The person who told the AI to create the music is the human that created it. This is how all music copyright works in practice, since artificial methods of gen

The person who told the AI to create the music is the human that created it. This is how all music copyright works in practice, since artificial methods of generating sound have been popular for over fifty years and humans who configured and used these artificial sound generators have received the same copyright protections as those who use traditional physical instruments. The question you're asking is: How complex does the sound/music generation need to get before copyright no longer applies? And I'd answer that there is no ceiling, whoever configured and activated the process of generating the artificial output owns the output. The output's complexity is irrelevant regardless of if it is a simple tone generator or a complex AI. There's no concept in law of Artificial Intelligence because it isn't distinguishable from existing methods of sound generation, just more complex version of the same. For example there's no self-aware or "alive" version of AI, and until there is I doubt we'll need any new laws to cover it.