralisation when close to the surface and iron-rich mafic rocks at depth, with possible presence of sulphides and/or graphite.