There should be a users/penetration threshold above which enforcing a cultural code of conduct waives immunity. GDPR uses a similar threshold, and unless you're seriously arguing that "trolling", "flaming", and "swearing" are being moderated successfully on e.g. Twitter, it seems practical.