"A number of art communities have even banned AI generated content."

I can't say I'm surprised. It can be very spammy and can crowd out original content. A lot of it is just not really interesting, and how many pieces of it are there which look pretty samey (or generic / tacky)?

It's not a problem unique to diffusion models either. When someone used to feed like a million anime pics into StyleGAN to train a model, there was also an issue where it would be "cool geekery" but would otherwise look boring.