We investigate the relationship between armed groups and large-scale mining firms in the Democratic Republic of Congo using geo-referenced data over 2000-2015. We start by showing that the pattern of links between armed bands and concession owners significantly departs from the random benchmark. To understand these patterns, we develop a statistical detection algorithm which flags owner-band dyads based on repeated interaction and anomalous bands’ movements. Results indicate that flagged dyads are significantly more likely to be observed together, also in very far away mineral concessions. We next explore the nature of these contacts. Our results are consistent with the interpretation that mining companies and armed bands engage in repeated interactions, where the latter help clear the territory from competing armed bands and destabilize the surrounding environment in a way that potentially allows to access cheaper labor. To conclude, we show an increase in mining production after the start of these relationships.