You might find that some providers simply disable comments, e.g. there's no particular reason that, say, the New York Times needs to support comments on articles for their business to remain viable.

Other sites, say, YouTube, that don't really exist without user content, might simply transfer liability to their users in their ToS during account setup. The net would be that YouTube continues its march to becoming, essentially, cableTV-like corporate media, where only professional orgs want to publish.

I'd be curious if sites like YT or even HN will feel the need to offer deletion of existing content, as a way of reducing their vulnerability surface? (Would that extend to github? People can be endlessly creative about expressing their opinion, see, eg, DeCSS) And if so, what impact would that have on training data?