Not censorship as much as embedding up-to-the-minute "fact checks" into GPT language models so it can give a smug reply about misinformation when someone naively asks it about a trending topic. You can then feed this data into search results to manipulate news articles and then wikipedia further down the chain.

Marketing corporations are also constantly pulling data for semantic analysis. Some of those corporations could be intelligence agency franchises. Real-time is not super critical here, but it's still valuable.