It is possible to distinguish between censorship and spam filtering. In the case of censorship, the speaker wants to say something and the listener wants to hear it and the censor prevents this. In the case of spam filtering, the spammer wants to say something and the listener doesn't want to hear it and voluntarily requests that a third party filter it out, with the option to individually disable this filtering.
Now, someone could implement censorship and call it a spam filter, e.g. it filters spam and also disfavored facts and people leave it on because there is only a single on/off toggle and they don't want to be deluged with spam. But what this implies is that a "spam filter" with uncorrectable "false positives" is equivalent to censorship.