There is a difference between AI voice (neural network) and a voice changer (that uses your voice sound waves but just changes pitch and stuff) or an honest voice synthesizer (that sounds like a crappy voice synthesizer). Disallowing the latter two would make no sense as these are among standard technologies. But, if the produced voice is of high quality and contains personality traits that were not present in the original recording then you can be sure that a neural network was involved and so it is not allowed.