This is based on this proposition and WGAN-GP paper has a proof on that. This is a property of the Lipschitz function. So the paper uses it to compute the penalty. I have not come up a non-mathematical way to explain it so if you want to know more, you may have to read the proof directly.
