In this excellent Guardian post ‘It’s the screams of the damned!’ The eerie AI world of deepfake music‘, the implications of modern AI in the field of music are detailed. It seems quite possible nowadays to imitate the voice of any singer including those of the past. And copyright laws won’t generally apply.

This implies as well of course that it is quite possible to imitate perfectly the voice of anybody, which creates at once a potential for fakeness for all sorts of voice recordings nowadays.
It is also the start of an era where the position of the artist will change, with an onus for song- and musicwriters and potentially less for singers, unless they manage to unleash emotional impact. Actually as the economy of music is moving progressively towards live performance, recordings become more and more of a commodity and artists will need to be better performing live.
An interesting point is that copyright law is different in each country. It appears that in the US there is a law “against impersonating famous people for commercial purposes” but not in the UK or in many countries. We can expect some substantial law changes to protect the unwary, artists and their heirs against usage of its voice at least for commercial purpose. Another situation where the law will have to catchup soon!
In any case AI is also upending a large part of the music industry like it is upending many other industries, and that is made easier because at the end music is only a succession of bits, which can be taught to AI. Welcome to a new world of music!