first, just to clarify, I am not saying all moderation should be automatic. That is what I said in my first point, but in my second point, moderation is still manual and delegated to another person. The only difference is that you can very easily opt-out of it without losing anything else, or you can override it.

so, instead of moderation being something tighly coupled with a community or space where people post, it is instead something separate. You can “subscribe” to a moderation policy managed by someone or a group of people, and anything they ban (automatically or manually) applies to you without extra effort. The benefit to this is that if you ever regret this “subscription”, you dont lose out on the entire community. You can simply just change the moderation policy.

To answer your other points:

• legal concerns: I think it will always be hard to please all lawmakers. But I think this approach would be coupled with a censorship proof model. It is a protocol that is hard to outright ban, as another instance can spring up anywhere to provide a gateway to the rest.
• this is still possible. The “community” in this case is the group of people subscribing to a particular moderation policy. The key is that unsubscribing from this policy is extremely easy without much loss. User freedom is satisfied