Maybe their excuse will be that it can't be proven whether or not the deep learning model is discriminating based on this or that, it's just taking in as much data as possible?