309
Why This Award-Winning Piece of AI Art Can’t Be Copyrighted
(www.wired.com)
This is a most excellent place for technology news and articles.
You can already do that with over fitting or getting lucky with duplicates in the dataset(<6*10^-5% chance for stable diffusion). The resulting work is still protected under the original creator's copyright because it's almost identical. Same way that if you photocopy an image it doesn't remove the copyright even though slight details will be off due to inaccuracies.