Personally I think Curt Yarvin had a decent idea about how to handle this type of thing with Urbit. Of course Urbit itself is terminally esoteric and nigh unusable, but here are the principles:
“Urbit will do everything possible to help you block content you don't want to see. It does nothing to help you block content you don't want others to see.
...
If you post content on Urbit which may offend or disturb any significant group of users, it is your responsibility to mark it properly for automatic blocking.”
https://urbit.org/posts/beliefs-and-principles/
There was a wordier post related to this but I think some stuff might have been purged with his separation from Tlon.
Anyway, the point is, if I’m writing from an alt-right perspective it’s my responsibility to tag it as such. Properly tagged, anti-semitism (for example) would not be an abuse of the platform. I don’t see how this is any less enforceable than the outright banning that e.g. Twitter practices now, and in fact it may be quite a bit easier, as you’ll get fewer “outlaws” who have to cycle through alts every week or two to get their message out to their followers.
One problem of course is establishing a small and universally recognized set of tags that work well enough for most people. Instead of very specific “content warning” types of things, it’s probably good enough to indicate broad political allegiance - black flag, red flag, brown flag etc. It’s rare that just a single word in isolation sets people off - offense taken seems to have more to do with the worldview motivating a given statement.
Then there’s the question of asymmetry. Say, for example, that As are mortally offended by B content but Bs don’t seem to care much about A content. In effect, you’d be creating an unfair burden on Bs simply because they’re more tolerant than As.
NLP algorithms can obviously help here, as well as crowdsourced tagging. So a single A user might have to be exposed to a B-type post to apply the tag, but we might consider that an acceptable loss over a large enough base.
This of course all assumes good faith. I suspect that As will simply be furious that Bs are able to talk to other Bs and neutral Cs, even if they never see those posts.