There is a decent legal argument to be made that §230 doesn't immunize platforms for the speech of their algorithm, to the extent that said speech is different from the speech of the underlying content. (A simple, if absurd, example of this would be if I ran a web forum and then created a highlight page of all of the defamatory comments people posted, then I'm probably liable for defamation.)

The problem of course is that it's difficult to disentangle the speech of algorithmic moderation from the speech of the content being moderated. And the minor issue that the vast majority of things people complain about is just plain First Amendment-protected speech, so it's not like the §230 protections actually matter as the content isn't illegal in the first place.