I’d guess that it is doing voice to text, then standard automatic moderation on the text, rather than a new AI that understand hateful sounds. Just a guess though. At this point, you could run the voice to text on local machines and pass that off to the server. Of course that means modders could disable the protection, but the vast majority of users wouldn’t be able to do that. It would also give the added benefit of transcriptions for players that can’t hear voice chat.