Almost every site and service you use on the Internet involves content created by someone other than the operator. Most obviously, social media networks such as Facebook, Twitter and YouTube rely on their users to create content. But so do Wikipedia and reviews sites such as Yelp and TripAdvisor. The vast majority of news and commentary websites allow users to post comments on each article. Email, messaging services, and video chat all empower users to communicate with each other. And it’s not just content created by users. Google and other search engines catalog websites created by others and attempt to offer users the most relevant results for their search. Each of those websites relies on a complex system of “intermediaries” to host their content, register their domain name, protect those sites against attack, serve the ads that sustain those sites, and much more. Google, Apple, Microsoft, Amazon and others offer third-party apps for download on mobile devices, personal computers, televisions, gaming consoles, watches, and much else.
How did such a complex ecosystem evolve? Try, for a moment, to look at the Internet the way Charles Darwin looked at the complexity of the organic world in the last paragraph of On The Origin of Species. In describing how the simple laws of evolution could produce the wondrous diversity of life we see all around us, Darwin wrote:
It is interesting to contemplate an entangled bank, clothed with many plants of many kinds, with birds singing on the bushes, with various insects flitting about, and with worms crawling through the damp earth, and to reflect that these elaborately constructed forms, so different from each other, and dependent on each other in so complex a manner, have all been produced by laws acting around us. These laws, taken in the largest sense, being Growth with Reproduction; inheritance which is almost implied by reproduction; Variability from the indirect and direct action of the external conditions of life, and from use and disuse; a Ratio of Increase so high as to lead to a Struggle for Life, and as a consequence to Natural Selection, entailing Divergence of Character and the Extinction of less-improved forms. Thus, from the war of nature, from famine and death, the most exalted object which we are capable of conceiving, namely, the production of the higher animals, directly follows. There is grandeur in this view of life, with its several powers, having been originally breathed into a few forms or into one; and that, whilst this planet has gone cycling on according to the fixed law of gravity, from so simple a beginning endless forms most beautiful and most wonderful have been, and are being, evolved.
Now throw in lawyers and America’s uniquely litigious culture: how far would evolution have progressed if every step had been subject to litigation? In the digital world, this is not a facetious question. Many “natural” “laws” have shaped the evolution of today’s Internet. Moore’s Law: processor speeds, or overall processing power for computers, will double every two years. Metcalfe’s law: “the effect of a telecommunications network is proportional to the square of the number of connected users of the system.” But one statutory law has made it all possible: Section 230 of the Communications Decency Act of 1996.
The law’s premise is simple: “Content creators—including online services themselves—bear primary responsibility for their own content and actions. Section 230 has never interfered with holding content creators liable. Instead, Section 230 restricts only who can be liable for the harmful content created by others.” Specifically, Section 230(c)(1) protects “interactive computer service” (ICS) providers and users from civil actions or state (but not federal) criminal prosecution based on decisions they make as “publishers” with respect to third-party content they are in no way responsible for developing. In its 1997 Zeran decision, the Fourth Circuit became the first appellate court to interpret this provision, concluding that “lawsuits seeking to hold a service provider liable for its exercise of a publisher’s traditional editorial functions—such as deciding whether to publish, withdraw, postpone or alter content—are barred.” A second immunity, Section 230(c)(2)(A) protects (in different ways) decisions to “restrict access to or availability of material that the provider or user considers to be . . . objectionable,” even if the provider is partly responsible for developing it.
These two immunities, especially (c)(1), allow these ICS providers and users to short-circuit the expensive and time-consuming litigation process quickly, usually with a motion to dismiss—and before incurring the expense of discovery. In this sense, Section 230 functions much like anti-SLAPP laws, which, in 30 states, allow defendants to quickly dismiss strategic lawsuits against public participation (SLAPPs)—brought by everyone from businesses fearful of negative reviews to politicians furious about criticism—that seek to use the legal system to silence speech.
Without Section 230 protections, ICS providers would face what the Ninth Circuit, in its landmark Roommates decision, famously called “death by ten thousand duck-bites:” liability at the scale of the billions of pieces of content generated by users of social media sites and other third parties every day. As that court explained, “section 230 must be interpreted to protect websites not merely from ultimate liability, but from having to fight costly and protracted legal battles.” For offering that protection, Section 230(c)(1) has been called the “the twenty-six words that created the Internet.”
What competition and consumer-protection claims does Section 230 actually bar with respect to content-moderation decisions? That depends on whether, when making such decisions, ICS providers and users are exercising the editorial discretion protected by the First Amendment for traditional media operators and new media operators alike under Miami Herald Publishing Co. v. Tornillo, 418 U.S. 241 (1974). Section 230’s protection is effectively co-extensive with the First Amendment’s. Like anti-SLAPP legislation, Section 230 merely supplies defendants a procedural shortcut to bypass the litigation process and quickly reach the end result that the First Amendment already assures them. That result is that they can moderate content as they see fit—except, just as newspapers may not engage in economic conduct that harms competition, nor are ICS providers acting “as publishers” or “in good faith” when they make content moderation decisions for anti-competitive reasons, and with anti-competitive effect, rather than editorial judgment. The First Amendment protects the right of newspapers to reject ads for editorial reasons, yet a “publisher may not accept or deny advertisements in an ‘attempt to monopolize . . . any part of the trade or commerce among the several States . . . .’” In consumer protection law, deception and other causes of action based on misleading advertising always require that marketing claims be capable of being disproven objectively.
Until recently, most of the debate over rethinking Section 230 has been about increasing liability for content that ICS providers and users do decide to publish. Most notably, in 2017, Congress amended Section 230 for the first time: websites have never been immune from federal criminal prosecution for sex trafficking or anything else, but the Stop Enabling Sex Traffickers Act (SESTA) allows for state criminal prosecutions and civil liability while strengthening federal law on the promotion of prostitution.
Congress is again debating legislation aimed at combatting child sexual exploitation (CSE) as well as child sexual abuse material (CSAM): the EARN IT Act would create vast, vague liability for enabling not only the transmission of CSAM videos and images but also communications between minors and adults that might amount to solicitation or “grooming.” EARN IT could cause online communications services to compromise encryption or other security features, lead to mandatory age and identity verification, and prompt unconstitutional suppression of anonymous speech.
A flurry of other pending bills would amend Section 230 to hold websites liable for a variety of other content, from misinformation to opioids. Democrats and civil rights groups have particularly focused on claims that Section 230 has obstructed enforcement of anti-discrimination law—even though the Department of Housing and Urban Development (HUD) has sued Facebook anyway for allowing buyers of housing ads to target audiences on the basis of race, religion, or gender. Section 230 won’t bar that suit if a court decides that Facebook was responsible, even “in part,” for “the creation or development of [those ads],” just as Roommates.com was held responsible for housing discrimination because it allowed its users to specify race-based preferences in posting, and searching, housing ads on the site. “Internet platforms are generally protected from liability for third-party content,” explains former Rep. Chris Cox (R-CA), Section 230’s primary author, “unless they are complicit in the development of illegal content, in which case the statute offers them no protection.”
In 2020, the political focus of the debate around Section 230 shifted dramatically—from holding ICS providers liable for not doing enough to moderate potentially unlawful or objectionable content to holding them responsible for doing too much content moderation, or at least, moderating content in ways that some find “biased.” Republicans blame Section 230 for preventing them from enforcing antitrust and consumer protection laws against “Big Tech” for discriminating against conservatives in how they treat third-party content. In May, President Trump signed an Executive Order that ordered an all-out attack on Section 230 on multiple fronts. At the President’s direction, the Department of Justice proposed to add an exemption to Section 230 for antitrust claims. Likewise both DOJ and the National Telecommunications Information Administration (NTIA) want to make it easier to sue tech companies for how they moderate third-party content, label it, and decide how to present it to their users. Both DOJ and NTIA claim that courts have misinterpreted Section 230, but while DOJ wants Congress to amend the law, NTIA has filed a petition (as directed by the Executive Order) asking the Federal Communications Commission (FCC) to reinterpret the law by rulemaking. Because the NTIA Petition (and the comments filed in response to it, including reply comments by NTIA) provides the most detailed explanation of the new Republican party line on Section 230, it provides a concrete basis for much of the discussion below. TechFreedom’s comments and reply comments in that docket provide a more detailed response than possible here. As requested by the Executive Order, multiple Republican lawmakers have introduced legislation to do the same thing, allowing lawsuits for alleged bias in content moderation.
Democrats have yet to sign onto legislation that would directly regulate bias. Bipartisan draft legislation in the Senate, the PACT Act, would require ICS providers other than “small business providers” to (i) “reasonably inform users about the types of content that are allowed on the interactive computer service,” (ii) follow certain procedures (providing notice to affected users and allowing appeals) after taking down content they created, and (iii) issue annual transparency reports. The FTC could sue ICS providers for failing to take down “policy-violating content” within 14 days of being provided notice of it. But unlike Republican proposals, the PACT Act would not allow ICS providers to be sued for how they interpret or apply their terms of service, nor would it require any kind of neutrality or limit the criteria on which content can be moderated. Thus, this draft bill avoids the most direct interference with the editorial discretion of ICS providers in choosing which content they want to carry. The bill would make two other major amendments to Section 230: it would let state Attorneys General enforce federal civil law and require platforms to take down user content that a court has deemed illegal. While still problematic, compared to other bills to amend Section 230, the PACT Act shows a better understanding of what the law currently is and how content moderation is done Rep. Jan Schakowsky (D-IL) has floated a discussion draft of the Online Consumer Protection Act, which would, notwithstanding any protection of Section 230, require all “social media companies” (regardless of size) to “describe the content and behavior the social media platform or online marketplace permits or does not permit on its service” in both human-readable and machine-readable form. The FTC would punish any violations of terms of service as deceptive practices.
This chapter focuses on these and other complaints about “content moderation” because ICS providers are far more likely to be subject to unfair competition and consumer protection claims (meritorious or otherwise) for choosing not to carry, or promote, some third party’s content. The term is used throughout to include not only blocking or removing third party content or those who post it, but also limiting the visibility of that content (e.g., in “Safe Mode”), making it ineligible for amplification through advertising (even if the content remains up), or making it ineligible for “monetization” (i.e., having advertising appear next to it).
We begin by providing an introduction to this statute, its origins, and its application by the courts since 1996. We then explain the interaction between Section 230 and the antitrust and consumer-protection laws. We conclude by discussing current proposals to amend or reinterpret Section 230 aimed specifically at making it easier to bring antitrust and consumer protection suits for content moderation decisions.
Recently, DOJ has claimed “the core objective of Section 230 [is] to reduce online content harmful to children.” This is an accurate summary of what the Communications Decency Act, as introduced in the Senate by Sen. John Exon (D-NE), aimed to accomplish. But it is a complete misrepresentation of Section 230, an entirely separate piece of legislation first introduced in the House by Rep. Chris Cox (R-CA) and Ron Wyden (D-OR) as the Internet Freedom and Family Empowerment Act. As Cox recently explained:
One irony . . . persists. When legislative staff prepared the House-Senate conference report on the final Telecommunications Act, they grouped both Exon’s Communications Decency Act and the Internet Freedom and Family Empowerment Act into the same legislative title. So the Cox-Wyden amendment became Section 230 of the Communications Decency Act—the very piece of legislation it was designed to rebuke. Today, with the original Exon legislation having been declared unconstitutional, it is that law’s polar opposite which bears Senator Exon’s label.
Rejecting the idea that Section 230 was “a necessary counterpart” to the rest of the Communications Decency Act as Trump’s DOJ has claimed, Cox notes:
The facts that the Cox-Wyden bill was designed as an alternative to the Exon approach; that the Communications Decency Act was uniformly criticized during the House debate by members from both parties, while not a single Representative spoke in support of it; that the vote in favor of the Cox-Wyden amendment was 420-4; and that the House version of the Telecommunications Act included the Cox-Wyden amendment while pointedly excluding the Exon amendment—all speak loudly to this point.
Newt Gingrich, then Speaker of the House, as Cox notes:
slammed the Exon approach as misguided and dangerous. “It is clearly a violation of free speech, and it’s a violation of the right of adults to communicate with each other,”[said Gingrich], adding that Exon’s proposal would dumb down the internet to what censors believed was acceptable for children to read. “I don’t think it is a serious way to discuss a serious issue,” he explained, “which is, how do you maintain the right of free speech for adults while also protecting children in a medium which is available to both?”
Section 230 provided a simple answer to that question: empowering the providers and users of Internet services to decide for themselves how to approach content moderation. Instead of one right answer for the entire country, there would be a diversity of approaches from which users could choose—a “Utopia of Utopias,” to borrow the philosopher Robert Nozick’s famous phrase. As Rep. Cox put it, “protecting speech and privacy on the internet from government regulation, and incentivizing blocking and filtering technologies that individuals could use to become their own censors in their own households” were among the core purposes of Section 230.
What’s also important to note is that only Section 230, a separate bill that was merged with CDA, is still standing. The Communications Decency Act, which did aim to protect children online, was found to be unconstitutional in a landmark 7-2 decision by the Supreme Court. Justice John Paul Stevens wrote that the CDA placed “an unacceptably heavy burden on protected speech” that “threaten[ed] to torch a large segment of the Internet community.”
Section 230 empowered not only content controls within each household but, perhaps even more importantly, choice among communities with different approaches: Facebook, Twitter, YouTube, Gab, and Parler all offer markedly different “utopias” among which users—and advertisers—are free to choose.
It is commonly said that that “Section 230(c)(1) . . . was intended to preserve the rule in Cubby v. Compuserve: platforms that simply post users’ content, without moderating or editing such content, have no liability for the content.” In fact, Section 230 voided both Cubby, Inc. v. CompuServe Inc. and Stratton Oakmont, Inc. v. Prodigy Servs. Co. Both decisions created a version of the “Moderator’s Dilemma,” in which websites have a perverse incentive to avoid content moderation because it will increase their liability. Prodigy, more famously, held websites liable by virtue of attempting to moderate content. Compuserve found no liability, but made clear that this finding depended on the fact that CompuServe had not been provided adequate notice of the defamatory content, implying that such notice would trigger a takedown obligation under a theory of distributor liability. Thus, Compuserve created a perverse incentive not to become aware of the unlawful nature of user content. Similarly, many critics of Section 230 insist the statute is being misapplied because the companies it protects are “obviously publishers”—when the entire purpose of 230(c)(1), as we will see next, is to say that the traditional distinctions among publishers, distributors, and other actors are irrelevant.
To avoid misinterpreting Section 230, practitioners must parse the text of the statute carefully. “The CDA . . . does more than just overrule Stratton Oakmont,” as the Tenth Circuit observed. “To understand the full reach of the statute, we will need to examine some of the technical terms used in the CDA.” Let us do exactly that.
As Section 230 has become more politicized, elected officials and advocates on both sides of the aisle have decried the law as a subsidy to “Big Tech.” This claim is false in at least three ways. First, all three of Section 230’s immunity provisions are worded the same way, protecting any “provider or user of an interactive computer service.” For example, President Trump has used Section 230(c)(1) to dismiss a lawsuit against him for retweeting another user’s content. Every user of social media is likewise protected by 230(c)(1) when they reshare third party content. Second, Section 230 applies to far more than just “social media sites.” The term “information content provider” encompasses a wide variety of services, and thus allows the statute to be truly technologically neutral. Third, as we and other experts noted in a declaration of “principles for lawmakers” concerning the law last year:
Section 230 applies to services that users never interact with directly. The further removed an Internet service—such as a DDOS protection provider or domain name registrar—is from an offending user’s content or actions, the more blunt its tools to combat objectionable content become. Unlike social media companies or other user-facing services, infrastructure providers cannot take measures like removing individual posts or comments. Instead, they can only shutter entire sites or services, thus risking significant collateral damage to inoffensive or harmless content. Requirements drafted with user-facing services in mind will likely not work for these non-user-facing services.
Section 230(e) preserves claims raised under federal criminal law, “any law pertaining to intellectual property,” the Electronic Communications Privacy Act and “any similar state law,” certain sex trafficking laws, and “any State law that is consistent with this section.”
Thus, Section 230’s immunities covers civil claims, both state and federal, and all state criminal liability—insofar as they either seek to hold a defendant liable as a publisher ((c)(1)), for content removal decisions ((c)(2)(A)), or for providing tools for content removal decisions to others ((c)(2)(B)), as discussed in the following sections.
The National Association of Attorneys General (NAAG) has asked Congress to amend Section 230 to preserve state criminal liability. But broad preemption of state criminal law reflected a deliberate policy judgment, as Rep. Cox has noted:
the essential purpose of Section 230 is to establish a uniform federal policy, applicable across the internet, that avoids results such as the state court decision in Prodigy. The internet is the quintessential vehicle of interstate, and indeed international, commerce. Its packet-switched architecture makes it uniquely susceptible to multiple sources of conflicting state and local regulation, since even a message from one cubicle to its neighbor inside the same office can be broken up into pieces and routed via servers in different states. Were every state free to adopt its own policy concerning when an internet platform will be liable for the criminal or tortious conduct of another, not only would compliance become oppressive, but the federal policy itself could quickly be undone. All a state would have to do to defeat the federal policy would be to place platform liability laws in its criminal code. Section 230 would then become a nullity. Congress thus intended Section 230 to establish a uniform federal policy, but one that is entirely consistent with robust enforcement of state criminal and civil law.
C. The (c)(1) Immunity Protects the Kind of Editorial
Decisions Made by Publishers, Including Content Moderation
The vast majority of Section 230 cases are resolved under 230(c)(1), a provision that is more complicated than it may seem at first blush: “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.”
First, (c)(1) protects ICS providers and users—but only insofar as they are not themselves information content providers (ICPs)—i.e., not responsible, even “in part, for the creation or development of information provided through the Internet or any other interactive computer service.” Rep. Cox (R-CA), has called “in part” the two most important words in the statute. “The statute’s fundamental principle,” he explains, “is that content creators should be liable for any illegal content they create.”
Thus, online publications and social networks alike are protected against liability for user comments unless they modify those comments in a way that “contributes to the alleged illegality—such as by removing the word ‘not’ from a user’s message reading ‘[Name] did not steal the artwork’ in order to transform an innocent message into a libelous one.” But these sites are not protected for articles created by their employees or agents. By the same token, Section 230(c)(1) does not apply to warning labels added by Twitter to tweets posted by President Trump or any other user—but nor does adding that label to the tweet make Twitter responsible for the contents of the tweet: Trump remains the ICP of that tweet and Twitter remains the ICP only of the label.
In multiple cases, courts have allowed lawsuits to proceed because they found that websites were at least partially responsible for content creation. The Federal Trade Commission has won two important victories. In 2009, the Tenth Circuit ruled that Section 230 did not protect Accusearch from being sued for illegally selling private telephone records because “a service provider is ‘responsible’ for the development of offensive content only if it in some way specifically encourages development of what is offensive about the content.” The court concluded:
Accusearch was responsible for the . . . conversion of the legally protected records from confidential material to publicly exposed information. Accusearch solicited requests for such confidential information and then paid researchers to obtain it. It knowingly sought to transform virtually unknown information into a publicly available commodity. And as the district court found and the record shows, Accusearch knew that its researchers were obtaining the information through fraud or other illegality.
Similarly, in FTC v. Leadclick Media, LLC, the Second Circuit Court of Appeals allowed the FTC to proceed with its deceptive advertising claims against Leadclick. “While LeadClick did not itself create fake news sites to advertise products,” the court concluded that the company had:
[R]ecruited affiliates for the LeanSpa account that used false news sites. LeadClick paid those affiliates to advertise LeanSpa products online, knowing that false news sites were common in the industry. LeadClick employees occasionally advised affiliates to edit content on affiliate pages to avoid being “crazy [misleading],” and to make a report of alleged weight loss appear more “realistic” by reducing the number of pounds claimed to have been lost. LeadClick also purchased advertising banner space from legitimate news sites with the intent to resell it to affiliates for use on their fake news sites, thereby increasing the likelihood that a consumer would be deceived by that content.
Second, the (c)(1) immunity “precludes courts from entertaining claims that would place a computer service provider in a publisher’s role.” In its 1997 Zeran decision, the Fourth Circuit became the first appellate court to interpret this provision, concluding that “lawsuits seeking to hold a service provider liable for its exercise of a publisher’s traditional editorial functions—such as deciding whether to publish, withdraw, postpone or alter content—are barred.” Thus, courts have held that (c)(1) bars claims that an internet company unlawfully hosted a third party’s defamatory blog post; claims that a social-media platform unlawfully provided an outlet for, or algorithmically “promoted” the content of, terrorists; and claims that an Internet service provider’s failure to remove comments from a chat room amounted to breach of contract and a violation of the civil-rights laws.
D. Section 230(c)(2)(A) Protects Only Content
Moderation Decisions, and Works Differently from (c)(1)
The vast majority of Section 230 cases are resolved under 230(c)(1). Following Zeran, courts have resolved many content moderation cases under both (c)(1) and (c)(2)(A), or simply (c)(1) because the analysis is easier. The two provide somewhat overlapping, but different protections for content moderation. Subparagraph 230(c)(2)(A) says that:
No provider or user of an interactive computer service shall be held liable on account of any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected . . . .
Note how much this provision differs from (c)(1). On the one hand, (c)(2)(A)’s protections are broader: they do not depend on the ICS/ICP distinction, so (c)(2)(A) applies even when an ICS provider or user is responsible, at least in part, for developing the content at issue. On the other, its immunity is narrower, protecting only against liability only for content moderation, not liability for content actually published. Furthermore, that protection is contingent on a showing of good faith. As we shall see, much of the current policy debate over rewriting or reinterpreting Section 230 rests on the notion that only (c)(2)(A) should protect content moderation, and that the kinds of content moderation decisions it protects should be narrowed significantly.
While it is often said that Section 230 contains two immunities—(c)(1) and (c)(2)—it actually contains three, as (c)(2)(A) and (c)(2)(B) work differently and serve different functions. While (c)(2)(A) protects ICS users and providers in making content removal decisions, (c)(2)(B) protects “any action taken to enable or make available to information content providers or others the technical means to restrict access to material described in [(c)(2)(B)].” There has been relatively little litigation involving this provision, most of which involves anti-malware tools. But (c)(2)(B) also protects makers of the parental control tools built into operating systems for computers, mobile devices, televisions, gaming consoles, etc., as well as additional tools and apps that users can add to those devices. Likewise, (c)(2)(B) protects other tools offered to users, such as Social Fixer, a popular online application that allows consumers to tailor the content they receive on Facebook. With a few clicks, consumers can hide posts involving specified keywords or authors, filter out political content, or block targeted advertisements. Like parental tools, Social Fixer puts the consumer in control; in the language of Section 230(c)(2)(B), it provides the “technical means” by which the consumer chooses what fills her computer screen.
In Marshall’s Locksmith v. Google, the D.C. Circuit dismissed, under Section 230, false advertising and Sherman Act claims by locksmiths who claimed that “Google, Microsoft, and Yahoo! have conspired to ‘flood the market’ of online search results with information about so-called ‘scam’ locksmiths, in order to extract additional advertising revenue.” The case makes clear that Section 230 does bar some antitrust claims, but also merits discussion because of the potentially confusing nature of its discussion of “neutrality.”
Plaintiffs alleged that the way search engines displayed “scam” locksmith listings on their map tools alongside those for “legitimate” locksmiths “allows scam locksmiths to amplify their influence.” They argued that Section 230(c)(1) ought not apply because the maps constituted “[e]nhanced content that was derived from third-party content, but has been so augmented and altered as to have become new content and not mere editorialization . . . .”
The court disagreed. “We have previously held,” it noted, “that ‘a website does not create or develop content when it merely provides a neutral means by which third parties can post information of their own independent choosing online.’” The defendants in the case at hand, it continued, had “use[d] a neutral algorithm” to “convert third-party indicia of location into pictorial form.” The defendants’ algorithms were “neutral” in that they did “not distinguish between legitimate and scam locksmiths . . . .” The (c)(1) immunity thus defeated the locksmiths’ claims.
“Because the defendants’ map pinpoints hew to the third-party information from which they are derived,” the court did not need to “decide precisely when an entity that derives information can be considered to have ‘creat[ed] or develop[ed]’ it.” This and other lines in the opinion might give the impression that “neutrality” is a prerequisite to all forms of Section 230 immunity. But “neutrality” is usually not such a condition. To hold otherwise would nullify the many opinions holding that a defendant enjoys Section 230 immunity whenever it acts as a publisher. No publisher that exercises editorial discretion can be wholly “neutral.” As Prof. Eric Goldman notes, “The ‘neutrality’ required in this case relates only to the balance between legal and illegal content.”
Does Section 230 bar otherwise valid antitrust claims? No. Properly understood, both the (c)(1) and (c)(2)(A) immunities protect decisions to moderate or prioritize third-party content exactly inasmuch as the First Amendment itself would apply. Just as the First Amendment does not offer complete immunity to newspapers for how they handle third-party content, neither does Section 230.
A. Both (c)(1) and (c)(2)(A) Protect the Exercise of a Publisher’s
Editorial Discretion Inasmuch as It Would Be Protected by the First Amendment
The (c)(1) immunity bars only those claims that hold an ICS provider (or user) liable as the “publisher” of content provided by another—i.e., “deciding whether to publish, withdraw, postpone or alter content.” In other words, (c)(1) protects the exercise of editorial discretion protected by the First Amendment itself: “The choice of material to go into a newspaper, and the decisions made as to limitations on the size and content of the paper, and treatment of public issues and public officials—whether fair or unfair—constitute the exercise of editorial control and judgment.” “Since all speech inherently involves choices of what to say and what to leave unsaid . . . for corporations as for individuals, the choice to speak includes within it the choice of what not to say.” In Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston, the Supreme Court barred the city of Boston from forcing organizers’ of the St. Patrick’s Day parade to include pro-LGBTQ individuals, messages, or signs that conflicted with the organizer’s beliefs. The “general rule,” declared the court, is “that the speaker has the right to tailor the speech, applies not only to expressions of value, opinion, or endorsement, but equally to statements of fact the speaker would rather avoid.” Courts have recognized that social media publishers have the same rights under Miami Herald as newspapers to reject content (including ads) provided to them by third parties.
The (c)(2)(A) immunity does not explicitly turn on whether the cause of action involves the ICS provider being “treated as the publisher,” yet it clearly protects a subset of the cluster of rights protected by Miami Herald: decisions to “restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected.” Including the broad catch-all “otherwise objectionable” ensures that this immunity is fully coextensive with the First Amendment right of ICS providers and users, recognized under Hurley, to “avoid” speech they find repugnant.
As the Tenth Circuit has noted, “the First Amendment does not allow antitrust claims to be predicated solely on protected speech.” Thus, Moody’s publication of an article giving a school district’s bonds a negative rating could not give rise to an antitrust claim. In all of the cases cited by the plaintiff, “the defendant held liable on an antitrust claim engaged in speech related to its anticompetitive scheme,” but in each, that speech was incidental to the antitrust violation. Notably, in none of these cases was the “speech” at issue the exercise of editorial discretion not to publish, or otherwise associate with, the speech of others. Antitrust suits against web platforms must be grounded in economic harms to competition, not the exercise of editorial discretion. Prof. Eugene Volokh (among the nation’s top free speech scholars) explains:
it is constitutionally permissible to stop a newspaper from “forcing advertisers to boycott a competing” media outlet, when the newspaper refuses advertisements from advertisers who deal with the competitor. Lorain Journal Co. v. United States, 342 U.S. 143, 152, 155 (1951). But the newspaper in Lorain Journal Co. was not excluding advertisements because of their content, in the exercise of some editorial judgment that its own editorial content was better than the proposed advertisements. Rather, it was excluding advertisements solely because the advertisers—whatever the content of their ads—were also advertising on a competing radio station. The Lorain Journal Co. rule thus does not authorize restrictions on a speaker’s editorial judgment about what content is more valuable to its readers.
This is true even against “virtual monopolies.” As Volokh explains, the degree of a media company’s market power does not diminish the degree to which the First Amendment protects its editorial discretion.
For example, a federal court dismissed an antitrust lawsuit alleging that Facebook had attempted to eliminate an advertising competitor by blocking users of a browser extension that super-imposed its own ads onto Facebook’s website:
Facebook has a right to control its own product, and to establish the terms with which its users, application developers, and advertisers must comply in order to utilize this product. . . . Facebook has a right to protect the infrastructure it has developed, and the manner in which its website will be viewed.
The court added:
There is no fundamental right to use Facebook; users may only obtain a Facebook account upon agreement that they will comply with Facebook’s terms, which is unquestionably permissible under the antitrust laws. It follows, therefore, that Facebook is within its rights to require that its users disable certain products before using its website.
Antitrust law is so well-settled here that the court did not even need to mention the First Amendment. Instead, the court simply cited Trinko: “as a general matter, the Sherman Act ‘does not restrict the long recognized right of [a] trader or manufacturer engaged in an entirely private business, freely to exercise his own independent discretion as to parties with whom he will deal.’” That “discretion” closely parallels the First Amendment right of publishers to exclude content they find objectionable under Miami Herald.
Just as The Lorain Journal was not acting “as a publisher” when it refused to host ads from advertisers that did not boycott its radio competitor, it was also not acting “in good faith.” In this sense, the two concepts serve essentially the same function as limiting features. Each ensures that the immunity to which it applies does not protect conduct that would not be protected by the First Amendment. Thus, either immunity could be defeated by showing that the actions at issue were not matters of editorial discretion, but of anti-competitive conduct.
Epic Games recently began offering users of Fortnite, its wildly popular gaming app, the ability to make in-game purchases directly from Fortnite’s website, thus avoiding the 30% fee charged for all purchases made through Apple’s App Store. Because this violated Apple’s longstanding terms of service for the App Store, Apple banned the game. Fortnite responded by filing an antitrust suit. Apple has not raised Section 230 as a defense, and for good reason: Apple may or may not have violated the antitrust laws, but it was clearly not operating as a “publisher” when it enforced rules intended to protect a key part of its business model.
By contrast, if Facebook refused to carry ads from publishers that also bought ads on one of Google’s ad networks, the situation would be exactly the same as in Lorain Journal. Facebook’s potential liability would not be because the site was being “treated as the publisher” of ads it refused to run. The same rule should hold in other cases, even though they may be more complicated.
While Google’s AdSense network may appear to be very different from Facebook—Facebook is clearly the “publisher” of its own site while AdSense merely fills ad space on webpages published by others—AdSense is no less an ICP than is Facebook. Thus, Google can claim Section 230 immunity for decisions it makes to limit participation in AdSense (i.e., on which sites AdSense ads appear). Still, Google would be no more protected by Section 230 than The Lorrain Journal was by the First Amendment if Google refused to provide display ads via its AdSense network on third party websites simply because those sites also showed display ads from other ad networks.
A somewhat more complicated case is presented by the European Commission’s 2016 complaint against Google for how the company handled exclusivity in a different product: AdSense for Search, which displays ads alongside search results in the custom search engines Google makes available for third party websites to embed on their sites (generally to search the site’s contents). (The Commission fined Google €1.5 billion in 2019.) The exclusivity clauses Google imposed from 2006 to 2009 seem little different from Lorrain Journal or the AdSense hypothetical discussed above. But starting in 2009, Google began replacing exclusivity clauses with “Premium Placement” clauses that worked differently. These clauses required “third parties to take a minimum number of search ads from Google and reserve the most prominent space on their search results pages to Google search ads.”  Again, it is hard to see how Google could convince an American court that excluding participation in AdSense for Search based on these restrictions would be an exercise in a publisher’s editorial judgment—and a plaintiff would have a good argument that these requirements were not imposed “in good faith.” (Whether or not such arguments would prevail under U.S. antitrust law would, of course, be an entirely separate question.) But Google’s requirement that “competing search ads cannot be placed above or next to Google search ads” is at least a harder question. Google would have a much stronger argument that this requirement prevented users from mistaking ads provided by third parties from its own ads. Google might object to those third-party ads being shown right next to its own search results not only because their (potentially) lower quality could diminish user expectations of the value of Google’s search ads across the board, including on Google.com, but also because Google has no editorial control over those ads, which might include content that violates Google’s terms of service.
This case would be essentially the same as Google’s refusal to allow participation in its AdSense network by webpages that display content that violates Google’s terms of service—that Google ads do not appear next to hate speech, misinformation, etc. Many large publications have found that Google will refuse to run ads on specific pages because content in the user comments section violated its terms of service. When the scale of such content has become large enough, Google has refused to run ads on the site altogether unless the site separates the comments section from the article, and runs ads only on the article. Some sites, like Boing Boing, have done just that, but as we shall see, Google’s enforcement of this policy recently sparked a political uproar.
Essentially the same analysis would apply over blocking apps from the app store: Apple, Google, Amazon and Microsoft all block pornography apps from their stores. All have banned Gab, the “free speech” social network where little but pornography is moderated, from offering an app in their stores. Gab may, in some sense, be a competitor to these companies, but each would have an easy argument that they were acting as “publishers,” and in “good faith,” when they banned the Gab app.
Now consider competition claims that are even more clearly about the exercise of editorial discretion—claims regarding blocking specific users or publications. When a plaintiff brings both a general antitrust claim and a specific claim about content removal, Section 230 bars the latter but not the former. In Brittain v. Twitter, for example, Craig Brittain sued Twitter for suspending four of his accounts. The court dismissed all claims under Section 230, save an antitrust claim that did not “arise out of Twitter’s deletion of the Brittain Accounts.” That claim, which attacked Twitter’s alleged “monopolization” of the “social media networking market,” was assessed separately under the antitrust laws. (It failed.)
But at least in one case, a plaintiff clearly alleged that account suspension violated a state unfair competition law. The case offers the most thorough explanation of why Section 230 bars such claims about “unfair” or “biased” exercises of editorial discretion. When Jared Taylor, a leading white supremacist, first joined Twitter in 2011, the site did not bar (indeed, it all but pledged not to censor) extremist content. But the site reserved the right to revise its terms of service, and in 2017 it “announced ‘updated . . . rules’” on hateful and violent content. These included a requirement that users “not affiliate with organizations that . . . use or promote violence against civilians to further their causes.” Acting under these rules, Twitter banned Taylor and his publication. Taylor then sued Twitter for blocking him (and his white-nationalist publication American Renaissance) from the site.
Although it dismissed most of Taylor’s lawsuit under Section 230, the trial court let an unfair-competition claim proceed. The California Court of Appeal reversed, holding that Section 230 barred all of Taylor’s claims. The core question, the court observed, is whether the plaintiff’s claim attacks decisions the defendant made while acting as a publisher. Such publishing decisions, the court continued, include decisions to “restrict or make available certain material.” Thus, a defendant’s decision to allow a user to have, or to bar a user from having, a social-media account falls squarely within the Section 230 immunity. “Any activity,” the court concluded, “that can be boiled down to deciding whether to exclude material that third parties seek to post online is perforce immune under section 230.” The court’s full analysis offers an excellent summary of current case law.
Conceding error on remand, the trial judge wrote: “I now realize that . . . [i]n the evaluation of plaintiffs’ UCL claim, framing the salient question as whether Twitter is being treated as a publisher[,] or [as the party] whose speech is the basis for the claim, is a single indivisible question.” In other words, as we have argued, being treated “as a publisher” means being held liable on the basis of the exercise of free speech rights, including the editorial decision to remove objectionable content.
D. Both the First Amendment and Section 230 Protect the Right of Ad Platforms to Disassociate from Content They Find Objectionable
Section 230(c)(1) protects a website operator’s right to “demonetize” third-party content creators who upload content to the site—i.e., leaving their content up, but declining to run ads next to it. Thus, for example, a court dismissed a suit alleging violations of the Sherman and Lanham acts after YouTube “demonetized the ‘Misandry Today’ channel (which purported to expose anti-male bias).” The court equated removal of content and demonetization: “Both fall under the rubric of publishing activities.”
The same result applies when publishers of any interactive computer service refuse to allow their content to appear on third party pages—whether that content is an ad or other embeddable function. Consider the most controversial case of “demonetization” by an advertising platform: “NBC News attempted,” the co-founders of the rightwing website The Federalist wrote in June 2020, “to use the power of Google to cancel our publication.” After NBC published an exposé about racist comments posted by readers of The Federalist on its articles, Google told the site that it would no longer run ads on the site unless the site moderated content that violated the AdSense Terms of Service. The Federalist never filed a lawsuit—and for good reason: both the First Amendment and Section 230 would have blocked any antitrust (or other) suit.
The Federalist received, not a dose of politically motivated punishment, but a common lesson in the difficulty of content moderation. Websites across the political spectrum have had their content flagged in the same way. Boing Boing (a tech news site) handled the situation by putting user comments for each article (with Google ads) on a separate page (without ads). TechDirt (a leading site about technology that no one would ever accuse of being “right wing”) kept user comments on the same page as each article.
The difficulties of content moderation cannot, constitutionally, be solved by the NTIA Petition or any other government action. Writing about his site’s experience with these issues, TechDirt’s editor Mike Masnick explained:
[T]here are currently no third-party ads on Techdirt. We pulled them down late last week, after it became impossible to keep them on the site, thanks to some content moderation choices by Google. In some ways, this is yet another example of the impossibility of content moderation at scale . . . .
The truth is that Google’s AdSense (its third-party ad platform) content moderation just sucks. In those earlier posts about The Federalist’s situation, we mentioned that tons of websites deal with those “policy violation” notices from Google all the time. Two weeks ago, it went into overdrive for us: we started receiving policy violation notices at least once a day, and frequently multiple times per day. Every time, the message was the same, telling us we had violated their policies (they don’t say which ones) and we had to log in to our “AdSense Policy Center” to find out what the problem was. Every day for the ensuing week and a half (until we pulled the ads down), we would get more of these notices, and every time we’d log in to the Policy Center, we’d get an ever rotating list of “violations.” But there was never much info to explain what the violation was. Sometimes it was “URL not found” (which seems to say more about AdSense’s shit crawler than us). Sometimes it was “dangerous and derogatory content.” Sometimes it was “shocking content.”
Other large publications have experienced similar problems. Consider Slate (generally considered notably left-of-center), whose experience illustrates the limits of algorithmic content moderation as a way of handling the scale problem of online content:
Last Thursday, Google informed Slate’s advertising operations team that 10 articles on the site had been demonetized for containing “dangerous or derogatory content.” The articles in question covered subjects like white supremacy, slavery, and hate groups, and most of them quoted racial slurs. They included pieces on the racist origins of the name kaffir lime, the 2017 police brutality movie Detroit, Joe Biden’s 1972 Senate run, and a Twitter campaign aimed at defaming Black feminists, which all had quotes containing the N-word . . . .
Needless to say, the articles were not promoting the discriminatory ideologies affiliated with these slurs but rather reporting on and analyzing the context in which they were used.
Once flagged by the algorithm, the pages were not eligible to earn revenue through Ad Exchange. Slate appealed the moderation decisions through Google’s ad platform last Thursday morning, as it normally would when a demonetization it feels is unjustified occurs. Not long after, as part of the reporting of this story, I contacted Google’s communications department, whose personnel said they would contact the engineering team to look into it. The pages were subsequently remonetized by Friday morning.
So, yes, the tool Google uses to decide whether it wants to run its ads next to potentially objectionable content is highly imperfect. Yes, everyone can agree that it would be better if this tool could distinguish between discussions of racism and racism itself—and exactly the same thing can be said of the algorithms Facebook and (to a lesser degree) Twitter apply to moderate user content on their sites. But these simply are not problems for the government to solve. The First Amendment requires us to accept that the exercise of editorial discretion will always be messy. This is even more true of digital media than it is of traditional media, as Masnick’s Impossibility Theorem recognizes: “Content moderation at scale is impossible to do well. More specifically, it will always end up frustrating very large segments of the population and will always fail to accurately represent the ‘proper’ level of moderation of anyone.”
Google has a clear First Amendment right not to run its ads next to content it finds objectionable. If websites want to run Google ads next to their comments section, they have a contractual obligation to remove content that violates Google’s terms of service for their advertising platform. Simply relying on users to downrank objectionable content, rather than removing it altogether—as The Federalist apparently did—will not suffice if the objectionable content remains visible on pages where Google Ads appear. It is worth noting that, while TechDirt might reasonably expect its readers to downrank obviously racist content, the exact opposite has happened on The Federalist:
(These comments use the term “jogger” as code for the n-word, in reference to the racially motivated murder of “Ahmaud Arbery, the young black man who was shot dead in February while he was running in a suburban neighborhood in Georgia.”)
Given the inherent challenges of content moderation at scale—and given how little attention most comment sections receive—it would be unfair to generalize about either The Federalist or its overall readership based on such examples. Nonetheless, Google has a right—guaranteed by the First Amendment—to insist that its ads not appear next to, or even anywhere on a site that hosts, a single such exchange. In exercising this right, Google also protects the right of advertisers not to be associated with racism and other forms of content they, in their sole discretion, deem objectionable. In making such decisions, Google is exercising its editorial discretion of AdSense, which clearly qualifies as an ICS. An advertising broker, after all, acts as a publisher no less than a newspaper does. Both a newspaper and an ad broker decide where content goes and how it is presented; the one just does this with stories, the other with ads. Thus, Section 230(c)(1) protects Google’s right to exercise editorial discretion in how it places ads just as it protects Google’s right to exercise such discretion over what content appears on YouTube.
Even more than antitrust, Republicans have also insisted that consumer protection law should be used to police “bias.” Multiple Republican legislative proposals would use existing consumer protection law to enforce promises (allegedly) made by social media companies to remain “neutral” in content moderation and other decisions about configuring the user experience. The Executive Order contemplates the FTC and state attorneys general using consumer protection law to declare unfair or deceptive “practices by entities covered by section 230 that restrict speech in ways that do not align with those entities’ public representations about those practices.” The NTIA Petition to the FCC argues:
if interactive computer services’ contractual representations about their own services cannot be enforced, interactive computer services cannot distinguish themselves. Consumers will not believe, nor should they believe, representations about online services. Thus, no service can credibly claim to offer different services, further strengthening entry barriers and exacerbating competition concerns.
The underlying premise of this argument false. As explained below, consumer protection law cannot enforce such claims because, to be enforceable, claims must be specific, objectively verifiable, and related to commercial transactions, not social policy. These requirements are ultimately rooted in the First Amendment itself. The category of “representations” about content moderation that could, perhaps, be enforced in court would be accordingly narrow.
Perhaps recognizing that the statements of neutrality they point to (often made in Congressional hearings in response to badgering by Republican lawmakers) will not give rise to liability today, NTIA proposes to require social media providers, as a condition of 230(c)(2)(A) protection for their content moderation practices, to “state plainly and with particularity the criteria the interactive computer service employs in its content-moderation practices.” (NTIA also asks the FCC to reinterpret Section 230 so that (c)(1) no longer protects content moderation.) Furthermore, NTIA would require each ICS provider to commit not to “restrict access to or availability of material on deceptive or pretextual grounds” and to “apply its terms of service or use to restrict access to or availability of material that is similarly situated to material that the interactive computer service intentionally declines to restrict.” Similar concepts can be found in multiple Republican bills. Such transparency mandates will also fail under the First Amendment.
Republicans now find themselves in the awkward position of advocating something like the FCC’s old Fairness Doctrine, under which broadcasters were required to ensure that all “positions taken by responsible groups” received airtime. Republicans fiercely opposed the Fairness Doctrine long after President Reagan effectively ended it in 1987—all the way through the 2016 Republican platform.
A. Community Standards Are Non-Commercial Speech,
which Cannot Be Regulated via Consumer Protection Law
The Federal Trade Commission has carefully grounded its deception authority in the distinction long drawn by the Supreme Court between commercial and non-commercial speech in Central Hudson Gas Elec. v. Public Serv. Comm’n, 447 U.S. 557 (1980) and its progeny. Commercial speech is speech which “[does] no more than propose a commercial transaction.” In Pittsburgh Press Co. v. Human Rel. Comm’n, the Supreme Court upheld a local ban on referring to sex in the headings for employment ads. In ruling that the ads at issue were not non-commercial speech (which would have been fully protected by the First Amendment), it noted: “None expresses a position on whether, as a matter of social policy, certain positions ought to be filled by members of one or the other sex, nor does any of them criticize the Ordinance or the Commission’s enforcement practices.” In other words, a central feature of commercial speech is that it is “devoid of expressions of opinions with respect to issues of social policy.” This is the distinction FTC Chairman Joe Simons was referring to when he told lawmakers this summer that the issue of social media censorship is outside the FTC’s remit because “our authority focuses on commercial speech, not political content curation.”
While “terms of service” for websites in general might count as commercial speech, the kind of statement made in “community standards” clearly “expresses a position on . . . matter[s] of social policy.” Consider just a few such statements from Twitter’s “rules”:
Violence: You may not threaten violence against an individual or a group of people. We also prohibit the glorification of violence. Learn more about our violent threat and glorification of violence policies.
Terrorism/violent extremism: You may not threaten or promote terrorism or violent extremism. . . .
Abuse/harassment: You may not engage in the targeted harassment of someone, or incite other people to do so. This includes wishing or hoping that someone experiences physical harm.
Hateful conduct: You may not promote violence against, threaten, or harass other people on the basis of race, ethnicity, national origin, caste, sexual orientation, gender, gender identity, religious affiliation, age, disability, or serious disease.
Each of these statements clearly “expresses a position on . . . a matter of social policy,” and therefore is clearly non-commercial speech that merits the full protection of the First Amendment under the exacting standards of strict scrutiny. “If there is any fixed star in our constitutional constellation, it is that no official, high or petty, can prescribe what shall be orthodox in politics, nationalism, religion, or other matters of opinion or force citizens to confess by word or act their faith therein.” Consumer protection law simply does not apply to such claims—because of the First Amendment’s protections for non-commercial speech.
B. The First Amendment Does Not Permit Social Media
Providers to Be Sued for “Violating” their Current Terms of Service, Community Standards, or Other Statements About Content Moderation
Given the strong protection a website’s terms of service enjoy under the First Amendment, it is no surprise that attempts to enforce statements about political neutrality via consumer protection law have failed—and that no American consumer protection agency has actually pursued such claims in an enforcement action.
Understanding why such claims are unenforceable starts in the rhetorical hotbed that is cable news. In 2004, MoveOn.org and Common Cause asked the FTC to proscribe Fox News’ use of the slogan “Fair and Balanced” as a deceptive trade practice. The two groups acknowledged that Fox News had “no obligation whatsoever, under any law, actually to present a ‘fair’ or ‘balanced’ presentation of the news,” but argued: “What Fox News is not free to do, however, is to advertise its news programming—a service it offers to consumers in competition with other networks, both broadcast and cable—in a manner that is blatantly and grossly false and misleading.” FTC Chairman Tim Muris (a Bush appointee) responded pithily: “I am not aware of any instance in which the [FTC] has investigated the slogan of a news organization. There is no way to evaluate this petition without evaluating the content of the news at issue. That is a task the First Amendment leaves to the American people, not a government agency.”
Deception claims always involve comparing marketing claims against conduct. Muris meant that, in this case, the nature of the claims (general claims of fairness) meant that their accuracy could not be assessed without the FTC sitting in judgment of how Fox News exercised its editorial discretion. The “Fair and Balanced” claim was not, otherwise, verifiable—which is to say that it was not objectively verifiable.
Prager “University,” the leading creator of “conservative” video content, attempted to make essentially the same deceptive marketing claim against YouTube in a case dismissed by the Ninth Circuit. Despite having over 2.52 million subscribers and more than a billion views of “5-minute videos on things ranging from history and economics to science and happiness,” PragerU sued YouTube for “unlawfully censoring its educational videos and discriminating against its right to freedom of speech.” Specifically, PragerU alleged that roughly a sixth of the site’s videos had been flagged for YouTube’s Restricted Mode, an opt-in feature that allows parents, schools, and libraries to restrict access to potentially sensitive content (and is turned on by fewer than 1.5% of YouTube users). After dismissing PragerU’s claims that YouTube was a state actor denied First Amendment protection, the Ninth Circuit ruled:
YouTube’s braggadocio about its commitment to free speech constitutes opinions that are not subject to the Lanham Act. Lofty but vague statements like “everyone deserves to have a voice, and that the world is a better place when we listen, share and build community through our stories” or that YouTube believes that “people should be able to speak freely, share opinions, foster open dialogue, and that creative freedom leads to new voices, formats and possibilities” are classic, non-actionable opinions or puffery. Similarly, YouTube’s statements that the platform will “help [one] grow,” “discover what works best,” and “giv[e] [one] tools, insights and best practices” for using YouTube’s products are impervious to being “quantifiable,” and thus are non-actionable “puffery.” The district court correctly dismissed the Lanham Act claim.
Roughly similar to the FTC’s deception authority, the Lanham Act requires proof that (1) a provider of goods or services made a “false or misleading representation of fact,” which (2) is “likely to cause confusion” or deceive the general public about the product. Puffery fails both requirements because it “is not a specific and measurable claim, capable of being proved false or of being reasonably interpreted as a statement of objective fact.” The FTC’s bedrock 1983 Deception Policy Statement declares that the “Commission generally will not pursue cases involving obviously exaggerated or puffing representations, i.e., those that the ordinary consumers do not take seriously.”
There is simply no way social media services can be sued under either the FTC Act (or state baby FTC acts) or the Lanham Act for claims they make today about their content moderation practices. Twitter CEO Jack Dorsey said this in Congressional testimony in 2018: “Twitter does not use political ideology to make any decisions, whether related to ranking content on our service or how we enforce our rules.” This claim is no less “impervious to being ‘quantifiable’” than YouTube’s claims.
Moreover, “[i]n determining the meaning of an advertisement, a piece of promotional material or a sales presentation, the important criterion is the net impression that it is likely to make on the general populace.” Thus, isolated statements about neutrality or political bias (e.g., in Congressional testimony) must be considered in the context of the other statements companies make in their community standards, which broadly reserve discretion to remove content or users. Furthermore, the FTC would have to establish the materiality of claims, i.e., that an “act or practice is likely to affect the consumer’s conduct or decision with regard to a product or service. If so, the practice is material, and consumer injury is likely, because consumers are likely to have chosen differently but for the deception.” In the case of statements made in Congressional testimony or in any other format besides a traditional advertisement, the Commission could not simply presume that the statement was material. Instead, the Commission would have to prove that consumers would have acted differently but for the deception. These requirements—objective verifiability (about facts, rather than opinions) and materiality—are not simply doctrinal formalities of current consumer protection law that could be bypassed by new legislation. Instead, they reflect the outer boundaries of how far the government may go in regulating speech without violating the First Amendment. In short, the First Amendment allows the government to enforce marketing claims, but not to police the exercise of editorial discretion.
C. The First Amendment Will Not Permit the Government to Compel More Specific Claims
Because current claims by social media companies about their editorial practices are “impervious to being ‘quantifiable,’ and thus are non-actionable ‘puffery,’” the NTIA Petition and several Republican bills propose to require more specific statements. But does anyone seriously believe that the First Amendment would—whether through direct mandate or as the condition of tax exemption, subsidy, legal immunity, or some other benefit—permit the government to require book publishers to “state plainly and with particularity the criteria” (as NTIA proposes) by which they decide which books to publish, or newspapers to explain how they screen letters to the editor, or cable TV news shows to explain which guests they book—let alone enforce adherence to those criteria the exercise of editorial discretion?
Startlingly, NTIA professes to see nothing unconstitutional about such a quid pro quo, at least when quo is legal immunity:
To the contrary, the Supreme Court has upheld the constitutionality of offers special liability protections in exchange for mandated speech. In Farmer’s Union v. WDAY, the Court held that when the federal government mandates equal time requirement for political candidates—a requirement still in effect, this requirement negates state law holding station liable for defamation for statements made during the mandated period. In other words, the Court upheld federal compelled speech in exchange for liability protections. Section 230’s liability protections, which are carefully drawn but come nowhere near to compelling speech, are just as constitutionally unproblematic if not more so.
The problem with this argument should be obvious: the 1959 Supreme Court decision cited by NTIA involved only broadcasters, which do not enjoy the full protection of the First Amendment. Thus, this argument is no more applicable to non-broadcast media than is the Supreme Court’s 1969 decision Red Lion upholding the Fairness Doctrine for broadcasters. Yet the NTIA’s “plain and particular criteria” mandate goes far beyond even what the Fairness Doctrine required. Such disclosure mandates offend the First Amendment for at least three reasons. First, community standards and terms of service are themselves non-commercial speech. Deciding how to craft them is a form of editorial discretion protected by the First Amendment, and forcing changes in how they are written is itself a form of compelled speech—no different from forcing a social media company’s other statements about conduct it finds objectionable on, or off, its platform. “Mandating speech that a speaker would not otherwise make necessarily alters the content of the speech.” The Court said that in a ruling striking down a North Carolina statute that required professional fundraisers for charities to disclose to potential donors the gross percentage of revenues retained in prior charitable solicitations. The Court declared that the “the First Amendment guarantees ‘freedom of speech,’ a term necessarily comprising the decision of both what to say and what not to say.” The Court will not allow compelled disclosure for something that is relatively objective and quantifiable, how could it allow compelled disclosure of something far more subjective?
Second, forcing a social media site to attempt to articulate all of the criteria for its content moderation practices while also requiring those criteria to be as specific as possible will necessarily constrain what is permitted in the underlying exercise of editorial discretion. Community standards and terms of service are necessarily overly reductive; they cannot possibly anticipate every scenario. If the Internet has proven anything, it is that there is simply no limit to human creativity in finding ways to be offensive in what we say and do in interacting with other human beings online. It is impossible to codify “plainly and with particularity” all of the reasons why online content and conduct may undermine Twitter’s mission to “serve the public conversation.” Thus, compelling disclosure of decision-making criteria and limiting media providers to them in making content moderation decisions will necessary compel speech on a second level: forcing the website to carry speech it would prefer not to carry.
Third, even if the government argued that the criteria it seeks to compel social media providers to disclose are statements of fact (about how they conduct content moderation) rather than statements of opinion, the Court has explicitly rejected such a distinction. Citing cases in which the court had struck down compelled speech requirements, such as displaying the slogan “Live Free or Die” on a license plate, the Court noted:
These cases cannot be distinguished simply because they involved compelled statements of opinion while here we deal with compelled statements of “fact”: either form of compulsion burdens protected speech. Thus, we would not immunize a law requiring a speaker favoring a particular government project to state at the outset of every address the average cost overruns in similar projects, or a law requiring a speaker favoring an incumbent candidate to state during every solicitation that candidate’s recent travel budget. Although the foregoing factual information might be relevant to the listener, and, in the latter case, could encourage or discourage the listener from making a political donation, a law compelling its disclosure would clearly and substantially burden the protected speech.
The same is true here: the First Amendment protects Twitter’s right to be as specific, or as vague, as it wants in defining what constitutes “harassment,” “hateful conduct,” “violent threats,” “glorification of violence,” etc.
V. Even Rolling Back Section 230’s Immunities
Would Not Create Liability under Antitrust or Consumer Protection Law Based on the Exercise of Editorial Discretion
Current law is clear: The First Amendment bars liability under antitrust or other competition laws for content moderation. Section 230, both (c)(1) and (c)(2)(A), provide a procedural short-cut that allows defendants to avoid having to litigate cases that would ultimately fail anyway under the First Amendment—just as anti-SLAPP laws do in certain defamation cases. This Administration has tried to sidestep current law in two ways.
In a 2017 opinion, Justice Anthony Kennedy called social media the “modern public square” and declared that “to foreclose access to social media altogether is to prevent the user from engaging in the legitimate exercise of First Amendment rights.” Many, across the political spectrum, have cited this passage to insist that their proposed regulations would actually vindicate the First Amendment rights of Internet users.
For example, the NTIA petition breezily asserts that “social media and other online platforms . . . function, as the Supreme Court recognized, as a 21st century equivalent of the public square.” NTIA cites the Supreme Court’s recent Packingham decision: “Social media . . . are the principal sources for knowing current events, checking ads for employment, speaking and listening in the modern public square, and otherwise exploring the vast realms of human thought and knowledge.” The Executive Order goes even further: “Communication through these channels has become important for meaningful participation in American democracy, including to petition elected leaders. These sites are providing an important forum to the public for others to engage in free expression and debate. Cf. PruneYard Shopping Center v. Robins, 447 U.S. 74, 85-89 (1980).” The Executive Order suggests that the First Amendment should constrain, rather than protect, the editorial discretion of social media operators because social media are de facto government actors.
Both the Order and the Petition omit a critical legal detail about Packingham: it involved a state law restricting the Internet use of convicted sex offenders. Justice Kennedy’s simile that social media is “a 21st century equivalent of the public square” merely conveys the gravity of the deprivation of free speech rights effected by the state law. Packingham says nothing whatsoever to suggest that private media companies become de facto state actors by virtue of providing that “public square.” On the contrary, in his concurrence, Justice Alito expressed dissatisfaction with the “undisciplined dicta” in the majority’s opinion and asked his colleagues to “be more attentive to the implications of its rhetoric” likening the Internet to public parks and streets.
The Executive Order relies on the Supreme Court’s 1980 decision in Pruneyard Shopping Center v. Robins, treating shopping malls as public fora under California’s constitution. NTIA makes essentially the same argument, by misquoting Packingham, even without directly citing Pruneyard. NTIA had good reason not to cite the case: it is clearly inapplicable, stands on shaky legal foundations on its own terms, and is antithetical to longstanding conservative positions regarding private property and the First Amendment. In any event, Pruneyard involved shopping malls (for whom speech exercised on their grounds was both incidental and unwelcome), not companies for which the exercise of editorial discretion lay at the center of their business. Pruneyard has never been applied to a media company, traditional or new. The Supreme Court ruled on a very narrow set of facts and said that states have general power to regulate property for certain free speech activities. The Supreme Court, however, has not applied the decision more broadly, and lower courts have rejected Pruneyard’s application to social media. Social media companies are in the speech business, unlike businesses which incidentally host the speech of others.
In a line of cases following Miami Herald, the Supreme Court consistently upheld the First Amendment right of media outlets other than broadcasters (a special case discussed below). In Reno v. American Civil Liberties Union, 521 U.S. 844, 870 (1997), the Court made clear that, unlike broadcasters, digital media operators enjoy the same protections in exercising their editorial discretion as newspapers:
some of our cases have recognized special justifications for regulation of the broadcast media that are not applicable to other speakers . . . Those factors are not present in cyberspace. Neither before nor after the enactment of the CDA have the vast democratic forums of the Internet been subject to the type of government supervision and regulation that has attended the broadcast industry. Moreover, the Internet is not as “invasive” as radio or television.
Since Reno, the Court has steadfastly refused to create such carveouts for new media. While striking down a state law restricting the purchase of violent video games, Justice Scalia declared: “whatever the challenges of applying the Constitution to ever-advancing technology, the basic principles of freedom of speech and the press, like the First Amendment’s command, do not vary when a new and different medium for communication appears.”
In general, media companies do not qualify as state actors merely because they provide “platforms” for others’ speech. A private entity may be considered a state actor when the entity exercises a function “traditionally exclusively reserved to the State.” In a 2019 case, Manhattan v. Halleck, the Supreme Court held that “operation of public access channels on a cable system is not a traditional, exclusive public function.” “Under the Court’s cases, those functions include, for example, running elections and operating a company town,” but not “running sports associations and leagues, administering insurance payments, operating nursing homes, providing special education, representing indigent criminal defendants, resolving private disputes, and supplying electricity.” Justice Kavanaugh, writing for the five conservatives Justices, concluded the majority opinion as follows: “merely hosting speech by others is not a traditional, exclusive public function and does not alone transform private entities into state actors subject to First Amendment constraints.” While Halleck did not involve digital media, the majority flatly rejected the argument made by the Executive Order for treating digital media as public fora. The lower courts have “uniformly concluded that digital internet platforms that open their property to user-generated content do not become state actors.”
The Executive Order, the NTIA Petition, and a flurry of Congressional proposals argue that courts have misinterpreted Section 230. We dissected these claims in detail in our comments and reply comments in the FCC proceeding, which provide a detailed analysis for anyone interested in the merits of their arguments (or complete lack thereof). Below follows an abbreviated explanation.
The Executive Order, NTIA’s Petition and multiple Republican bills propose to disqualify social media completely for the (c)(1) immunity by redefining the line between an “interactive computer service”—the providers or users of which are covered by (c)(1)—and an “information content provider,” which are never protected by (c)(1). Specifically, NTIA proposes the following definition: “‘responsible, in whole or in part, for the creation or development of information’ includes substantively contributing to, modifying, altering, presenting or prioritizing with a reasonably discernible viewpoint, commenting upon, or editorializing about content provided by another information content provider.” Parts of this definition are uncontroversial: Section 230 has never applied to content that a website creates itself, so, yes, “adding special responses or warnings [to user content] appear to develop and create content in any normal use of the words.” There is simply no confusion in the courts about this. Similarly, “modifying” or “altering” user content may not be covered today, as the Ninth Circuit explained in Roommates:
A website operator who edits user-created content—such as by correcting spelling, removing obscenity or trimming for length—retains his immunity for any illegality in the user-created content, provided that the edits are unrelated to the illegality. However, a website operator who edits in a manner that contributes to the alleged illegality—such as by removing the word “not” from a user’s message reading “[Name] did not steal the artwork” in order to transform an innocent message into a libelous one—is directly involved in the alleged illegality and thus not immune.
But then the Petition veers off into radically reshaping current law when it claims that “prioritization of content under a variety of techniques, particularly when it appears to reflect a particularly [sic] viewpoint, might render an entire platform a vehicle for expression and thus an information content provider.”
Once again, NTIA is trying to place the exercise of editorial discretion beyond the protection of (c)(1), despite the plain language of that provision preventing ICS providers from being held liable “as the publisher” of third party content. What the Supreme Court said in Miami Herald is no less true of website operators: “The choice of material to go into a newspaper, and the decisions made as to limitations on the size and content of the paper, and treatment of public issues and public officials—whether fair or unfair—constitute the exercise of editorial control and judgment.” As the Ninth Circuit has noted, “the exclusion of ‘publisher’ liability necessarily precludes liability for exercising the usual prerogative of publishers to choose among proffered material and to edit the material published while retaining its basic form and message.” NTIA is proposing a legal standard by which the government will punish digital media publishers for exercising that prerogative in ways this administration finds objectionable. While NTIA claims to advocate neutrality, its proposed regulation would be anything but content-neutral: it would allow certain community standards (such as against pornography) but disallow others (such against racism and misinformation).
Because they would regulate speech “based on the message a speaker conveys”—and, in particular, based on the government’s “disagreement with the [speaker’s] message”—the NTIA’s attempts to limit which grounds for content moderation qualify for Section 230 protections are content-based restrictions subject to strict scrutiny. Strict scrutiny “requires the Government to prove that the restriction furthers a compelling interest and is narrowly tailored to achieve that interest.” The NTIA’s proposals, which stand on no more than the government’s disagreement with private platforms’ editorial decisions—and, more, with the platforms’ perceived political leanings—would clearly fail.
And make no mistake—true “content neutrality,” devoid of any “prioritization of content,” is not a desirable outcome. In the context of government speech, the Supreme Court has noted that neutrality would require “every jurisdiction that has accepted a donated war memorial . . . to provide equal treatment for a donated monument questioning the cause for which the veterans fought.” Likewise, a “neutral” Internet platform would have to treat harassing, hateful, or violent messages like all other messages. Platforms forced to act as “neutral” public fora would become miserable, off-putting places, driving away users. The platforms’ very existence would be put at risk.
NTIA claims that that Zeran misread the statute, and insists that “Section 230(c)(1) applies to acts of omission—to a platform’s failure to remove certain content. In contrast, section 230(c)(2) applies to acts of commission—a platform’s decisions to remove. Section 230(c)(1) does not give complete immunity to all a platform’s ‘editorial judgments.’” Multiple Republican bills would amend Section 230 to reflect this view.
NTIA argues that courts have read “section 230(c)(1) in an expansive way that risks rendering (c)(2) a nullity.” The Petition claims interpreting Paragraph (c)(1) to cover decisions to remove content (as well as to host content) violates the statutory canon against surplusage because it renders (c)(2) superfluous. The plain text of the statute makes clear why this is not the case. Neither subparagraph of (c)(2) is rendered a “nullity” by the essentially uniform consensus of courts that Paragraph (c)(1) covers decisions to remove user content just as it covers decisions to leave user content up. The Ninth Circuit has already explained what work Subparagraph (c)(2)(A) does that Subsection (c)(1) does not:
Subsection (c)(1), by itself, shields from liability all publication decisions, whether to edit, to remove, or to post, with respect to content generated entirely by third parties. Subsection (c)(2), for its part, provides an additional shield from liability, but only for “any action voluntarily taken in good faith to restrict access to or availability of material that the provider . . . considers to be obscene . . . or otherwise objectionable.” § 230(c)(2)(A). Crucially, the persons who can take advantage of this liability are not merely those whom subsection (c)(1) already protects, but any provider of an interactive computer service. See § 230(c)(2). Thus, even those who cannot take advantage of subsection (c)(1), perhaps because they developed, even in part, the content at issue, see Roommates.Com, 521 F.3d at 1162-63, can take advantage of subsection (c)(2) if they act to restrict access to the content because they consider it obscene or otherwise objectionable. Additionally, subsection (c)(2) also protects internet service providers from liability not for publishing or speaking, but rather for actions taken to restrict access to obscene or otherwise objectionable content.
The (c)(2)(A) immunity ensures that, even if an ICS provider is shown to be partially responsible for content creation, its decision to remove content will not be grounds for liability (absent a lack of “good faith”). This belt-and-suspenders approach is crucial to serving the statute’s central purpose—removing disincentives against content moderation—because certain forms of content moderation may at least open the door for plaintiffs to argue that the ICS provider had become responsible for the content, and thus subject them to the cost of litigating that question at a motion to dismiss or the even greater cost of litigating past a motion to dismiss if the trial judge rules that they may have been responsible for the creation of that content. Discovery costs alone have been estimated to account as much as 90% of litigation costs.
The NTIA Petition argues that “the plain words of [(c)(2)(A)] indicate that this protection only covers decisions to restrict access to certain types of enumerated content. . . . [T]hese categories are quite limited and refer primarily to traditional areas of media regulation—also consistent with legislative history’s concern that private regulation could create family-friendly internet spaces.” The Petition makes two arguments to support this assertion—both of which have become talking points for Republican legislators proposing to amend Section 230 in the same ways NTIA seeks to have the FCC reinterpret the statute.
First, NTIA argues: “If ‘otherwise objectionable’ means any material that any platform ‘considers’ objectionable, then section 230(b)(2) offers de facto immunity to all decisions to censor content.” NTIA is clearly referring to the wrong statutory provision here; it clearly mean 230(c)(2)—yet “makes this same erroneous substitution on page 28, so it wasn’t just a slip of the fingers.” NTIA fails to understand how the (c)(2)(A) immunity works. This provision contains two distinct operative elements: (1) the nature of the content removed (a subjective standard) and (2) the requirement that the action to “restrict access to or availability” of that content be taken in good faith (an objective standard). Under the clear consensus of courts that have considered this question, the former does indeed mean “any material that any platform ‘considers’ objectionable” provided that the decision to remove it is taken in “good faith.” This has not created a “de facto immunity to all decisions to censor content” under (c)(2)(A) because the subjective standard of objectionability is constrained by the objective standard of good faith.
Second, NTIA invokes another canon of statutory construction: “ejusdem generis, which holds that catch-all phases [sic] at the end of a statutory lists should be construed in light of the other phrases.” The Ninth Circuit explained why this canon does not apply in its recent Malwarebytes decision:
the specific categories listed in § 230(c)(2) vary greatly: Material that is lewd or lascivious is not necessarily similar to material that is violent, or material that is harassing. If the enumerated categories are not similar, they provide little or no assistance in interpreting the more general category. We have previously recognized this concept. See Sacramento Reg’l Cty. Sanitation Dist. v. Reilly, 905 F.2d 1262, 1270 (9th Cir. 1990) (“Where the list of objects that precedes the ‘or other’ phrase is dissimilar, ejusdem generis does not apply”).
We think that the catchall was more likely intended to encapsulate forms of unwanted online content that Congress could not identify in the 1990s.
The categories of objectionable material mentioned in (c)(2)(A) are obviously dissimilar in the sense that matters most: their constitutional status. Unlike the other categories, “obscenity is not within the area of constitutionally protected speech or press.” Note also that five of these six categories include no qualifier, but the removal of “violent” content qualifies only if it is “excessively violent.” Merely asserting that the six specified categories “[a]ll deal with issues involving media and communications content regulation intended to create safe, family environments,” does not make them sufficiently similar to justify the invocation of ejusdem generis, in part because the term “safe, family environment” itself has no clear legal meaning. Harassment, for example, obviously extends far beyond the concerns of “family environments” and into the way that adults, including in the workplace, interact with each other.
But in the end, this question is another red herring: whether ejusdem generis applies simply means asking whether Congress intended the term to be “broad and all-inclusive” or “limited in scope.” This is, obviously, a profound constitutional question: does the term “otherwise objectionable” protect an ICS provider’s exercise of editorial discretion under the First Amendment or not? Ejusdem generis is a linguistic canon of construction, supporting logical inferences about the meaning of text; it is thus a far weaker canon than canons grounded in substantive constitutional principles. Here, the canon of constitutional avoidance provides ample justification for courts’ interpretation of otherwise “objectionable” as “broad and all-inclusive”:
[W]here an otherwise acceptable construction of a statute would raise serious constitutional problems, the Court will construe the statute to avoid such problems unless such construction is plainly contrary to the intent of Congress . . . . ‘The elementary rule is that every reasonable construction must be resorted to, in order to save a statute from unconstitutionality.’ This approach not only reflects the prudential concern that constitutional issues not be needlessly confronted, but also recognizes that Congress, like this Court, is bound by and swears an oath to uphold the Constitution.
Finally, because of the First Amendment questions involved, it is unlikely that any court would apply the deferential standard of Chevron to an FCC rule reinterpreting “otherwise objectionable” narrowly.
As noted at the outset, Paragraph 230(c)(2) contains two distinct immunities. While (c)(2)(A) protects content moderation, (c)(2)(B) protects those who “make available to information content providers or others the technical means to restrict access to material described in [(c)(2)(A)].” While the latter provision has been subject to little litigation, it merits attention here because the Supreme Court is currently deciding whether to grant cert in a case concerning how this provision intersects with competition law.
It makes sense that Congress grouped these two distinct immunities into Paragraph (c)(2) because both begin the same way (“No provider or user of an interactive computer service shall be held liable on account of. . . .”) and because subparagraph (c)(2)(B) incorporates by reference (c)(2)(A)’s list of grounds for content moderation (ending in “or otherwise objectionable”). But (c)(2)(B) plainly does not incorporate (c)(2)(A)’s “good faith” requirement. Yet the Ninth Circuit recently held otherwise, reading into the statute words that are not there for what are clearly policy, rather than legal, reasons. TechFreedom amicus brief supporting Malwarebytes’ petition for cert explains in detail why the Ninth Circuit made a reversible legal error. As a policy matter, we explain:
It would not make any sense to hold providers of filtering tools liable based on users’ decisions to “purchase, install, and utilize” those tools to tailor their Internet experiences. Zango, 568 F.3d at 1176. Nor would it make sense to hold developers liable when they provide those tools to other ICS providers. Yet the Ninth Circuit’s ruling authorizes both.
The Malwarebytes majority’s creation of an “anticompetitive animus” exception to Section 230(c)(2)(B) was purportedly driven by a “warning” Judge Fisher voiced a decade ago in a concurring opinion in another case: that if the statute were construed according to its “literal terms,” Section 230(c)(2)(B) could immunize “covert, anti-competitive blocking” of desirable content “without the user’s knowledge.” But this concern evinces a significant misunderstanding of the statute. In Judge Fisher’s hypothetical, the decision by the developer of a software program that covertly blocks competitors’ content would not implicate subsection (c)(2)(B) in the first place; when an ICS provider acts independently of the end user (or parent or some other intermediary, e.g., a school or library administrator) to restrict access to material on its own service, its conduct falls squarely under subsection (c)(2)(A). And to be protected by that provision, the provider must act with “good faith.”
C. Even If Section 230 Were Amended or Reinterpreted as Republicans Propose, “Biased” Content Moderation Would Still Be Protected by the First Amendment
In all three cases, even if Congress were to amend Section 230 as Trump’s DOJ has proposed, the First Amendment would still bar antitrust suits predicated on “biased” editorial judgment rather than economic injury. Thus, the net effect of these changes would be to force ICS providers to litigate the kinds of lawsuits discussed above on more complex First Amendment grounds instead of being able to rely on Section 230. This would be exactly equivalent to repealing an anti-SLAPP law—with the same effect: unleashing costly litigation that may be an effective tool of harassment, but will not succeed in court.
Section 230 does not exempt ICS providers from the antitrust or consumer protection laws, but it does bar liability for decisions they make as publishers with respect to content developed wholly by others, or for decisions they make in “good faith” to remove or restrict access to content on their services, regardless of who created it. As we have seen, neither immunity protects them from the kind of anti-competitive conduct that traditional publishers can be sued for. This does not include claims of bias in content moderation. Nor will Section 230 permit consumer protection law claims that could not be brought against traditional publishers for their exercise of editorial discretion—including claims that biased content moderation violates vague promises of “neutrality.” Those who demand that Section 230 be amended, or reinterpreted, to allow such claims should remember what President Reagan said when he vetoed legislation to restore the Fairness Doctrine back in 1987:
We must not ignore the obvious intent of the First Amendment, which is to promote vigorous public debate and a diversity of viewpoints in the public forum as a whole, not in any particular medium, let alone in any particular journalistic outlet. History has shown that the dangers of an overly timid or biased press cannot be averted through bureaucratic regulation, but only through the freedom and competition that the First Amendment sought to guarantee.
 Liza Gross, Darwin’s Tangled Bank in Verse, PLOS Biologue (Dec. 5, 2012), https://biologue.plos.org/2012/12/05/darwins-tangled-bank-in-verse/.
 Darwin’s Origin of Species has been rightly called the “most poetic thing ever written about nature.” Id.
 Moore’s Law, Encyclopædia Britannica (Dec. 26, 2019), https://www.britannica.com/technology/Moores-law.
 Steve Gibson, Network Marketing | Metcalfe’s Law, Medium (May 12, 2020), https://medium.com/@steve_98816/network-marketing-metcalfes-law-973f22d4a4d8.
Liability for User-Generated Content Online, Principles for Lawmakers (July 11, 2019), https://digitalcommons.law.scu.edu/cgi/viewcontent.cgi?article=2992&context=historical.
 Zeran v. Am. Online, 129 F.3d 327, 330 (4th Cir. 1997); see also Barrett v. Rosenthal, 146 P.3d 510, 515 (Cal. 2006).
 47 U.S.C. § 230(c)(2)(A).
 Austin Vining & Sarah Matthews, Introduction to Anti-SLAPP Laws, Reporters Comm. for Freedom of the Press, https://www.rcfp.org/introduction-anti-slapp-guide/#:~:text=As%20of%20fall%202019%2C%
 Congress has considered, but not yet enacted, federal anti-SLAPP legislation. Citizen Participation Act of 2020, H.R. 7771, 116th Cong. (2d Sess. 2020).
 Fair Hous. Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157, 1174 (9th Cir. 2008).
 See, e.g., Jeff Kosseff, The Twenty Six Words That Created The Internet (2019).
 Lorain Journal v. United States, 342 U.S. 143, 156 (1951).
 S. 1693, 115th Cong. (2017).
 See S. 3398, 116th Cong. (Mar. 5, 2020).
 See, e.g., See Something Say Something Online Act of 2020, S. 4758, 116th Cong. (Sept. 29, 2020), https://www.manchin.senate.gov/imo/media/doc/2020_0928%20See%20Something%20Say%20Something%20Online%20Act.pdf?cb; Online Content Policy Modernization Act, S. 4632 116th Cong. (Sept. 21, 2020), https://www.judiciary.senate.gov/imo/media/doc/S4632.pdf?fbclid=IwAR1R_u6DUfbcmRCe-7izU93LGbtT497oLU-WwmJX_3cPdU4t67BhfW-NXCI/.
 Charge of Discrimination, HUD v. Facebook, Inc. FHEO No. 01-18-0323-8 (U.S. Dep’t of Hous. & Urban Dev., Mar. 28, 2019).
 47 U.S.C. § 230(f)(3).
 Fair Hous. Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157 (9th Cir. 2008).
 Christopher Cox, The Origins and Original Intent of Section 230 of the Communications Decency Act, J.L. & Tech., ¶ 1 (Aug. 27, 2020), https://jolt.richmond.edu/2020/08/27/the-origins-and-original-intent-of-section-230-of-the-communications-decency-act/ (emphasis added).
 Preventing Online Censorship at Section 4(c), Exec. Order No. 13925, 85 Fed. Reg. 34079, 34080 (June 2, 2020) [hereinafter Executive Order].
 U.S. Dept. of Justice, Ramseyer Draft Legislative Reforms to Section 230 of the Communications Decency Act (June 17, 2020), https://www.justice.gov/file/1319331/download (“Nothing in this section shall be construed to prevent, impair, or limit any civil action brought under the Federal antitrust laws.”).
 Nat’l Telecomm. & Info. Admin., Petition for Rulemaking of the Nat’l Telecomm. & Info. Admin. (July 27, 2020) [hereinafter NTIA Petition], https://www.ntia.gov/files/ntia/publications/ntia_petition_for_
 TechFreedom, Comments in the Matter of Nat’l Telecomms. & Info. Admin. Petition for Rulemaking to Clarify provisions of Section 230 Of the Communications Act of 1934 (Sept. 2, 2020), https://techfreedom.
org/wp-content/uploads/2020/09/NTIA-230-Petition-Comments-%E2%80%93-9.2.2020.pdf; TechFreedom, Reply Comments (Sept. 17, 2020), https://techfreedom.org/wp-content/uploads/2020/09/NTIA-230-Petition-Reply-Comments-9.17.2020.pdf.
 Ending Support for Internet Censorship Act, S. 1914, 116th Cong. (June 19, 2019), https://www.hawley.senate.gov/sites/default/files/2019-06/Ending-Support-Internet-Censorship-Act-Bill-Text.pdf (removing Section 230 protections for big tech companies unless they submit to an external audit that proves by clear and convincing evidence that their algorithms and content removal practices are politically neutral); Limiting Section 230 Immunity to Good Samaritans Act, S. 3983, 116th Cong. (June 17, 2020); Stop the Censorship Act, H.R. 4027, 116th Cong. (July 25, 2019), https://www.hawley.
senate.gov/sites/default/files/2020-06/Limiting-Section-230-Immunity-to-Good-Samaritans-Act.pdf (denying Section 230 protections unless ICS providers update their terms of service to operate in good faith and pay a $5,000 or actual damages or higher plus attorney’s fees if they violate that promise, and defining ”good faith” to bar selective enforcement); Online Freedom and Viewpoint Diversity Act, S. 4534, 116th Cong. (Sept. 8, 2020), https://www.congress.gov/bill/116th-congress/senate-bill/4632 (“amend[ing] the Communications Act of 1934 to modify the scope of protection from civil liability for ‘good Samaritan’ blocking and screening of offensive material . . . .”); Stopping Big Tech Censorship Act, S. 4062, 116th Cong. (June 24, 2020), https://www.congress.gov/bill/116th-congress/senate-bill/4062 (“amend[ing] section 230 of the Communications Act of 1934 to require that providers and users of an interactive computer service meet certain standards to qualify for liability protections.”).
 Platform Accountability and Consumer Transparency Act, S. 4066, 116th Cong. § 2(9) (June 24, 2020), https://www.congress.gov/116/bills/s4066/BILLS-116s4066is.pdf. (“The term ‘‘small business provider’’ means a provider of an interactive computer service that, during the most recent 24-month period— (A) received fewer than 1,000,000 monthly active users or monthly visitors; and (B) accrued revenue of less than 19 $25,000,000.”).
 Daphne Keller, CDA 230 Reform Grows Up: The PACT Act Has Problems, But It’s Talking About The Right Things, Ctr. for Internet & Soc’y (July 16, 2020), https://cyberlaw.stanford.edu/blog/2020/07/cda-230-reform-grows-pact-act-has-problems-it%E2%80%99s-talking-about-right-things.
 Rep. Schakowsky, Discussion Draft, §§ 5(a)(1)(B), 7(a)(1) (Oct. 2, 2020), https://techfreedom.org/wp-content/uploads/2020/10/Online-Consumer-Protection-Act_discussion-draft_.pdf.
 U.S. Dept. of Justice, Department of Justice’s Review of Section 230 of the Communications Decency Act of 1996 (Sept. 23, 2020), https://www.justice.gov/ag/department-justice-s-review-section-230-communications-decency-act-1996?utm_medium=email&utm_source=govdelivery.
 Cox, supra note 20, ¶ 43.
 NTIA Petition, supra note 23.
 Cox, supra note 20, ¶ 64.
 Id. ¶ 29 (quoting Gingrich Opposes Smut Rule for Internet, N.Y. Times, June 22, 1995, A20).
 Robert Nozick, Anarchy, State, And Utopia 311-312 (1974) (“[T]here will not be one kind of community existing and one kind of life led in utopia. Utopia will consist of utopias, of many different and divergent communities in which people lead different kinds of lives under different institutions. Some kinds of communities will be more attractive to most than others; communities will wax and wane. People will leave some for others or spend their whole lives in one. Utopia is a framework for utopias, a place where people are at liberty to join together voluntarily to pursue and attempt to realize their own vision of the good life in the ideal community but where no one can impose his own utopian vision upon others.”).
 Cox, supra note 20, ¶ 24.
 Reno v. Am. Civil Liberties Union, 521 U.S. 844 (1997).
 Id. at 882.
 Daniel Barnhizer & George Mocsary, Contract Law Professors Comment on the NTIA Petition for Rulemaking and Section 230 of the Communications Act of 1934, at 2 (Filed Sept. 3, 2020), https://ecfsapi.fcc.gov/file/10902206336251/Contract%20230%20Comments-barnhizer%20mocsary.pdf; see also Nat’l Telecomm. & Info. Admin., Reply Comments in the Matter of Section 230 of the Communications Act of 1934, File No. RM-11862, 18 (Sept. 17, 2020), https://www.ntia.gov/files/ntia/
 776 F. Supp. 135 (S.D.N.Y. 1991).
 1995 WL 323710 (N.Y. Sup. Ct., May 24, 1995) (unpublished).
 Cubby, 776 F. Supp. 135 (S.D.N.Y. 1991). Unlike Stratton Oakmont, the Cubby court found no liability, but made clear that this finding depended on the fact that CompuServe had not been provided adequate notice of the defamatory content, thus implying (strongly) that such notice would trigger a takedown obligation under a theory of distributor liability. Id. at 141.
 See Adam Candeub & Mark Epstein, Platform, or Publisher? City Journal (May 7, 2018), https://www.city-journal.org/html/platform-or-publisher-15888.html; Adam Candeub, Social Media Platforms or Publishers? Rethinking Section 230, The Am. Conservative (June 19, 2019), https://www.theamericanconservative.
 Fed. Trade Comm’n v. Accusearch Inc., 570 F.3d 1187, 1195 (10th Cir. 2009).
 47 U.S.C. § 230(c).
 Cristiano Lima, Before Bashing Tech’s Legal Shield, Trump Used It To Defend Himself in Court, Politico (June 4, 2020), https://www.politico.com/news/2020/06/04/tech-legal-trump-court-301861.
 Liability for User-Generated Content Online, Principles for Lawmakers, at 2 (July 11, 2019), https://
 47 U.S.C. § 230(e)(2).
 47 U.S.C. § 230(e)(4).
 47 U.S.C. § 230(e)(5).
 47 U.S.C. § 230(e)(3).
 Nat’l Ass’n of Att’ys Gen., Letter Regarding Amendment of Communications Decency Act (May 23, 2019), https://www.naag.org/assets/redesign/files/sign-on-letter/CDA%20Amendment%20NAAG%20Let
 Cox, supra note 20, ¶ 48.
 47 U.S.C. § 230(c)(1).
 47 U.S.C. § 230(f)(3) (“The term ‘information content provider’ means any person or entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service.”).
 Armchair discussion with Former Congressman Cox, Back to the Future of Tech Policy, YouTube (Aug. 10, 2017), https://www.youtube.com/watch?time_continue=248&v=iBEWXIn0JUY.
 Cox, supra note 20, ¶ 1.
 Fair Hous. Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157, 1169 (9th Cir. 2008).
 See, e.g., Enigma Software Grp. USA, LLC v. Bleeping Computer LLC, 194 F. Supp. 3d 263 (S.D.N.Y. 2016) (pseudonymous user responsible for posting defamatory content acted as the agent of the website).
 Fed. Trade Comm’n v. Accusearch Inc., 570 F.3d 1187, 1199 (10th Cir. 2009).
 Fed. Trade Comm’n v. LeadClick Media, LLC, 838 F.3d 158, 164 (2d Cir. 2016).
 Id. at 176.
 Zeran v. Am. Online, 129 F.3d 327, 330 (4th Cir. 1997).
 Id.; see also Barrett v. Rosenthal, 146 P.3d 510, 515 (Cal. 2006).
 Bennett v. Google, LLC, 882 F.3d 1163, 1168 (D.C. Cir. 2018) (“[T]he decision to print or retract is fundamentally a publishing decision for which the CDA provides explicit immunity. . . . [I]f Bennett takes issue with Pierson’s post, her legal remedy is against Pierson himself as the content provider, not against Google as the publisher.”).
 Force v. Facebook, Inc., 934 F.3d 53, 69-70 (2d Cir. 2019) (Section 230(c)(1) immunity because “Facebook does not edit (or suggest edits) for the content that its users—including Hamas—publish”); id. at 70 (“making information more available is, again, an essential part of traditional publishing.”).
 Noah v. AOL Time Warner Inc., 261 F. Supp. 2d 532, 538 (E.D. Va. 2003) (holding that the plaintiff’s claims “that AOL is liable for its refusal to intervene and stop the allegedly harassing statements” seeks “to place AOL in a publisher’s role, in violation of [Section] 230.”).
 Elizabeth Banker, Internet Association: A Review of Section 230’s Meaning & Application Based On More Than 500 Cases, Internet Association, at 10 (July 27, 2020), https://internetassociation.org/wp-content/uploads/2020/07/IA_Review-Of-Section-230.pdf (concluding, based on a study of more than 500 Section 230 cases, that Section 230 “typically” protects content moderation “through application of subsection (c)(1) rather than (c)(2).”).
 Eric Goldman, Online User Account Termination and 47 U.S.C. § 230(c)(2), 2 U.C. Irvine L. Rev. 659, 664 (2012).
 47 U.S.C. § 230(c)(2)(A).
 See infra at 51.
 See infra at 49-53.
 See infra at 53-57.
 47 U.S.C. § 230(c)(2)(B). Actually, (c)(2)(B) says “material described in paragraph (1),” but courts uniformly recognize this as a “typographical error.” Zango, Inc. v. Kaspersky Lab, Inc., 568 F.3d 1169, 1173 n.5 (9th Cir. 2009).
 See, e.g., Enigma Software Grp. USA, LLC v. Malwarebytes, Inc., 946 F.3d 1040 (9th Cir. 2019); Zango, Inc. v. Kaspersky Lab, Inc., 568 F.3d 1169 (9th Cir. 2009); Fair Hous. Council v. Roommates.com, LLC, 521 F.3d 1157 (9th Cir. 2008); Carafano v. Metrosplash.com, Inc., 339 F.3d 1119 (9th Cir. 2003); PC Drivers Headquarters, LP v. Malwarebytes Inc., 371 F. Supp. 3d 652 (N.D. Cal. 2019); PC Drivers Headquarters, LP v. Malwarebytes, Inc., No. 1:18-CV-234-RP, 2018 U.S. Dist. WL 2996897 (W.D. Tex. Apr. 23, 2018); Barnes v. Yahoo!, Inc., 570 F.3d 1096 (9th Cir. 2009); E360insight, LLC v. Comcast Corp., 546 F. Supp. 2d 605 (N.D. Ill. 2008); Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102 (9th Cir. 2007).
 Social Fixer, https://www.socialfixer.com (last visited June 9, 2020).
 Marshall’s Locksmith Serv. Inc. v. Google, LLC, 925 F.3d 1263, 1265 (D.C. Cir. 2019).
 Id. at 1265.
 Id. at 1268 (citation omitted).
 Id. at 1270-71 (quoting Klayman v. Zuckerberg, 753 F.3d 1354, 1358 (D.C. Cir. 2014)).
 Id. at 1270, 1271.
 Id. at 1271.
 Zeran v. Am. Online, Inc., 129 F.3d 327, 330 (4th Cir. 1997).
 Eric Goldman, D.C. Circuit Issues Sweeping Pro-Section 230 Opinion–Marshall’s Locksmith v. Google, Tech. & Marketing L. Blog (June 7, 2019), https://blog.ericgoldman.org/archives/2019/06/d-c-circuit-issues-sweeping-pro-section-230-opinion-marshalls-locksmith-v-google.htm.
 Zeran v. Am. Online, Inc., 129 F.3d 327, 330 (4th Cir. 1997).
 Miami Herald Publ’g Co. v. Tornillo, 418 U.S. 241, 258 (1974).
 Pac. Gas & Elec. Co. v. Pub. Utilities Comm’n of Cal., 475 U.S. 1, 11, 16 (1986) (plurality opinion) (emphasis in original).
 515 U.S. 557, 573 (1995).
 Id. at 573.
 See, e.g., Langdon v. Google, Inc., 474 F. Supp. 2d 622 (D. Del. 2007).
 47 U.S.C. § 230(c)(2)(A).
 Jefferson Cty. Sch. Dist. No. R-1 v. Moody’s Inv’r Servs., 175 F.3d 848, 860 (10th Cir. 1999).
 Id. at 859 (providing as support the following string citation: “See, e.g., Federal Trade Comm’n v. Superior Court Trial Lawyers Ass’n, 493 U.S. 411, 430-32 (1990) (upholding finding that an attorneys’ association’s boycott of assignments to cases involving indigent defendants violated the antitrust laws even though the boycott had an expressive component); National Society of Professional Engineers v. United States, 435 U.S. 679 (1978) (upholding finding that a professional association’s ban on competitive bidding for engineering services violated the antitrust laws even though one means of carrying out the ban was through the publication of an ethical code); American Society of Mechanical Eng’rs v. Hydrolevel Corp., 456 U.S. 556 (1982) (upholding finding that professional association violated the antitrust laws through the issuance of an inaccurate safety report used to undermine a competitor’s product); Wilk v. American Medical Ass’n, 895 F.2d 352, 357-58, 371 (7th Cir. 1990) (upholding finding that a medical association’s boycott of chiropractors violated the antitrust laws even though one means of enforcing the boycott was through the association’s code of ethics). More generally, the School District relies on decisions holding that the First Amendment does not provide publishers with immunity from antitrust laws. See Citizen Publishing Co v. United States, 394 U.S. 131, 135 (1969) (upholding injunction prohibiting newspaper publishers from engaging in joint operating agreement) . . . .”).
 Newspaper Printing Corp. v. Galbreath, 580 S.W. 2d 777, 779 (Tenn. 1979) (“Newspaper publishers may refuse to publish whatever advertisements they do not desire to publish and this is true even though the newspaper in question may enjoy a virtual monopoly in the area of its publication.”).
 Eugene Volokh & Donald Falk, First Amendment Protection for Search Engine Search Results, at 22 (UCLA School of Law Research Paper No. 12-22, April 20, 2012) (emphasis added).
 Id. at 23 (Volokh explains: “the Ninth Circuit has concluded that even a newspaper that was plausibly alleged to have a ‘substantial monopoly’ could not be ordered to run a movie advertisement that it wanted to exclude, because ‘[a]ppellant has not convinced us that the courts or any other governmental agency should dictate the contents of a newspaper.’ Associates & Aldrich Co. v. Times Mirror Co., 440 F.2d 133, 135 (9th Cir. 1971). And the Tennessee Supreme Court similarly stated that, ‘[n]ewspaper publishers may refuse to publish whatever advertisements they do not desire to publish and this is true even though the newspaper in question may enjoy a virtual monopoly in the area of its publication.’ Newspaper Printing Corp. v. Galbreath, 580 S.W. 2d 777, 779 (Tenn. 1979).”)
 Sambreel Holdings LLC v. Facebook, Inc., 906 F. Supp. 2d 1070, 1075-76 (S.D. Cal. 2012).
 Id. at 1080.
 Id. at 1075 (quoting Verizon Commc’ns Inc. v. Law Offices of Curtis V. Trinko, 540 U.S. 398, 408 (2004)).
 See Miami Herald, 418 U.S. at 258; see supra note 88 and associated text (discussing Miami Herald).
 Complaint for Injunctive Relief, Epic Games, Inc. v Apple Inc., No. 3:20-cv-05640 (N.D. Cal. Aug. 13, 2020), https://www.courtlistener.com/recap/gov.uscourts.cand.364265/gov.uscourts.cand.364265.1.0_1.pdf.
 European Commission, Antitrust: Commission Takes Further Steps in Investigations Alleging Google’s Comparison Shopping and Advertising-Related Practices Breach EU Rules, European Commission (July 14, 2016), https://ec.europa.eu/commission/presscorner/detail/en/IP_16_2532.
 Inappropriate Content, Google, Inc., https://support.google.com/googleplay/android-developer
/answer/9878810 (“We don’t allow apps that contain or promote sexual content or profanity, including pornography, or any content or services intended to be sexually gratifying. Content that contains nudity may be allowed if the primary purpose is educational, documentary, scientific or artistic, and is not gratuitous.”); App Store Review Guidelines, Apple, § 1.1.4, https://developer.apple.com/app-store/review/guidelines/ (banning “[o]vertly sexual or pornographic material”); Microsoft Store Policies, Microsoft, § 11.7, https://docs.microsoft.com/en-us/windows/uwp/publish/store-policies.
Available Apps, Gab, https://apps.gab.com/ (last visited Oct. 12, 2020) (“Gab is banned from the Google Play Store and Apple App Store for refusing to censor speech for Google and Apple”). Unlike Apple, Android does allow users to turn on the ability “sideload” apps that are not verified or screened through the official Play store.
 Brittain, who briefly ran in the 2018 Arizona Republican Senate primary, is best known for having founded isanybodydown.com, a site that encouraged users to post revenge porn and identifying information for those associated with it. Antonia Noori Farzan, Revenge Porn Operator Craig Brittain Running for Jeff Flake’s Senate Seat, Phoenix New Times (Nov. 3, 2017), https://www.phoenixnewtimes.com/
 Brittain v. Twitter, Inc., 2019 WL 2423375, at *3 (N.D. Cal. June 10, 2019).
 Id. at *4.
 See, e.g., S. Poverty Law Ctr., Jared Taylor, https://www.splcenter.org/fighting-hate/extremist-files/individual/jared-taylor (last visited Oct. 5, 2020) (quoting Taylor as saying, in 2005, “[b]lacks and whites are different. When blacks are left entirely to their own devices, Western civilization—any kind of civilization—disappears.”).
 Associated Press, Twitter Suspends White Nationalists as New Rules Take Effect, L.A. Times (Dec. 18, 2017), https://www.latimes.com/business/technology/la-fi-twitter-white-nationalists-20171218-story.html.
 Twitter v. Superior Court ex rel Taylor, A154973 (Cal. App. Ct. Aug. 17, 2018), https://digitalcommons.
 Id. at 195.
 Id. at 196 (quoting Fair Hous. Council of San Fernando Valley v. Roommates.Com, LLC, 521 F.3d 1157, 1170-71 (9th Cir. 2008); see also id. at 195 (“‘[Section] 230 precludes courts from entertaining claims that would place a computer service provider in a publisher’s role. Thus, lawsuits seeking to hold a service provider liable for its exercise of a publisher’s traditional editorial functions—such as deciding whether to publish, withdraw, postpone or alter content—are barred.’ (Barrett v. Rosenthal (2006) 40 Cal.4th 33, 43 (Barrett), quoting Zeran v. America Online, Inc. (4th Cir.1997) 129 F.3d 327, 330, italics added.) Barrett also noted that one ‘important purpose of section 230 was to encourage service providers to self-regulate the dissemination of offensive material over their services.’ Barrett at 44. The immunity provided by section 230 was intended to shield service providers from the fear of liability that might deter them from ‘blocking and screening offensive material . . . .’”) (some internal quotation omitted).
 Id. at 196 (“Thus, we ‘must ask whether the duty that the plaintiff alleges the defendant violated derives from the defendant’s status or conduct as a “publisher or speaker.” If it does, section 230(c)(1) precludes liability.’ (Cross v. Facebook, Inc. (2017) 14 Cal.App.5th 190, 207 (Cross), quoting Barnes v. Yahoo!, Inc. (9th Cir. 2009) 570 F .3d 1096, 1102 (Barnes).) We observe that California courts have held that a service provider’s decision ‘to restrict or make available certain material is expressly covered by section 230.’ Doe II v. MySpace, Inc. (2009) 175 Cal.App.4th 561, 573.) And federal courts have specifically ruled that a service provider’s exercise of control over user accounts is immunized by section 230. (Fields v. Twitter, Inc., supra, 217 F.Supp.3d at p. 1124 [“[T]he decision to furnish an account, or prohibit a particular user from obtaining an account, is itself publishing activity.”]; see also Riggs v. MySpace, Inc. (9th Cir. 2011) 444 Fed.Appx. 986, 987 [claims “arising from MySpace’s decisions to delete . . . user profiles on its social networking website yet not delete other profiles . . . were precluded by section 230(c)(1) of the Communications Decency Act.”]; Cohen v. Facebook, Inc. (E.D.N.Y. 2017) 252 F.Supp.3d 140, 157 [“Facebook’s choices as to who may use its platform are inherently bound up in its decisions as to what may be said on its platform, and so liability imposed based on its failure to remove users would equally derive from [Facebook’s] status or conduct as a “publisher or speaker.” ‘]; Sikhs for Justice “SFJ”, Inc. v. Facebook, Inc. (N.D. Cal. 2015) 144 F.Supp.3d 1088, 1094-1095 [CDA barred claim under title II of the Civil Rights Act of 1964 (42 U.S.C. § 2000a) alleging that Facebook “was motivated solely by unlawful discrimination” in blocking access to plaintiffs Facebook page in India; claim sought “to hold Defendant liable for Defendant’s decision ‘whether to publish’ third-party content”].) Indeed, ‘any activity that can be boiled down to deciding whether to exclude material that third parties seek to post online is perforce immune under section 230.’ (Fair Housing Council of San Fernando Valley v. Roommates.com, LLC (9th Cir. 2008) 521 F.3d 1157, 1170-1171 (en banc)).”) (some internal quotations omitted).
 Taylor v. Twitter, Inc., CGC 18-564460 at 16 (Cal. Super. Ct. Mar. 8, 2019), https://digitalcommons
 Lewis v. Google LLC, 2020 WL 2745253 (N.D. Cal. May 21, 2020).
 Id. at *9.
 Ben Domenech & Sean Davis, NBC Tries to Cancel a Conservative Website, Wall St. J. (June 17, 2020), https://www.wsj.com/articles/nbc-tries-to-cancel-a-conservative-website-11592410893.
 Mike Masnick, No, Google Didn’t Demonetize The Federalist & It’s Not an Example of Anti-Conservative Bias, Techdirt (June 16, 2020), https://www.techdirt.com/articles/20200616/14390744730/no-google-didnt-demonetize-federalist-not-example-anti-conservative-bias.shtml.
 Mike Masnick, Why Are There Currently No Ads On Techdirt? Apparently Google Thinks We’re Dangerous, Techdirt (Aug. 12, 2020), https://www.techdirt.com/articles/20200810/11335745081/why-are-there-currently-no-ads-techdirt-apparently-google-thinks-were-dangerous.shtml.
 Aaron Mak, Google’s Advertising Platform Is Blocking Articles About Racism, Slate (Aug. 13, 2020), https://slate.com/technology/2020/08/googles-ad-exchange-blocking-articles-about-racism.html.
 Mike Masnick, Masnick’s Impossibility Theorem: Content Moderation at Scale Is Impossible To Do Well, Techdirt (Nov. 20, 2019) (emphasis added), https://www.techdirt.com/articles/20191111/23032743367/
 Tess Owen, White Supremacists Have a Disgusting New Code for the N-Word After Ahmaud Arbery’s Death, Vice News (May 14, 2020), https://www.vice.com/en_us/article/bv88a5/white-supremacists-have-a-disgusting-new-code-for-the-n-word-after-ahmaud-arberys-death.
 See Ending Support for Internet Censorship Act, supra note 25.
 Executive Order, supra note 21.
 NTIA Petition, supra note 23, at 26.
 NTIA Petition, supra note 23, at 26.
 See Ending Support for Internet Censorship Act, supra note 25.
 Republican Platform 2016, at 12 (2016) (“We likewise call for an end to the so-called Fairness Doctrine, and support free-market approaches to free speech unregulated by government.”), https://prod-cdn-static.gop.com/media/documents/DRAFT_12_FINAL%5B1%5D-ben_1468872234.pdf.
 Pittsburgh Press Co. v. Human Rel. Comm’n, 413 U.S. 376, 385 (1973) (citing Valentine v. Chrestensen, 316 U.S. 52 (1942)).
 Id. at 385.
 Congressional Research Service, The Constitution of the United States of America, Analysis and Interpretation, S. Doc. No. 112-9, at 1248 (Centennial ed. 2013), https://www.law.cornell.edu/constitution-conan/amendment-1/commercial-speech.
 Leah Nylen, Trump Aides Interviewing Replacement for Embattled FTC Chair, Politico (Aug. 28, 2020), https://www.politico.com/news/2020/08/28/trump-ftc-chair-simons-replacement-404479.
 The Twitter Rules, Twitter, https://help.twitter.com/en/rules-and-policies/twitter-rules (last visited Aug. 31, 2020).
 Pittsburgh Press Co. v. Pittsburg Comm’n on Human Relations, 413 U.S. 376, 385 (1973).
 Bd. of Educ. v. Barnette, 319 U.S. 624, 642 (1943).
 Petition for Initiation of Complaint Against Fox News Network, LLC for Deceptive Practices Under Section 5 of the FTC Act, MoveOn.org and Common Cause (July 19, 2004),
 Id. at 2.
 Id. at 3.
 Statement of Fed. Trade Comm’n Chairman Timothy J. Muris on the Complaint Filed Today by MoveOn.org (July 19, 2004), https://www.ftc.gov/news-events/press-releases/2004/07/statement-federal-trade-commission-chairman-timothy-j-muris.
 Fed. Trade Comm’n, FTC Policy Statement on Deception, at 1 (Oct. 14, 1983) [hereinafter Deception Statement].
 PragerU, YouTube, https://www.youtube.com/user/PragerUniversity/about (last visited September 25, 2020).
 PragerU Takes Legal Action Against Google and YouTube for Discrimination, PragerU (2020), https://www.prageru.com/press-release/prageru-takes-legal-action-against-google-and-youtube-for-discrimination/.
 Dennis Prager, Don’t Let Google Get Away With Censorship, Wall St. J. (Aug. 6, 2019), https://www.wsj.com/articles/dont-let-google-get-away-with-censorship-11565132175.
 Your content & Restricted Mode, YouTube Help (2020), https://support.google.com/youtube/answer
 Prager Univ. v. Google LLC, 951 F.3d 991, 1000 (9th Cir. 2020) (internal citations omitted).
 15 U.S.C. § 1125 (a)(1).
 15 U.S.C. § 1125 (a)(1)(A).
 Coastal Abstract Serv. v. First Amer. Title, 173 F.3d 725, 731 (9th Cir. 1998).
 Deception Statement, supra note 148, at 4. The Commission added: “Some exaggerated claims, however, may be taken seriously by consumers and are actionable.” But the Commission set an exceptionally high bar for such claims:
For instance, in rejecting a respondent’s argument that use of the words “electronic miracle” to describe a television antenna was puffery, the Commission stated: Although not insensitive to respondent’s concern that the term miracle is commonly used in situations short of changing water into wine, we must conclude that the use of “electronic miracle” in the context of respondent’s grossly exaggerated claims would lead consumers to give added credence to the overall suggestion that this device is superior to other types of antennae.
 Twitter: Transparency and Accountability: Hearing Before the H. Comm. on Energy & Commerce, 115th Cong. (Sept. 5, 2018) (statement of Jack Dorsey, CEO, Twitter, Inc.), https://energycommerce.house.gov/sites/democrats.energycommerce.house.gov/files/documents/Testimony%20-Dorsey-FC-Hrg-on-Twitter-Transparency-and-Accountabilit-2018-09-05.pdf.
 Prager Univ. v. Google LLC, 951 F.3d 991, 1000 (9th Cir. 2020).
 Deception Statement, supra note 148, at 3.
 Id. at 1.
 As the DPS notes, “the Commission presumes that express claims are material. As the Supreme Court stated recently, ‘[i]n the absence of factors that would distort the decision to advertise, we may assume that the willingness of a business to promote its products reflects a belief that consumers are interested in the advertising.’” Id. at 5 (quoting Central Hudson Gas Elec. v. Public Serv. Comm’n, 447 U.S. 557, 567 (1980)); see also Geoffrey A. Manne, R. Ben Sperry & Berin Szóka, In the Matter of Nomi Technologies, Inc.: The Dark Side of the FTC’s Latest Feel-Good Case (ICLE Antitrust & Consumer Protection Research Program White Paper 2015-1), http://laweconcenter.org/images/articles/icle-nomi_white_paper.pdf.
 NTIA Reply Comments, supra note 40, at 37 (citing Farmers Educational & Cooperative Union v. WDAY, Inc., 360 U.S. 526-28 (1959)).
 395 U.S. 367, 386 (1969) (“Although broadcasting is clearly a medium affected by a First Amendment interest, differences in the characteristics of new media justify differences in the First Amendment standards.”); See supra note 88.
 See supra Section V.A.
 Riley v. Nat’l Fed’n of the Blind, 487 U.S. 781, 795 (1988).
 Id. at 797 (citing Miami Herald Publ’g Co. v. Tornillo, 418 U.S. 241, 256 (1974)).
 See The Twitter Rules, Twitter, https://help.twitter.com/en/rules-and-policies/twitter-rules (last visited Aug. 31, 2020).
 Wooley v. Maynard, 430 U.S. 705, 714 (1977).
 Riley, 487 U.S. at 797-98.
 Packingham v. North Carolina, 137 S. Ct. 1730, 1737 (2017)
 See, e.g., Komal S. Patel, Note, Testing the Limits of the First Amendment: How Online Civil Rights Testing is Protected Speech Activity, 118 Colum. L. Rev. 1473, 1508 (2018) (stating that the “[Packingham] Court implied that for First Amendment purposes, the internet might very well be considered a public forum.”); Executive Order, supra note 21, at 34080.
 NTIA Petition, supra note 23, at 7.
 Packingham, 137 S. Ct. at 1732.
 Executive Order, supra note 21, at 34082.
 137 S. Ct. at 1738, 1743 (Alito, J., concurring in judgement).
 447 U.S. 74, 85-89 (1980).
 See hiQ Labs, Inc. v. LinkedIn Corp., 273 F. Supp. 3d 1099, 1115-16 (N.D. Cal. 2017); Prager Univ. v. Google LLC, 951 F.3d 991, 1000 (9th Cir. 2020).
 Reno v. Am. Civil Liberties Union, 521 U.S. 844, 868 (1997).
 Brown v. Entm’t Merchs. Assn., 564 U.S. 786, 790 (2011).
 Jackson v. Metro. Edison Co., 419 U.S. 345, 352 (1974).
 Manhattan Cmty. Access Corp. v. Halleck, 139 S. Ct. 1921, 1930 (June 17, 2019) (holding that the private operator of a public access TV channel is not a state actor and not bound by the First Amendment in the operator’s programming choices).
 Id. at 1929.
 Prager Univ. v. Google LLC, 951 F.3d 991, 997 & n.3 (9th Cir. 2020) (citing Freedom Watch, Inc. v. Google, Inc., 368 F. Supp. 3d 30, 40 (D.D.C. 2019)) (“Facebook and Twitter . . . are private businesses that do not become ‘state actors’ based solely on the provision of their social media networks to the public.”); Green v. YouTube, LLC, 2019 WL 1428890, at *4 (D.N.H. Mar. 13, 2019) (stating there is no “state action giving rise to the alleged violations of [the plaintiff’s] First Amendment rights” by YouTube and other platforms that are “all private companies”); Nyabwa v. FaceBook, 2018 WL 585467, at *1 (S.D. Tex. Jan. 26, 2018) (“Because the First Amendment governs only governmental restrictions on speech, [the plaintiff] has not stated a cause of action against FaceBook.”); Shulman v. Facebook.com, 2017 WL 5129885, at *4 (D.N.J. Nov. 6, 2017) (Facebook is not a state actor); Forbes v. Facebook, Inc., 2016 WL 676396, at *2 (E.D.N.Y. Feb. 18, 2016) (“Facebook is a private corporation” whose actions may not “be fairly attributable to the state”); Doe v. Cuomo, 2013 WL 1213174, at *9 (N.D.N.Y. Feb. 25, 2013) (Facebook is not a state actor under the joint action test).
 NTIA Petition, supra note 23, at 42 (quoting 47 U.S.C. § 230(f)(3)).
 Id. at 41.
 Fair Hous. Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157, 1169 (9th Cir. 2008).
 NTIA Petition, supra note 23, at 40.
 Miami Herald Publ’g Co. v. Tornillo, 418 U.S. 241, 258 (1974). See generally supra Section IV.A.
 Batzel v. Smith, 333 F.3d 1018, 1031 (9th Cir. 2003).
 Reed v. Town of Gilbert, 576 U.S. 155, 163 (2015).
 Id. at 171.
 Pleasant Grove City v. Summum, 555 U.S. 460, 480 (2009).
 Indeed, the risk of such a content death-spiral helped drive the Supreme Court’s conclusion that government speech at a location does not automatically turn that location into a public forum. “[W]here application of forum analysis would lead almost inexorably to closing of the forum, it is obvious that forum analysis is out of place.” Id.
 NTIA Petition, supra note 23, at 27.
 See supra note 25.
 NTIA Petition at 28.
 “NTIA urges the FCC to follow the canon against surplusage in any proposed rule. Explaining this canon, the Supreme Court holds, ‘[a] statute should be construed so that effect is given to all its provisions, so that no part will be inoperative or superfluous, void or insignificant . . . .’ The Court emphasizes that the canon ‘is strongest when an interpretation would render superfluous another part of the same statutory scheme.’” Id. at 29.
 Elizabeth Banker, A Review of Section 230’s Meaning & Application Based on More Than 500 Cases, Internet Association, at 10 (July 27, 2020), https://internetassociation.org/publications/a-review-of-section-230s-meaning-application-based-on-more-than-500-cases/ (“Of the decisions reviewed pertaining to content moderation decisions made by a provider to either allow content to remain available or remove or restrict content, only 19 of the opinions focused on Section 230(c)(2). Of these, the vast majority involved disputes over provider efforts to block spam. The remainder were resolved under Section 230(c)(1), Anti-SLAPP motions, the First Amendment, or for failure to state a claim based on other deficiencies.”).
 Barnes v. Yahoo!, Inc., 565 F.3d, 560, 569-70; see also Fyk v. Facebook, Inc., 808 F. App’x 597, 598 (9th Cir. 2020) (reaffirming Barnes).
 Memorandum from Paul V. Niemeyer, Chair, Advisory Committee on Civil Rules, to Hon. Anthony J. Scirica, Chair, Committee on Rules of Practice and Procedure (May 11, 1999), 192 F.R.D. 354, 357 (2000).
 NTIA Petition, supra note 23, at 23.
 Id. at 31.
 Eric Goldman, Comments on NTIA’s Petition to the FCC Seeking to Destroy Section 230, Technology and Marketing L. Blog (Aug. 12, 2020), https://blog.ericgoldman.org/archives/2020/08/comments-on-ntias-petition-to-the-fcc-seeking-to-destroy-section-230.htm (“I have never seen this typo by anyone who actually understands Section 230. It’s so frustrating when our tax dollars are used to fund a B-team’s work on this petition (sorry for the pun).”)
 Cf. e360Insight, LLC v. Comcast Corp., 546 F. Supp. 2d 605 (N.D. Ill. 2008) (dismissing unfair competition claims as inadequately pled, but implying that better pled claims might make a prima facie showing of “bad faith” sufficient to require Comcast to establish its “good faith”).
 NTIA Petition, supra note 23, at 32 (citing Washington State Dep’t of Soc. & Health Servs. v. Guardianship Estate of Keffeler, 537 U.S. 371, 372 (2003)) (“[U]nder the established interpretative canons of noscitur a sociis and ejusdem generis, where general words follow specific words in a statutory enumeration, the general words are construed to embrace only objects similar to those enumerated by the specific words.”).
 Enigma Software Grp. U.S.A v. Malwarebytes, Inc., 946 F.3d 1040, 1052 (9th Cir. 2019). The Reilly court explained:
The phrase “other property” added to a list of dissimilar things indicates a Congressional intent to draft a broad and all-inclusive statute. In Garcia, the phrase “other property” was intended to be expansive, so that one who assaulted, with intent to rob, any person with charge, custody, or control of property of the United States would be subject to conviction under 18 U.S.C. § 2114. Where the list of objects that precedes the “or other” phrase is dissimilar, ejusdem generis does not apply. However, the statute at issue here falls into a different category. Because section 1292(1) presents a number of similar planning and preliminary activities linked together by the conjunction “or,” the principle of ejusdem generis does apply. “[O]r other necessary actions” in the statute before us refers to action of a similar nature to those set forth in the parts of the provision immediately preceding it. We have previously construed “or other” language that follows a string of similar acts and have concluded that the language in question was intended to be limited in scope—a similar conclusion to the one we reach today.
905 F.2d at 1270.
 Roth v. United States, 354 U.S. 476 (1957).
 Sacramento Reg’l Cty. Sanitation Dist. v. Reilly, 905 F.2d 1262, 1270 (9th Cir. 1990).
 DeBartolo Corp. v. Florida Gulf Coast Trades Council, 485 U.S. 568, 575 (1988) (quoting Hooper v. California, 155 U.S. 648, 657 (1895)). Accord Burns v. United States, 501 U.S. 129, 138 (1991); Gollust v. Mendell, 501 U.S. 115, 126 (1991).
 See, e.g., U.S. West v. FCC, 182 F.3d 1224 (10th Cir. 1999) (“It is seductive for us to view this as just another case of reviewing agency action. However, this case is a harbinger of difficulties encountered in this age of exploding information, when rights bestowed by the United States Constitution must be guarded as vigilantly as in the days of handbills on public sidewalks. In the name of deference to agency action, important civil liberties, such as the First Amendment’s protection of speech, could easily be overlooked. Policing the boundaries among constitutional guarantees, legislative mandates, and administrative interpretation is at the heart of our responsibility. This case highlights the importance of that role.”).
 47 U.S.C. § 230(c)(2)(B).
 Enigma Software Grp. USA, LLC v. Malwarebytes, Inc., 946 F.3d 1040 (9th Cir. 2019); Zango, Inc. v. Kaspersky Lab, Inc., 568 F.3d 1169 (9th Cir. 2009); Fair Hous. Council v. Roommates.com, LLC, 521 F.3d 1157 (9th Cir. 2008); Carafano v. Metrosplash.com, Inc., 339 F.3d 1119 (9th Cir. 2003); PC Drivers Headquarters, LP v. Malwarebytes Inc., 371 F. Supp. 3d 652 (N.D. Cal. 2019); PC Drivers Headquarters, LP v. Malwarebytes, Inc., No. 1:18-CV-234-RP, 2018 U.S. Dist. WL 2996897 (W.D. Tex. Apr. 23, 2018); Barnes v. Yahoo!, Inc., 570 F.3d 1096 (9th Cir. 2009); E360insight, LLC v. Comcast Corp., 546 F. Supp. 2d 605 (N.D. Ill. 2008); Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102 (9th Cir. 2007).
 Enigma Software Grp. U.S.A, 946 F.3d 1040, 1052 (9th Cir. 2019).
 Brief for TechFreedom, as Amici Curiae Supporting Petitioner, Malwarebytes, Inc., v. Enigma Software Grp. U.S.A, 2020 WL 6037214 (No. 19-1284) (Oct. 13, 2020), https://techfreedom.org/wp-content
 Id. at 8-9.
 Zango, 568 F.3d at 1178-79 (Fisher, J., concurring) (emphasis omitted); see Enigma, 946 F.3d at 1045.
 Schools and libraries can use a variety of filtering tools to limit access by their students, visitors and patrons to content the schools or libraries deem objectionable. For example, parents and institutional administrators alike can opt-in to YouTube’s Restricted Mode, which will then automatically block videos flagged as being in six categories: drugs and alcohol, sexual situations, violence, mature subjects, profane and mature language; and incendiary and demeaning content. Your content & Restricted Mode, Youtube, https://support.google.com/youtube/answer/7354993?hl=en#:~:text=Restricted%20Mode%20is%20an%20optional,off%20for%20viewers%20by%20default (last visited June 11, 2020). Congress clearly intended 230(c)(2)(B) to cover such technologies, and specifically included “libraries or educational institutions” in the definition of “interactive computer service.” 47 U.S.C. § 230(f)(2).
 Message from the President Vetoing S. 742, S. Doc. No. 10-100, at 2 (1987), https://www.senate.gov