It's a bit more than that. AI is writing new music in the style of an existing artist. What I haven't seen, which means I am not certain, is AI writing in its own style. That to me would be purely original and require multiple compositions with a consistent style.
As for sounding like a voice, that's the easy part. It is generating a voice, not altering a voice. That certainly can be done algorithmically, and probably isn't considered AI. Perhaps, if it get so far as generating emotion in a voice that would be considered AI.
To put a more political spin on things, AI can be used to generate "deep fakes", which includes video. If deep fakes get good enough they could generate video of footage of Trump committing an illegal act for instance.
What I haven’t seen, which means I am not certain, is AI writing in its own style.
= = =
Like some left-handed kid who learns to play guitar that is strung for a right-hander?