Interesting, so the implication is that web companies are generating advertising from user-submitted content, therefore it's been read by the company, so therefore any cross-user communication needs to be reviewed by a person or a sufficiently accurate AI for endorsement through publication.

So in your framework, transmission of user content + advertising based on that content = culpability. I wonder if social media companies would figure out a way to legally prove transmission of a message without viewing its contents as a way to avoid culpability and maintain some level of profitability.

This is like suggesting that if you build a spam filter, you need to build a filter that scans content and catches and reports every type of crime possible to commit ever.