Accurate assessment of fault-related CO2 migration hazard is required to deploy geologic carbon storage at the gigaton scale. First, we present a novel methodology, PREDICT, to model the intrinsic permeability of faults in siliciclastic sequences. PREDICT models realizations of the fault core consistent with the stratigraphy, and computes the probability distributions for the directional components (dip-normal, strike-parallel and dip-parallel) of the fault-scale permeability tensor. PREDICT accounts for uncertainty in the geologic variables influencing fault permeability and was developed for scenario building and risk management. Second, we show how to leverage PREDICT to build geologically-realistic fault leakage scenarios using a model of the Miocene offshore Texas, Gulf of Mexico. The process includes selection of anisotropic, upscaled fault permeability values from PREDICT’s output, upscaling of multiphase-flow fault properties (relative permeability and capillary pressure), and CO2-brine numerical simulation for hundreds of years. CO2 migration through the fault and into overlying units is tracked in each scenario, and results are compared with SGR-based fault property modeling.