You need both. LLMs can, I think, do the bulk of removing posts that break community guidelines, but you need moderators to define and adjust the guidelines. Most would also like to have a human to escalate a dispute to.
Google is famous for having almost solely automated support, at it absolutely sucks at doing almost anything. AI only moderation would go the same way.
> but you need moderators to define and adjust the guidelines
The comments above you are suggesting that global guidelines are unnecessary. Instead, they suggest you don't need moderation at all when LLMs now give us the technology to filter out the stuff individual users don't want to see based own their own personal policies. I am sure you can come up with reasons to dispute that, but "you need moderators to do the thing you say is no longer necessary" doesn't add to the discussion.