Well… since excising training data from an existing model is, at this point in time, computationally infeasible in addition to being unprovable… I’m gonna press X to doubt.