As an exception to most regulations that we hear about from China, this approach actually seems well considered - something that might benefit people and work.

Similar regulations should be considered by other countries. Labeling generated content at the source, hopefully without the metadata being too extensive (this is where China might go off the handle) would help avoid at least two things:

• casual deception
• training AI with material generated by another AI, leading to degradation of ability to generate realistic content