I find it difficult to create a censorship model that can distinguish between "nuanced counterfactual take" and "contrarian falsehood". I think a lot of people already know this is a problem (e.g. censorship of the Wuhan lab theory).

I am fairly biased though -- I would rather 100 dangerous falsehoods get shared (even if it results in a lot of people believing wrong things) than a even 1 true fact get censored and that puts me in the weird position of often defending people and takes I disagree with and dislike.