In this presentation we advance a fully nonparametric model for the cumulative incidence functions in a competing risks framework by employing a species of nonparametric kernel regression within the context of an EM-based methodology. We relax the commonly invoked independence assumption between the failure times on the one hand and the observed censoring types and attendant masking sets, if any, on the other. The covariates created for use in the kernel regression will effectively act as the dependence mechanism, and will produce smoothed estimators of the competing risks failure probabilities. Our model is designed to be the most automated and data-driven approach possible when modeling competing risks in the presence of the aforementioned complicating factors of censoring, masking, and dependence, and will therefore be of particular interest to actuaries and statisticians who construct multiple-decrement life tables.