[nine_k]

I think the same approach as with spam is best.

If you can determine that a particular piece of information belongs to some unpleasant or dangerous category, mark it as that, like a spam filter does, or vaguely how FB does with posts. Importantly, do not remove it unless the law forces you to.

Then let users switch between the filtered and unfiltered view, or look at the analog of the "Spam" folder.

This allows to study pieces of information that are deemed "controversial" or even "malevolent" and make your own opinion, if you're so inclined.

Even more, it could apply several different cultural filters, like "content likely offensive for X", where X is a major religious or cultural group, much the same way as many providers currently mark content as inappropriate for children.

[silentsea90]

Twitter does add warnings to tweets that are questionable in their claims (anti vax, fake news etc). Is that what you were thinking?

[smhost]

and then there will be a million "Top 10 things the government doesn't want you to know" videos that amplify whatever it was they were trying to censor

[nine_k]

But the point is in not trying to censor.