Well, I'll give you one example. Piefed has tools designed to suppress new AI posters. I don't know if other instances use it - probably not, but Piefed.social does use this. Can it get it wrong and incorrectly alert us to users who are not using AI? Absolutely. Does this bother you? Legitimately curious.
I imagine every instance has the capacity to turn some cogs to censor content, but will choose not to. Sopuli for itself absolutely is blocking some instances though.