The law often limits the liability of an intermediary for the speech it carries. And rightly so, because imposing liability on intermediaries can induce them to filter out questionable content and this “collateral censorship” risks suppressing much lawful, even highly beneficial, speech. The “collateral censorship” rationale has its limits, though, and correspondingly, so should the applicability of intermediary immunity. The worry with collateral censorship is not just that intermediaries censor, but that they censor more than an original speaker would in the face of potential liability. Increased censorship, in turn, is the product of applying liability targeted at original speakers to entities whose interests diverge from original speakers. Where the “intermediary” has the interests of an original speaker, and so should be regarded as one, or where the form of liability already takes into account the intermediary’s interests, collateral censorship is not the problem, and immunity is not the right response. This understanding should, in particular, inform the interpretation of section 230 of the Communications Decency Act of 1996, a federal statute that broadly immunizes Internet intermediaries from speech torts and many other forms of liability. In cases involving the republication of e-mails, questionnaires, member screening, and contract claims, among others, courts have begun to explore the limits of this immunity, but have done so haphazardly and inconsistently, having largely lost sight of the underlying rationale for immunity. Focusing on the conditions that generate problematic collateral censorship provides a principled basis upon which to define the limits of intermediary immunity generally, and section 230 in particular.