I'm specifically talking about that part being put through an AI voice model (RVC), which is VERY different from saying it was generated with AI which is what I'm assuming you got out of my message, I don't blame you though, there's a million different AI technologies nowadays so my bad for not specifying what I meant.
I'm gonna make this as explicitly clear as I possibly could:
When an acapella with multiple harmonies in the background (or in simpler terms, Background vocals.) gets processed through an RVC model (an AI Voice Model, you know, the thing that was all the craze before generative AI became a thing) the model messes up it's detection, since there's multiple voices going on at the same time, it tries to imitate the melody of every line in the original acapella at the same time, which it obviously can't.
The most obvious giveaway is the fact that the actual background vocals randomly come in after that part is over
THAT'S what I think is going on with THAT specific part of the song... it's not that impossible or unbelievable, especially because it's been done in the past in the exact same manner. A good example is how before Men In Black leaked completely, snippets were being passed around which were just like I described here, the original song with the acapella being put through AI.