I do think this is the most reasonable compromise here, but to answer your question.
Why should artists not use generative models for their art process?
Simply, it takes away a lot of the creative part of the art process away. You might say that these models are just more efficiently doing what humans do, taking in source material, and reinterpreting it in a different context.
These models however are not creative, they can only ever recreate patterns from the training data. This is completely different from humans who can create wholly original ideas and patterns.
Putting this process in the middle of art creation will probably lead to even more stagnation and cultural death.