There's an issue of AI art itself not having copyright protection regardless of the source material used for training, at least in the USA, due to a lack of human involvement in the final result - even photographers have to aim and frame, as well as make various adjustments based on environment. All an AI user does is act as a client making a request, and the generator does all the real lifting. The US Copyright Office has currently been going with this policy, and other countries will likely follow suit.
It's still technically the wild west in courts for the most part, but big companies will likely stay away from AI art in their commercial projects except maybe as use in concept work when planning out a project.
The copyright protection technicality on that isn't that broad. It's if the entire work is AI made. If the art is used in, say, a Disney movie, that becomes transformative enough that the movie itself still has copyright protection.
Of course, then that would allow people to use the picture within their work to create other works which... I mean fanfiction, fangames, and fanart are all already a thing. Yes, making a profit off these types of things can be murky depending on the context, but it is done. I don't actually see big companies shying away from using AI art in the long term, personally, just because of that drawback.
Heck, there are news orgs (Mostly tech related) with AI generated articles and cover art for those articles already that had some controversy about that because of the AI being confidently wrong a few times.
Also, while I wouldn't call generating AI art skilled or creative work, I wouldn't agree with the characterization that it's effortless. Adjusting prompting, getting the right settings and continuously generating takes time and effort and even then it will often require editing to get something reasonably decent. It's not just "push this button" and you get more or less what you want.
Of course, ultimately I doubt the technology will stay in a state where it requires that much effort in the long term, either, and I do agree that purely/mostly AI created works should not be able to be copyrighted.
But:
because AI art generators won't necessarily veer far away from someone else's copyrighted designs to produce results that can be considered non-derivative or fair use, and there's no guarantee that AI art will be legal to use in commercial products in the future.
While it's murky it's not
that murky. The cases where AI reproduce something close to an original artist's work are very rare unless the model's been trained to do exactly that, and in the cases where it's not it's normally some very, very precise prompting.
The only thing I think that's going to be up for debate in the courts is how models are generated.