Fundamentally I think the gold standard should be that anything generated by ai is public domain and that dmca anything that using copyrighted work.

But at the end of the day, the issue at hand is really the way copyright works.

Since so far the "copyright" we got at the moment is mostly in practice only ever a right for the middleman (publisher, record companies, etc. ) which are important to the industry just not remotely as beneficial to the actual creator(s).

If there was a clear cut licensing system (license to stream, license to use once, lifetime license and ai training license) it would solve potentially the ai issue, since some of the profit would go to the creator.

But since copyright is based on who owns it (based on usually capital purchase) like a monopoly, instead of an inherent creator(s) , it's at odds with the creator(s) and automation.