What kind of hate speech could you possibly generate with ChatGPT that doesn't already exist in the wild, free to copy and spread?
The sophisticated bad actor won’t generate straight-up hate speech that will just get filtered/blocked. They will be master of ten thousand bot accounts that work slowly to build up a plausibly innocuous posting history (including holding realistic conversations either between themselves or with real users), then start subtly manipulating conversations towards a predetermined political end.
Basically everything that political troll farms and media do already, but automated on a massive scale.
Where’s the automated defence against this?
Do you need the automated defense against it?
What people are up in arms about is that an "AI" saying it gives "legitimacy" to the views. Tai was just something that repeated what people said to it; ChatGPT does basically the same but more advanced, so if you ask it "how do you solve poverty" and it barfs out something horribly racist, people misinterpret that as being "supportive" scientifically somehow.