Geologic carbon storage (GCS) is one of the few climate-change mitigation technologies that can be deployed at the scale necessary to significantly offset current anthropogenic CO2 emissions, especially in combination with so-called negative emission technologies1. Much work over the past two decades has helped improve our understanding of the physics of CO2 migration and trapping, and storage capacity of geologic formations2; however, a critical question that remains to be better understood is the potential for CO2 leakage into shallower formations by migration along geologic structures like fractures and faults. The impact of faults on subsurface fluid flow is controlled by (1) their internal structure, which can be highly complex; and (2) the flow properties of the altered rock, which can be highly heterogeneous. Failure to either accurately characterize or account for these factors limits predictions of CO2 migration within fault zones, and needs to be addressed if large-scale GCS is to be deployed3-6. Here, we address this question by developing a 3D computational model representative of the geology in offshore Texas waters at the scale of tens of kilometers7. We employed the computer code MRST8 to simulate CO2 injection into a saline aquifer, and recorded fluid pressures and CO2 saturations within the injection layer, along the fault zone, and in overlaying formations. Of particular importance is that we simulate CO2 injection into a poorly-lithified Miocene sand, typical of “soft” and permeable sedimentary formations that offer excellent potential for storing CO2 at scale4,7. We modeled the fault architecture using discrete cells along different architectural elements. Given the large uncertainty associated with the petrophysical properties of the faults, we considered a wide range of values that account for the stratigraphic sequence and process-based fault fabric in this type of sediments9. This provides a rigorous testing ground to conduct a sensitivity analysis of CO2 along-fault leakage as a function of the various subsurface flow parameters, and to identify—quantitatively—the critical geologic features that will guide site selection and reduce leakage hazard.