Testing whether a Chinese deep learning model is censored is like testing if water is wet.