Section 230 is twenty-six words, and most of the debate skips them

Section 230 is a 1996 provision of the Communications Decency Act whose operative clause runs twenty-six words: no provider of an interactive computer service shall be treated as the publisher of information provided by another. The text, at 47 U.S.C. § 230(c)(1) as published in the United States Code, is shorter than nearly every headline written about it, and the debates it anchors usually turn on what those words skip as much as what they say. This analysis reads the provision as written, then examines the strongest argument on each side of the reform question.

Media News Watch publishes analysis of the record, not legal advice, and takes no position on any proposal. The standard applied here is symmetrical: the provision's defenders and critics both get read at their strongest, from the documents.

What does the text actually say?

The provision has two operative parts. Clause (c)(1), the twenty-six words, bars treating a service as the publisher of third-party content. Clause (c)(2) adds a second shield: good-faith removal of objectionable material is protected in its own right. The pairing is deliberate, and it is the part of the design most often flattened in summary: the law protects both hosting decisions and moderation decisions, in separate clauses, per the statutory text.

Two definitions carry weight. An interactive computer service, under § 230(f)(2), includes any service offering access to the internet — a definition written before social platforms existed. And nothing in the text requires political neutrality; that condition is absent from the statute, a point on which the case law has been consistent since the 1997 ruling in Zeran v. America Online, the first major appellate decision applying the provision.

What does the provision not protect?

The exceptions are enumerated in the statute itself, and they are wider than the public debate usually registers. Federal criminal law, intellectual property claims, and the sex-trafficking carve-out added by the 2018 FOSTA amendment are all outside the shield, per § 230(e). The provision also does not cover a platform's own speech — its original content, its advertising claims, its product statements — because the shield extends only to information provided by another content provider.

What the provision does not do is impose obligations. It contains no transparency requirement, no appeal process, no moderation standard. It is a defense to liability, not a regulatory scheme; platforms' moderation systems exist under it, not because of it. That asymmetry — a strong defense paired with no affirmative duties — is the structural fact beneath most reform proposals, whatever their direction.

Is the reform debate engaging the actual text?

The strongest criticism, developed in legal scholarship and congressional testimony over the past decade, is conditional immunity: critics argue the provision as interpreted lets platforms make curation choices at scale without the editing liability a newspaper would carry for comparable decisions. The strongest defense, equally developed in scholarship and in platform amicus filings, is that the defense is what allows moderation itself — that a publisher-standard liability would push platforms toward either removing far more speech or hosting far less of it, with the first amendment applying to the government rather than to the platforms either way.

Both positions engage the text. The debate that does not is the version that treats § 230 as a neutrality mandate or as a full exemption from law; neither reading appears in the twenty-six words or in the enumerated exceptions. Proposals in Congress have included conditioning the shield on neutrality review and narrowing it by category of content; per the Congressional Research Service's summaries of reform proposals, none has been enacted as of 2024.

What the record establishes is a provision doing exactly what its text says, applied by courts to services its drafters did not imagine. What remains genuinely open is the policy question — whether the 1996 balance serves a 2020s information system — and that question is decided by legislatures, not by the statute's existing words.