Section 230, the 1996 Liability Shield, and the Fight to Rewrite It

In October 1994, an anonymous user on a Prodigy bulletin board called Money Talk posted that a Long Island brokerage was engaged in fraud and that its president was a criminal. The brokerage was Stratton Oakmont, the firm later made famous by a movie about its founder, Jordan Belfort, and by a long list of securities regulators. Stratton Oakmont sued Prodigy for libel, asking for $200 million. In May 1995 a New York state judge ruled that Prodigy could be treated as the publisher of the post, and so could be held liable for it, precisely because Prodigy had tried to keep its boards family-friendly by screening content. The company had tried to moderate, and that effort made it responsible for everything it missed. Within months, two congressmen had a fix. It became twenty-six words of federal law that now sit underneath almost everything people do online.

In short: Section 230 says websites are generally not liable as publishers for what their users post, and that they can remove content in good faith without taking on that liability. It was written to undo the perverse lesson of a 1995 case that punished a company for moderating. The fight over rewriting it now divides both parties, and the best reforms target specific harms rather than the whole shield.

The problem a Wall Street firm handed Congress

Before Stratton Oakmont, the leading case pointed the other way. In 1991 a federal court in New York had ruled in Cubby v. CompuServe that CompuServe was more like a bookstore or a library than a newspaper. It did not review what users posted in a forum, so it was a distributor, and distributors are liable for defamation only if they know or should know about it.

Put the two decisions side by side and the incentive is obvious. CompuServe did nothing and won. Prodigy tried to clean up its boards and lost. Any online service reading those cases would conclude that the safest legal strategy was to look away from everything: no filters, no moderators, no removal of obscenity or harassment. In 1995 that was an alarming conclusion, because Congress was in the middle of a panic about children finding pornography online.

Representatives Chris Cox, a California Republican, and Ron Wyden, an Oregon Democrat, proposed a different answer. Their bill would protect services that chose to moderate, and it would make clear that hosting someone else's speech did not make a service the speaker. It was folded into the Communications Decency Act, which was itself Title V of the Telecommunications Act of 1996. President Bill Clinton signed the whole package on February 8, 1996.

What the twenty-six words actually say

The central sentence, Section 230(c)(1), reads: "No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider."

That sentence does two quiet things. It covers users as well as companies, so a person who forwards an email or retweets a post is not liable for what someone else wrote. And it applies to any interactive computer service, which courts have read to include search engines, social networks, review sites, comment sections, email providers and online marketplaces.

The second part, Section 230(c)(2), protects services from liability for any action taken voluntarily and in good faith to restrict access to material they consider objectionable. This is the part aimed directly at the Prodigy problem. A site can remove spam, nudity, harassment or anything else it decides it does not want, and it cannot be sued for that choice on the theory that removing some content makes it responsible for the rest.

What the law does not do is just as important, and it is often misunderstood. Section 230 does not protect a company for content it creates itself. It does not override federal criminal law, so it is no defense to a federal prosecution. It does not touch intellectual property law, which is why copyright disputes online run through a separate notice-and-takedown system created in 1998. It does not apply to certain federal communications privacy law. And since 2018, when Congress passed a pair of bills known as FOSTA and SESTA, it does not protect sites from certain claims related to sex trafficking.

Its odd survival in 1997

Most of the Communications Decency Act did not survive. Its main provisions made it a crime to send indecent material to minors online, and civil liberties groups challenged them immediately. In 1997, in Reno v. ACLU, the Supreme Court struck down the indecency provisions as a violation of the First Amendment. The Court described the internet as a uniquely open medium that deserved full protection, the opposite of how it had treated broadcasting in cases about radio.

Section 230 was in a different part of the law and was not challenged. So the irony stands: the law Congress passed to restrict online speech was mostly thrown out, and the provision meant to encourage moderation became the legal foundation of the open internet.

The same year, a federal appeals court read Section 230 broadly in Zeran v. America Online. A man named Kenneth Zeran had been the target of anonymous posts on AOL advertising offensive T-shirts about the Oklahoma City bombing, listing his phone number. He received threatening calls and asked AOL to take the posts down. He sued when it was slow to act. The Fourth Circuit ruled that Section 230 barred his claims even though AOL had been notified, reasoning that liability on notice would push services to remove any content anyone complained about. That decision set the pattern for the next two decades: courts treated the shield as wide and strong.

Where the courts have drawn lines

The shield has limits, and courts have found them in specific places.

In 2008 the Ninth Circuit, sitting en banc, ruled in a case about Roommates.com that the site lost protection for questions it required users to answer about sex, sexual orientation and family status, because the site itself had designed those questions and helped develop the allegedly discriminatory content. A site that materially contributes to what makes content unlawful becomes, in part, a content provider itself.

In 2021 the same court ruled in Lemmon v. Snap that Section 230 did not bar a lawsuit by parents of young men killed in a high-speed crash. The parents alleged that Snapchat's speed filter, which recorded how fast a user was going, encouraged dangerous driving. The court said the claim was about the negligent design of the product, not about any message a user had posted. That reasoning, separating product design from user speech, has become the main road for lawsuits against platforms over harm to children, and many of those cases are working through state and federal courts.

The Supreme Court has largely stayed out. In 2023 it heard Gonzalez v. Google, a case asking whether YouTube's recommendation algorithm was protected when it suggested terrorist recruitment videos. Then it declined to decide the Section 230 question at all, sending the case back after ruling in a companion case, Twitter v. Taamneh, that the plaintiffs had not shown the platforms aided and abetted terrorism. In 2024, in Moody v. NetChoice, the Court said that when platforms curate and moderate their feeds, that activity is generally expressive and protected by the First Amendment, and it sent challenges to Texas and Florida social media laws back for more analysis. Notice what that means for reform: even without Section 230, the First Amendment protects a great deal of moderation.

The two parties' complaints point in opposite directions

Section 230 is unusual in having critics in both parties who want to change it for opposite reasons.

Many Republicans argue that large platforms use the moderation shield to remove conservative voices, and they want to condition protection on political neutrality or narrow the good-faith clause. During his first term Donald Trump signed an executive order targeting Section 230 and vetoed a defense authorization bill in part because it did not repeal it; Congress overrode that veto. Many Democrats argue the opposite: that platforms do too little, profiting from harassment, fraud, misinformation and content harmful to children while hiding behind immunity.

Those complaints cannot both be satisfied by the same reform. A law that pushes platforms to remove more harmful content will also lead them to remove more contested political speech. A law that punishes them for removing political content will make them more cautious about removing anything.

That is the honest core of the problem, and it is why I think broad repeal is the wrong goal. Without Section 230, the first casualties would not be the largest companies. They can afford lawyers and armies of moderators. The casualties would be small forums, local news comment sections, independent review sites and new competitors who could not survive a stream of defamation suits. The likely result of repeal would be more power for the incumbents, not less, along with the old Prodigy problem: sites either moderating everything aggressively or not at all.

The case for change, made carefully

None of this means the law should be left as it is. The internet of 1996 was bulletin boards and early web pages. The internet of 2026 includes recommendation engines that decide what billions of people see, ad systems that profit from engagement, and generative AI tools that produce text and images that no user wrote.

Here are the reforms that I think have the strongest case.

Product design. The Lemmon v. Snap line of cases can be clarified in statute: Section 230 should not protect a company from liability for the design of its own product, such as features that are dangerous for children, when the claim does not depend on treating the company as the publisher of a particular user's speech.

Paid content. When a platform is paid to run an advertisement, it has a commercial relationship with the content and a chance to review it. Narrowing immunity for paid ads, particularly fraudulent ones, is a targeted change that would not touch ordinary user speech.

Generative AI. It is an open question whether output generated by an AI model is information provided by another information content provider at all. Courts have not settled it, and Congress could clarify that companies are responsible for what their own AI systems produce.

Transparency. Platforms could be required to disclose how their moderation and recommendation systems work and to give researchers access to data, without the government dictating any particular moderation decision.

Each of those is narrower than repeal, and each targets a harm people in both parties recognize. There is a precedent for this kind of cross-party action. FOSTA and SESTA passed with overwhelming bipartisan majorities in 2018, though civil liberties groups and sex worker advocates argued, with some evidence, that the law made people less safe by pushing ads offline and that it was rarely used by prosecutors. That outcome is a warning: a carve-out written in a hurry can miss its target. Reform should be specific and should be tested against who it actually protects.

What the shield means for a small site

Debates about Section 230 usually focus on the biggest platforms, which makes sense, since they host most of the speech. But the law matters most, in practical terms, to everyone else.

Think of a neighborhood forum where people post about lost pets, crime and the city council. Think of a site that collects reviews of local contractors, a hobby board for model railroad builders, a school's parent group, a nonprofit's comment section or a small newspaper that lets readers argue under its stories. Each of those hosts speech written by other people. Each will sometimes host a post that someone considers defamatory. Without Section 230, each would face a choice: screen everything before it appears, which most cannot afford, or stop hosting user posts at all.

The shield also protects the people who do the moderating. Wikipedia's articles are written and policed by volunteers. Reddit's communities are run largely by volunteer moderators who set and enforce their own rules. When Reddit filed a brief in the Gonzalez case, it stressed that point: the users who remove posts and recommend content could be exposed to liability if the law were narrowed carelessly. That is the part of Section 230 most people never think about. It protects users as well as companies, and much of the internet's moderation is done by users.

None of that settles the harder questions about the largest platforms. It is a reason to write reforms with care, and to ask, of any proposal, who could afford to comply with it.

What you can do

Proposals to change Section 230 come back in nearly every Congress, often attached to children's online safety bills. A few ways to make sure the people writing it hear from you:

• Contact your representatives early, before a bill reaches the floor. Use the lookup tool to find your House member, senators and state legislators, with their phone numbers, contact forms and latest votes. Ask specifically whether they support targeted changes on product design and paid advertising rather than full repeal.
• Watch for state action, too. Many states have passed or proposed online safety laws for minors, and several are being challenged in court. Court Watch covers Supreme Court opinions and the emergency docket, which is where those fights often end up.
• Look for the bills that drew both parties. The common-ground tracker lists bills both parties passed together, and children's online safety has been one of the few tech areas with genuine bipartisan support.
• Follow the broader story of platform power. Our topic page on Big Tech and AI gathers reporting on platform accountability, AI regulation and the companies that run the internet.

The Prodigy judge in 1995 was applying old law faithfully. The result was a rule that punished anyone who tried to clean up a public space. The lesson of Section 230 is that the details of liability shape behavior, and that lesson applies just as much to any rewrite.