Yes, AI is involved. When OpenAI is configured, posts from authors without a good reputation get checked automatically: clean ones are approved, near-certain violations rejected, everything else waits for a human moderator.
Authors who are admins or have a good reputation…
It uses the existing OpenAI integration (the stack lists OpenAI for batch embeddings, commit relevance checks and moderation), but the knowledge base doesn't name a specific moderation model. So I don't know which one.