If the "safe harbor" you're referring to is CDA230, you have it quite backwards. They're shielded under that statute even if they ban people and delete things s

If the "safe harbor" you're referring to is CDA230, you have it quite backwards. They're shielded under that statute even if they ban people and delete things selectively, regardless of the reason. Under the legal environment at the time CDA230 was enacted, they already had a shield. What they didn't have was the ability to moderate their own platforms and still keep that shield, they had to pick one or the other because if they screwed up and missed something then suddenly they were liable again. So CDA230 actually encouraged them to moderate their platforms, arbitrarily if they want to. The "moderation" they were expected to do was mostly to remove "indecent" content, but they aren't required to do that or forbidden from removing anything else. So at least in that respect, Twitter and Apple are doing exactly what they were expected to do.