A generative model is trained on a dataset and then used to sample new synthetic training examples. What is a key risk of this approach?