Rewriting the Law that Built the Internet

This article is co-authored by Luo Xuhong and Stacey Ngiam. Cover image by Stacey Ngiam.

Image credit: ProPublica

Imagine if you had to wait weeks on end for approval by TikTok to post your video of you doing the latest TikTok trend. Yes, even that silly Instagram story you took of your lunch today will take weeks before Instagram approves for you to post it. This will be the reality of social media if Section 230 of America’s Communication Decency Act (CDA) of 1996 did not exist, a world where every social media post has to be vetted for legal exposure before it gets published. 

Section 230, commonly referred to as the 'twenty-six words that created the Internet', establishes a shield for Internet Service Providers (ISPs) and user-generated content platforms, including social media networks and online forums. Under this framework, such intermediaries are treated as distributors rather than publishers of third-party content and are therefore insulated from civil liability arising from material posted by users. Essentially, this means that if a contributor of Wikipedia had added false information about Donald Trump on his wikipedia page, Wikipedia itself is not liable for a defamation suit by Trump and his administration since they were not the publisher of the content. Instead, only the user who uploaded the content can be sued by Trump.

With this umbrella, platforms like Instagram and TikTok were able to grow rapidly, hosting vast quantities of user content without legal screening. But as concerns mount over misinformation, hate speech, and inappropriate content reaching young users, a growing chorus now asks whether Section 230 gives Big Tech too much cover, and too little incentive to manage what happens on their platforms.

Where it all began

Section 230 emerged to resolve the “moderator’s dilemma” created by two conflicting court decisions on intermediary liability: Cubby Inc. v. CompuServe Inc. in 1991 and Stratton Oakmont Inc. v. Prodigy Services Co. in 1995.

In Cubby, Inc. v. CompuServe Inc., CompuServe had hosted third-party content through its CompuServe Information Service (CIS), including the newsletter Rumorville. When defamatory statements about Cubby Inc.’s database, Skuttlebut, and its founder Robert Blanchard were published in Rumourville, Cubby sued, arguing that CompuServe should be punished for its role as a publisher. But the court rejected this argument. They saw CompuServe as a distributor, not publisher as it neither exercised editorial control over the content nor had prior knowledge of the defamatory statements, and was hence not liable. This case established the principle that liability hinged on the platform’s role: a distributor who disseminates content versus a publisher who plays a crucial role in the editorial process.

However, in Stratton Oakmont, Inc. v. Prodigy Services Co., the court reached the opposite conclusion. Prodigy Services Company had implemented content guidelines, moderators, and filtering software. Hence, when defamatory posts targeted brokerage firm Stratton Oakmont, the court ruled that Prodigy’s moderation efforts amounted to editorial control, classifying it as a publisher and holding it liable. These conflicting results created the “moderator’s dilemma”: platforms that did not moderate could avoid liability, while those that attempted to remove harmful content risked being treated as publishers. A clear incentive was established: passivity was rewarded; responsible moderation was not. But moral responsibilities argue otherwise. How could platforms be discouraged from creating a more respectful online community? 

Image credit: Stacey Ngiam

Section 230 was hence introduced to provide immunity for online intermediaries while encouraging good-faith content moderation. It allowed for online service providers to be treated as distributors of third-party generated content, providing them immunity from any legal claims arising for user posts. But this did not mean a blanket immunity. These providers were still liable for their own content or actions and any copyright claims. They were also liable for cases where criminal content was published according to the federal criminal law such as sex trafficking claims under FOSTA-SESTA. 

This framework of broad immunity with some guidelines allows for the promotion of free speech on platforms hosting user content according to the First Amendment, while eliminating the “chilling effect” where platforms over-censor out of fear of tort-based liability.

Crucially however, one might realise a significant gap in the definitions of distributor and publishers. Distributors are usually argued to be those who did not have knowledge on the content of what is distributed. But what if distributors did know about the content? Zeran v. America Online, Inc. in 1997 thus provides a clarification. In the case, Zeran sued America Online after his name was attached to an inappropriate advertisement in 1997. The United States Court of Appeals for the Fourth Circuit applied the broad immunity granted under Section 230, and concluded that distributor liability is merely a subset of publisher liability. As such, attempts to impose notice-based distributor liability were equally barred under Section 230. 

Rising concerns with Section 230

Dozens of bills have been proposed to reform section 230 in some form. Image credit: The Lawfare Institute

Since the rise of fake news, deepfakes, and increasingly politicised social media use, debate has intensified over whether platforms should retain immunity under Section 230. This concern is amplified by the growing concentration of power among a few dominant platforms that significantly shape public discourse.

Conservative or right-leaning Republicans argue that platforms are not politically neutral. They contend that Section 230 immunity enables companies to selectively censor conservative viewpoints through content removal, shadow-banning, and algorithmic suppression, while leaving liberal viewpoints comparatively untouched. Because immunity does not impose neutrality obligations, platforms retain broad discretion over content moderation. During and after his first term, Donald Trump repeatedly called for reforms to condition Section 230 protection on standards of neutrality, transparency, and non-discrimination. In his current term, Trump has yet made significant announcements regarding the repeal of Section 230. While he currently has opted to build better relationships with social media platforms instead, his continued focus on preventing content censorship reflect the ongoing Republican concerns about ideological bias.

Conversely, Democrats and liberal commentators argue that Section 230 allows platforms to insufficiently moderate harmful content. They contend that broad immunity reduces incentives to curb misinformation, extremism, and politically destabilising propaganda, particularly when algorithms amplify such content. During his term, President Joe Biden has famously criticised Section 230 protections in the context of misinformation and platform accountability, calling for the removal of the section on multiple platforms. 

Despite their differences, both sides share a core concern: platform immunity allegedly enables irresponsible or unjust content governance. The dispute, however, centres less on constitutional free speech protections and more on how private platforms exercise their moderation power.

At this point, it is important to establish the relationship between Section 230 and free speech debates that are protected by the First Amendment. Section 230 itself is a statutory liability shield, protecting online platforms from being liable in civil suits regarding content posted on their platforms by third-party users. By reducing such liability, platforms do not feel pressured by potential legal obligations to overly censor content even when the content itself may actually be legally safe from a lawsuit, therefore promoting free speech. This means that with Section 230 in place, platforms can now focus wholeheartedly on regulating content based on prevailing federal laws like pornography and their own moral compass. 

Yet, the interest in repealing (or a phased sunset) Section 230 stems from a fundamental misunderstanding of the scope of its liability shield. As discussed earlier, Republican interest in repealing Section 230 over concerns about its supposed role in the censorship of conservative viewpoints. However, doing so would go against the principles of the state action doctrine stating that the First Amendment (that calls for the uncensored free speech of a person) only applies to state and federal governments. In fact, private actors (such as ISPs) themselves have First Amendment rights – in choosing what they display and what they exercise their discretion to censor, that is an exercise of the said right. To impose a requirement of neutrality is compelled speech. 

“A private party's collection of third-party content into a single speech product (the operators’ “repertoire” of programming) is itself expressive, and intrusion into that activity must be specially justified under the First Amendment” (Moody v NetChoice, citing Turner Broadcasting System v FCC)

In fact, a compelled speech approach has not gone down well in courts either: SCOTUS has previously struck down Republican attempts at legislating against viewpoint discrimination in Florida and Texas. In rejecting state statutes targeting content moderation policies of certain large platforms, SCOTUS held that the state laws interfere with protected speech under the First Amendment and were likely unconstitutional. 

So what will happen if section 230 is gone? 

“If you repeal Section 230, one of two things will happen. Either platforms will decide they don’t want to moderate anything, or platforms will moderate everything” – Sinan Aural 

With their liability shield removed, interactive computer services will be treated as publishers instead of distributors. In other words, these websites are expected to face a deluge of lawsuits over UGC hosted on their platforms that they are now liable for. 

While websites could theoretically choose from adopting a hands-off approach to all moderation or strictly moderating all content, most are likely to choose the latter in practice. Recall CompuServe: even though the court ruled that the firm could not be held liable owing to its lack of editorial control over its own forums, a similar platform today with absolutely no moderation in place would still be liable for posts that it ought to or had reason to know about in the absence of Section 230. This means that such a platform is still obliged to review complaints over unlawful content and maintain some form of a notice-and-takedown system. 

The removal of Section 230 is also expected to entrench the foothold of existing Internet giants. In response to the escalating costs of defending against lawsuits (Section 230 also facilitates the prompt dismissal of frivolous ones) and extensive content moderation, smaller market players without deep pockets would inevitably modify their business models and cease hosting UGC entirely.

Naturally, surviving internet companies would adopt the strictest possible standards of content moderation. It won’t be far-fetched to suggest that some might even resort to pre-screening content as well. However, even the largest platforms won’t have the capacity to review the legality of all content and takedown requests. Most would take the risk-averse approach and remove all challenged content altogether instead of dedicating resources to individually investigate the merits of each complaint, leading to a major chill in Internet speech.  

Excluding “Bad Samaritans” 

Amidst the raft of changes to Section 230 that have been floated over the years, one of the most promising proposals is the creation of a “Bad Samaritan” carveout to the sweeping immunity as recommended by the Department of Justice (reform 1a). 

“It makes little sense to immunise from civil liability an online platform that purposefully facilitates or solicits third-party content or activity that would violate federal criminal law.” 

The DOJ rationalises that the original legislative intent of Section 230 was to offer protection against lawsuits that would force then-nascent internet companies to choose between either screening for and blocking all illegal content, or alternatively opting out of any content moderation. Yet, broad interpretations of Section 230 immunity have increasingly detracted from the law’s intended purpose, offering internet firms the license to act with impunity in the online world.  

The DOJ adds a nuance in its proposed reform with the heightened “purposefully” mens rea (state of mind of the defendant). In other words, section 230 immunity does not apply to website operators that are responsible for the illegal conduct (see 9th Circuit precedent below) while those that “accidentally or negligently” facilitate illegal conduct would not lose immunity. In limiting the recommendation to violations of federal criminal law, the DOJ addresses the most egregious forms of content while leaving the core of existing immunity for defamation intact. 

“[i]f you don’t encourage illegal content, or design your website to require users to input illegal content, you will be immune.” – Ninth Circuit 

Post-Zeran, the Ninth Circuit set a precedent when it narrowed its interpretation of Section 230 immunity in Fair Housing Council of San Fernando Valley v. Roommates.com, LLC. Crucially, the Ninth Circuit identified a difference between websites that passively hosted third-party content and websites designed specifically to induce illegality. In establishing the material contribution test, the court held that design decisions of websites that result in illegal content will render Section 230 immunity inapplicable. 

As the questionnaire and drop-down menu created by Roommates.com was mandatory for users as part of the registration process, the court held that this qualified it as an information content provider instead of a mere (third-party) information transmitter and hence declined to extend Section 230 protections. 

While some federal courts (e.g. Sixth Circuit in Jones v. Dirty World Entm’t Recordings LLC and Tenth Circuit in FTC v. Accusearch, Inc.) have since followed the Ninth Circuit’s precedent in denying Section 230 immunity to internet companies that facilitate illegal conduct by design, the law is not settled on this matter. Instead, courts with broader interpretations of section 230 have reached opposite conclusions. 

Two notable examples involve Backpage (First Circuit) and Armslist (Seventh Circuit), where websites that attempted to profit from commercial sex trafficking and skirt federal firearm laws respectively still received section 230 immunity. Both cases would likely have resulted in different outcomes if other circuits had uniformly followed the Ninth Circuit’s approach set out in Roommates. 

In fact, an unsatisfactory outcome in Backpage eventually pressured Congress’ 2018 passing of FOSTA-SESTA, creating a new carveout in Section 230 immunity for the knowing facilitation of sex trafficking. Had the Ninth Circuit ruled on Armslist instead, it would have likely arrived at a more logical outcome. In such a hypothetical scenario, Armslist would have been denied immunity on the grounds that it promoted illegal sales of firearms that were later used in a murder. 

No analysis of how s.230 would look like in future would be complete without examining the ubiquity of algorithms in the Internet today.  

In the 3 decades since CDA was enacted, the Internet has undergone such drastic changes in ways that its drafters would have never envisioned. From its roots as a static experience that was largely occupied by forums, chatrooms and bulletin boards, today’s digital landscape has evolved into a dynamic realm that is uniquely personalised for every Internet user. 

The vast nature of the Internet means that platforms usually contain more information than users would ever be able to consume within their lifetime. Enter the role of recommendation algorithms: tools that identify and curate the content that users are most likely to engage with, from items at the top of your Amazon landing page to the order of breaking news articles on major news sites. Meanwhile, the development of generative AI tools that derive novel outputs based on their vast training datasets spotlights the limitations of Section 230 immunity.  

At present, courts have largely granted Section 230 immunity in cases involving recommendation algorithms - as per the neutral tools doctrine first outlined in Roommates, platforms using algorithmic features that merely rearranged existing content (such as Google’s search function as highlighted in O’Kroley v. Fastcase) would be shielded from liability under Section 230. 

The 2nd Circuit similarly found in Force v Facebook that Facebook retained immunity from terrorist propaganda as its algorithms applied “neutral ranking principles” to all content without specifically promoting any harmful material. 

However, the proliferation of Generative AI tools trained on vast datasets to provide novel output in response to user prompts blurs the lines of Section 230 immunity once more. Significant questions remain unanswered on whether such AI systems contribute to the “development” of new content based on the synthesis of information from their datasets. Would such predictive algorithms involved in the generation of original content be treated akin to a traditional publisher then? Until more clarity is offered, the issue of liability arising from defamatory or otherwise illegal content found in the output of GenAI models would be a significant one for developers and regulators alike. 

As Section 230 marks its 30th anniversary this year, the law may indeed be ripe for reform. Though perhaps not in the manner advocated for by Capitol Hill. 

References

1. Access Alert: Trump Justice Department Proposes Reforms to “Section 230” Online Liability Protections. (2020, September 24). Access Partnership. https://accesspartnership.com/opinion/access-alert-trump-justice-department-proposes-reforms-to-section-230-online-liability-protections/

2. Citron, D., & Franks, M. (2020). The Internet as a Speech Machine and Other Myths Confounding Section 230 Reform . University of Chicago Legal Forum, 2020, 3. https://chicagounbound.uchicago.edu/cgi/viewcontent.cgi?article=1662&context=uclf

3. Cornell Law School Legal Information Institute. (1996). 47 U.S. Code § 230 - Protection for private blocking and screening of offensive material. Legal Information Institute; Cornell Law School. https://www.law.cornell.edu/uscode/text/47/230

4. Cubby, Inc. v. CompuServe Inc., 776 F. Supp. 135 (S.D.N.Y. 1991). (n.d.). Justia Law. https://law.justia.com/cases/federal/district-courts/FSupp/776/135/2340509/

5. Dick, E., & Johnson, A. (2021, April 7). What We Learned When We Broke the Internet: Insights From the Section 230 Debate. Information Technology and Innovation Foundation. https://itif.org/publications/2021/04/07/what-we-learned-when-we-broke-internet-insights-section-230-debate/

6. Dickinson, G. M. (2025). Section 230: A Juridical History. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.5164697

7. Feiner, L. (2019, August 20). Republican-led report finds Facebook has “significant work” to earn conservatives’ trust. CNBC. https://www.cnbc.com/2019/08/20/republican-report-of-facebook-anti-conservative-bias-suggests-changes.html

8. Hamilton, I. A. (2020, November 15). If Biden revokes Section 230, it could decide Big Tech’s future. Business Insider. https://www.businessinsider.com/what-biden-administration-means-for-section-230-2020-11

9. Jacobson, L. (2026, February 19). Trump hasn’t spoken about Section 230 in more than three years. Politifact. https://www.politifact.com/truth-o-meter/promises/maga-meter-tracking-donald-trumps-2024-promises/promise/1688/sign-a-bill-revisiting-section-230-for-tech-platfo/article/3242/

10. Johnson, A., & Castro, D. (2021, February 22). Overview of Section 230: What It Is, Why It Was Created, and What It Has Achieved. Information Technology and Information Foundation. https://itif.org/publications/2021/02/22/overview-section-230-what-it-why-it-was-created-and-what-it-has-achieved/

11. Kern, R. (2022, September 8). White House renews call to “remove” Section 230 liability shield. POLITICO. https://www.politico.com/news/2022/09/08/white-house-renews-call-to-remove-section-230-liability-shield-00055771

12. Kerry, C. (2021, May 14). Section 230 reform deserves careful and focused consideration. Brookings. https://www.brookings.edu/articles/section-230-reform-deserves-careful-and-focused-consideration/

13. Kiss, N. (2025). The Twenty-Six Words That Created the Internet… and then Maybe, Kind Of, Destroyed Society: Understanding and Reforming Section 230 of the Communications Decency Act. Michigan Technology Law Review, 31.1, 131. https://doi.org/10.36645/mtlr.31.1.twenty-six

14. Lanier, J., Stanger, A., & Tang, A. (2024, October 30). Sunset and Renew: Section 230 Should Protect Human Speech, Not Algorithmic Virality. Ash Center. https://ash.harvard.edu/articles/sunset-and-renew-section-230-should-protect-human-speech-not-algorithmic-virality/

15. Lim, D. (2026, February 4). What would happen if Section 230 went away? A legal expert explains the consequences of repealing “the law that built the internet.” The Conversation. https://theconversation.com/what-would-happen-if-section-230-went-away-a-legal-expert-explains-the-consequences-of-repealing-the-law-that-built-the-internet-253326

16. Masnick, M. (2019, April 18). New Paper: Why Section 230 Is Better Than The First Amendment. Techdirt. https://www.techdirt.com/articles/20190411/09024441979/new-paper-why-section-230-is-better-than-first-amendment.shtml

17. Megan. (2025). Generative AI Meets Section 230: The Future of Liability and Its Implications for Startup Innovation. The University of Chicago Business Law Review. https://businesslawreview.uchicago.edu/print-archive/generative-ai-meets-section-230-future-liability-and-its-implications-startup

18. Miers, J. (2020, November 2). Your Problem Is Not With Section 230, But The 1st Amendment. Techdirt. https://www.techdirt.com/2020/11/02/your-problem-is-not-with-section-230-1st-amendment/

19. Robertson, A. (2021, February 9). Section 230 is 25 years old, and it’s never been more important. The Verge. https://www.theverge.com/22268421/cda-section-230-25th-anniversary-reform-stakes-big-tech-internet

20. Sarceño Robles, C. (2022). Section 230 is Not Broken: Why Most Proposed Section 230 Reforms Will Do More Harm Than Good, and How the Ninth Circuit Got it Right. FIU Law Review, 16(1). https://doi.org/10.25148/lawrev.16.1.15

21. Schnatterbeck, C. (2023, July 3). The 26 Words that Ended the Internet: The Bipartisan Dilemma Over Section 230. Columbia Political Review. https://www.cpreview.org/articles/2023/7/the-26-words-that-ended-the-internet-the-bipartisan-dilema-over-section-230

22. Terr, A. (2023, February 20). Why repealing or weakening Section 230 is a very bad idea. The Foundation for Individual Rights and Expression. https://www.thefire.org/news/why-repealing-or-weakening-section-230-very-bad-idea

23. US Department of Justice. (2020, June). Section 230 — Nurturing Innovation or Fostering Unaccountability? https://www.justice.gov/ag/media/1072971/dl?inline=

24. Waheed, N. (2024, September 4). Section 230 and its Applicability to Generative AI: A Legal Analysis. Center for Democracy and Technology. https://cdt.org/insights/section-230-and-its-applicability-to-generative-ai-a-legal-analysis/