Censorship that many people, myself included, agree with (I don't believe all censorship is harmful) but it is clearly censorship.

But the real problem is assuming everything an AI blocks will be an "ad." Why is this the only instance where the slippery slope doesn't apply? This is different than a normal ad blocker that uses specific heuristics to block specific elements - an AI will be employed to make "intelligent" editorial decisions based on interpreting the intent of the content. What happens when it decides - or is ordered to decide - that political advocacy or the promotion of certain ideas or contrary narratives amounts to "advertising" them?

It's literally just enforcing wrongthink, most likely using closed source centralized services run by corporations who will be harvesting and selling your data to advertisers anyway. But it's cool because ads. It's not cool when it's Nazis or doxxing transgender people or spreading conspiracy theories or plotting sedition, but it's definitely cool if Justin Whang wants to sell me on Nord VPN.