NetChoice v. Paxton

Case: 21-51178 Document: 00516474571 Page: 1 Date Filed: 09/16/2022 United States Court of Appeals for the Fifth Circuit United States Court of Appeals Fifth Circuit FILED September 16, 2022 No. 21-51178 Lyle W. Cayce Clerk NetChoice, L.L.C., a 501(c)(6) District of Columbia organization doing business as NetChoice; Computer Communications Industry Association, a 501(c)(6) non-stock Virginia Corporation doing business as CCIA, Plaintiffs—Appellees, versus Ken Paxton, in his official capacity as Attorney General of Texas, Defendant—Appellant. Appeal from the United States District Court for the Western District of Texas USDC No. 1:21-cv-840 Before Jones, Southwick, and Oldham, Circuit Judges. Andrew S. Oldham, Circuit Judge: * A Texas statute named House Bill 20 generally prohibits large social media platforms from censoring speech based on the viewpoint of its speaker. The platforms urge us to hold that the statute is facially unconstitutional and hence cannot be applied to anyone at any time and under any circumstances. * Judge Jones joins all but Part III.E and Part V.B.3 of this opinion. 1 Case: 21-51178 Document: 00516474571 Page: 2 Date Filed: 09/16/2022 No. 21-51178 In urging such sweeping relief, the platforms offer a rather odd inversion of the First Amendment. That Amendment, of course, protects every person’s right to “the freedom of speech.” But the platforms argue that buried somewhere in the person’s enumerated right to free speech lies a corporation’s unenumerated right to muzzle speech. The implications of the platforms’ argument are staggering. On the platforms’ view, email providers, mobile phone companies, and banks could cancel the accounts of anyone who sends an email, makes a phone call, or spends money in support of a disfavored political party, candidate, or business. What’s worse, the platforms argue that a business can acquire a dominant market position by holding itself out as open to everyone—as Twitter did in championing itself as “the free speech wing of the free speech party.” Blue Br. at 6 & n.4. Then, having cemented itself as the monopolist of “the modern public square,” Packingham v. North Carolina, 137 S. Ct. 1730, 1737 (2017), Twitter unapologetically argues that it could turn around and ban all pro-LGBT speech for no other reason than its employees want to pick on members of that community, Oral Arg. at 22:39–22:52. Today we reject the idea that corporations have a freewheeling First Amendment right to censor what people say. Because the district court held otherwise, we reverse its injunction and remand for further proceedings. 2 Case: 21-51178 Document: 00516474571 Page: 3 Date Filed: 09/16/2022 No. 21-51178 I. A. This case involves HB 20, a Texas statute that regulates large social media platforms. 1 The law regulates platforms 2 with more than 50 million monthly active users (“Platforms”), such as Facebook, Twitter, and YouTube. Tex. Bus. & Com. Code § 120.002(b). In enacting HB 20, the Texas legislature found that the Platforms “function as common carriers, are affected with a public interest, are central public forums for public debate, and have enjoyed governmental support in the United States.” It further found that “social media platforms with the largest number of users are common carriers by virtue of their market dominance.” Two sections of HB 20 are relevant to this suit. First is Section 7, which addresses viewpoint-based censorship of users’ posts. Section 7 provides: A social media platform may not censor a user, a user’s expression, or a user’s ability to receive the expression of another person based on: 1 The full text of HB 20 can be viewed here: https://perma.cc/9KF3-LEQX. The portions of HB 20 relevant to this lawsuit are codified at Texas Business and Commerce Code §§ 120.001–151 and Texas Civil Practice and Remedies Code §§ 143A.001–08. 2 HB 20 defines “social media platform” to include “an Internet website or application that is open to the public, allows a user to create an account, and enables users to communicate with other users for the primary purpose of posting information, comments, messages, or images.” Tex. Bus. & Com. Code § 120.001(1). The definition expressly excludes internet service providers, email providers, and any “online service, application, or website” that “consists primarily of news, sports, entertainment, or other information or content that is not user generated but is preselected by the provider,” and “for which any chat, comments, or interactive functionality is incidental to, directly related to, or dependent on the provision of [that] content.” Id. § 120.001(1)(A)– (C). 3 Case: 21-51178 Document: 00516474571 Page: 4 Date Filed: 09/16/2022 No. 21-51178 (1) the viewpoint of the user or another person; (2) the viewpoint represented in the user’s expression or another person’s expression; or (3) a user’s geographic location in this state or any part of this state. Tex. Civ. Prac. & Rem. Code § 143A.002(a). “Censor” means “to block, ban, remove, deplatform, demonetize, de-boost, restrict, deny equal access or visibility to, or otherwise discriminate against expression.” Id. § 143A.001(1). For Section 7 to apply, a censored user must reside in Texas, do business in Texas, or share or receive expression in Texas. Id. § 143A.004(a)–(b). This prohibition on viewpoint-based censorship contains several qualifications. Section 7 does not limit censorship of expression that a Platform “is specifically authorized to censor by federal law”; expression that “is the subject of a referral or request from an organization with the purpose of preventing the sexual exploitation of children and protecting survivors of sexual abuse from ongoing harassment”; expression that “directly incites criminal activity or consists of specific threats of violence targeted against a person or group because of their race, color, disability, religion, national origin or ancestry, age, sex, or status as a peace officer or judge”; or “unlawful expression.” Id. § 143A.006. Finally, Section 7 provides a narrow remedial scheme. If a Platform violates Section 7 with respect to a user, that user may sue for declaratory and injunctive relief and may recover costs and attorney’s fees if successful. Id. § 143A.007. The Attorney General of Texas may also sue to enforce Section 7 and may recover attorney’s fees and reasonable investigative costs if successful. Id. § 143A.008. Damages are not available. 4 Case: 21-51178 Document: 00516474571 Page: 5 Date Filed: 09/16/2022 No. 21-51178 The other relevant provision of HB 20 is Section 2. It imposes certain disclosure and operational requirements on the Platforms. These requirements fall into three categories. First, Platforms must disclose how they moderate and promote content and publish an “acceptable use policy.” Tex. Bus. & Com. Code §§ 120.051–52. This policy must inform users about the types of content allowed on the Platform, explain how the Platform enforces its policy, and describe how users can notify the Platform of content that violates the policy. Id. § 120.052(b). Platforms must also publish a “biannual transparency report.” Id. § 120.053. This report must contain various high-level statistics related to the Platform’s content-moderation efforts, including the number of instances in which the Platform was alerted to the presence of policy-violating content; how the Platform was so alerted; how many times the Platform acted against such content; and how many such actions were successfully or unsuccessfully appealed. See ibid. Last, Platforms must maintain a complaint-and-appeal system for their users. See id. §§ 120.101–04. When a Platform removes user-submitted content, it must generally explain the reason to the user in a written statement issued concurrently with the removal. Id. § 120.103(a). It also must permit the user to appeal the removal and provide a response to the appeal within 14 business days. Id. § 120.104. Section 2 includes various exceptions to these notice-and-appeal requirements. See id. § 120.103(b). Only the Texas Attorney General may enforce Section 2. Id. § 120.151. The Attorney General may seek injunctive relief but not damages. Ibid. B. NetChoice and the Computer & Communications Industry Association are trade associations representing companies that operate 5 Case: 21-51178 Document: 00516474571 Page: 6 Date Filed: 09/16/2022 No. 21-51178 Platforms covered by HB 20. They sued the Attorney General of Texas (“Texas”) on September 22, 2021, before HB 20 went into effect. The district court issued a preliminary injunction on December 1, 2021. It first held that Section 7 is facially unconstitutional. The court “start[ed] from the premise that social media platforms are not common carriers.” It then concluded that Platforms engage in “some level of editorial discretion” by managing and arranging content, and viewpoint-based censorship is part of that editorial discretion. It further held that this editorial discretion is protected by cases like Miami Herald Publishing Co. v. Tornillo, 418 U.S. 241 (1974). So according to the district court, HB 20’s prohibition on viewpoint-based censorship unconstitutionally interfered with the Platforms’ protected editorial discretion. The court did not explain why a facial attack on Section 7 was appropriate, other than asserting that Section 7 is “replete with constitutional defects” and the court believed “nothing . . . could be severed and survive.” The district court then held that Section 2 is facially unconstitutional. It reasoned that “Section 2’s disclosure and operational provisions are inordinately burdensome given the unfathomably large numbers of posts on these sites and apps.” Moreover, the court reasoned that Section 2 will “chill the social media platforms’ speech” by disincentivizing viewpoint-based censorship. Again, the court did not explain why a facial challenge to Section 2 was appropriate, other than stating that it imposes “onerously burdensome disclosure and operational requirements.” The district court also found that HB 20 discriminates based on content and speaker, because it permits censorship of some content (like specific threats of violence directed at a protected class) and only applies to large social media platforms. It then held that HB 20 fails any level of heightened scrutiny. Finally, it issued a preliminary injunction. 6 Case: 21-51178 Document: 00516474571 Page: 7 Date Filed: 09/16/2022 No. 21-51178 Texas timely appealed. On December 15, 2021, Texas moved for a stay of the preliminary injunction. We granted that motion on May 11, 2022. On May 31, 2022, in a 5–4 decision, the Supreme Court vacated our stay. Justice Kagan noted her dissent. Justice Alito, joined by Justice Thomas and Justice Gorsuch, authored a six-page dissenting opinion to argue that our stay should have remained undisturbed. II. We review the district court’s preliminary injunction for abuse of discretion. Atchafalaya Basinkeeper v. U.S. Army Corps of Eng’rs, 894 F.3d 692, 696 (5th Cir. 2018). A district court abuses its discretion if it grants an injunction based on clearly erroneous factual findings or erroneous conclusions of law. Ibid. A preliminary injunction is “an extraordinary remedy that may only be awarded upon a clear showing that the plaintiff is entitled to such relief.” Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 22 (2008). “A plaintiff seeking a preliminary injunction must establish that he is likely to succeed on the merits, that he is likely to suffer irreparable harm in the absence of preliminary relief, that the balance of equities tips in his favor, and that an injunction is in the public interest.” Id. at 20. III. The Platforms contend that Section 7 of HB 20 is facially unconstitutional. We disagree. We (A) first reject the Platforms’ facial overbreadth challenge because Section 7 does not chill speech; if anything, it chills censorship. Then we (B) turn to the First Amendment’s text and history, which offer no support for the Platforms’ claimed right to censor. Next, applying Supreme Court precedent, we (C) hold that Section 7 does not regulate the Platforms’ speech at all; it protects other people’s speech and regulates the Platforms’ conduct. Our decision (D) is reinforced by 47 U.S.C. 7 Case: 21-51178 Document: 00516474571 Page: 8 Date Filed: 09/16/2022 No. 21-51178 § 230, which reflects Congress’s judgment that the Platforms are not “speaking” when they host other people’s speech. Our decision (E) is still further reinforced by the common carrier doctrine, which vests the Texas Legislature with the power to prevent the Platforms from discriminating against Texas users. Finally, even if all of that’s wrong and Section 7 does regulate the Platforms’ speech, it (F) satisfies the intermediate scrutiny that applies to content-neutral rules. A. We begin with the First Amendment overbreadth doctrine. It (1) offers a facial constitutional remedy that protects speech. It (2) does not apply here because if Section 7 chills anything, it chills censorship. And the Platforms’ parade of whataboutisms proves their real complaint is a purely speculative one about how HB 20 will be enforced. The Platforms are therefore not entitled to pre-enforcement facial relief against Section 7. 1. The Platforms have asked a federal court to invalidate HB 20 in its entirety before Texas even tries to enforce it. 3 To put it mildly, pre- enforcement facial challenges to legislative acts are “disfavored for several 3 The plaintiff trade associations—which include every Platform subject to HB 20—asked the district court to find the statute could never be constitutionally enforced against them. They did so before the law could be enforced against anyone. See 13B Charles Alan Wright et al., Federal Practice and Procedure § 3532.3 (3d ed. Apr. 2022 Update) (stressing the “distinctions between the ripeness of broad attacks on the legitimacy of any regulation and the nonripeness of more particular attacks on more specific applications”). During briefing in the district court, the Platforms characterized their suit as a facial challenge to HB 20. The district court’s opinion thus properly treated this suit as a facial challenge, and the Platforms do not object to that characterization on appeal. 8 Case: 21-51178 Document: 00516474571 Page: 9 Date Filed: 09/16/2022 No. 21-51178 reasons.” Wash. State Grange v. Wash. State Republican Party, 552 U.S. 442, 450 (2008). Three bear emphasis here. First, the judicial power vested in us by Article III does not include the power to veto statutes. And that omission is no accident: The Founders expressly considered giving judges that power, and they decided not to do so. Several delegates at the Constitutional Convention suggested creating a “Council of Revision” consisting of federal judges and the executive. Jonathan F. Mitchell, The Writ-of-Erasure Fallacy, 104 Va. L. Rev. 933, 954 (2018). They wanted to empower this Council to veto Congress’s legislation, subject to congressional override. Ibid. A veto would render the legislation “void.” Ibid. But despite the best efforts of James Wilson and James Madison, the Convention rejected the proposal—three times over. Id. at 957–59. That means we have no power to “strike down,” “void,” or “invalidate” an entire law. See id. at 936 (explaining that “federal courts have no authority to erase a duly enacted law from the statute books” but have only the power “to decline to enforce a statute in a particular case or controversy” and “to enjoin executive officials from taking steps to enforce a statute”); Borden v. United States, 141 S. Ct. 1817, 1835–36 (2021) (Thomas, J., concurring in the judgment) (noting that “[c]ourts have no authority to strike down statutory text” and that “a facial challenge, if successful, has the same effect as nullifying a statute” (quotations omitted)); Kevin C. Walsh, Partial Unconstitutionality, 85 N.Y.U. L. Rev. 738, 756 (2010) (explaining that the Founders did not conceive of judicial review as the power to “strike down” legislation). Second, the judicial power vested in us by Article III is limited to deciding certain “Cases” and “Controversies.” U.S. Const. art. III, § 2. A federal court “has no jurisdiction to pronounce any statute, either of a state or of the United States, void, because irreconcilable with the constitution, except as it is called upon to adjudge the legal rights of litigants in actual 9 Case: 21-51178 Document: 00516474571 Page: 10 Date Filed: 09/16/2022 No. 21-51178 controversies.” Liverpool, N.Y. & Phila. S.S. Co. v. Comm’rs of Emigration, 113 U.S. 33, 39 (1885); accord Marbury v. Madison, 5 U.S. (1 Cranch) 137, 178 (1803). This limitation on federal jurisdiction to “actual controversies” prevents courts from “ancitipat[ing] a question of constitutional law in advance of the necessity of deciding it.” Liverpool, 113 U.S. at 39; see also Broadrick v. Oklahoma, 413 U.S. 601, 610–11 (1973) (“[U]nder our constitutional system courts are not roving commissions assigned to pass judgment on the validity of the Nation’s laws.”). And it makes pre- enforcement facial challenges a particularly nettlesome affair. Such suits usually do not present “flesh-and-blood legal problems with data relevant and adequate to an informed judgment.” New York v. Ferber, 458 U.S. 747, 768 (1982) (quotation omitted). Instead, they require the court “to consider every conceivable situation which might possibly arise in the application of complex and comprehensive legislation,” forcing courts to deploy the severe power of judicial review “with reference to hypothetical cases.” United States v. Raines, 362 U.S. 17, 21–22 (1960). Third, federalism. Invalidate-the-law-now, discover-how-it-works- later judging is particularly troublesome when reviewing state laws, as it deprives “state courts [of ] the opportunity to construe a law to avoid constitutional infirmities.” Ferber, 458 U.S. at 768. And “facial challenges threaten to short circuit the democratic process by preventing laws embodying the will of the people from being implemented in a manner consistent with the Constitution.” Wash. State Grange, 552 U.S. at 451. The respect owed to a sovereign State thus demands that we look particularly askance at a litigant who wants unelected federal judges to countermand the State’s democratically accountable policymakers. In accordance with the disfavor that attaches to pre-enforcement facial challenges, the legal standard for them is extraordinarily high. Ordinarily, plaintiffs bringing this sort of “facial challenge to a legislative 10 Case: 21-51178 Document: 00516474571 Page: 11 Date Filed: 09/16/2022 No. 21-51178 Act” must “establish that no set of circumstances exists under which the Act would be valid.” United States v. Salerno, 481 U.S. 739, 745 (1987); see also Americans for Prosperity Found. v. Bonta, 141 S. Ct. 2373, 2387 (2021). “Such a challenge is the most difficult to mount successfully.” City of El Cenizo v. Texas, 890 F.3d 164, 187 (5th Cir. 2018) (quotation omitted). The Platforms do not even try to show that HB 20 is “unconstitutional in all of its applications.” Wash. State Grange, 552 U.S. at 449. 4 Instead, their challenge is premised on First Amendment overbreadth doctrine. Under this doctrine, the Supreme Court has “recognized a second type of facial challenge, whereby a law may be invalidated as overbroad if a substantial number of its applications are unconstitutional, judged in relation to the statute’s plainly legitimate sweep.” Bonta, 141 S. Ct. at 2387 (quotation omitted). This doctrine is limited to “the First Amendment context.” Ibid. “Overbreadth is a judicially created doctrine designed to prevent the chilling of protected expression.” Massachusetts v. Oakes, 491 U.S. 576, 584 (1989) (plurality op.); see generally Lewis D. Sargentich, Note, The First Amendment Overbreadth Doctrine, 83 Harv. L. Rev. 844 (1970). As the seminal case explained, the overbreadth doctrine addresses “threat[s] to censure comments on matters of public concern.” Thornhill v. Alabama, 310 U.S. 88, 97 (1940). The doctrine’s rationale is that “[m]any persons, rather than undertake the considerable burden (and sometimes risk) of vindicating their rights through case-by-case litigation, will choose simply to abstain from protected speech—harming not only themselves but society as a whole, 4 For example, the Platforms do not argue that HB 20’s provision restricting censorship based on “a user’s geographic location in [Texas]” could not be constitutionally applied to them. Tex. Civ. Prac. & Rem. Code § 143A.002(a)(3). While they vigorously argue that viewpoint-based censorship is protected speech, they nowhere contend that the First Amendment protects censorship based on geographic location. 11 Case: 21-51178 Document: 00516474571 Page: 12 Date Filed: 09/16/2022 No. 21-51178 which is deprived of an uninhibited marketplace of ideas.” Virginia v. Hicks, 539 U.S. 113, 119 (2003) (citation omitted). Consistent with the overbreadth doctrine’s rationale, the Supreme Court has only applied it where there is a substantial risk that the challenged law will chill protected speech or association. See, e.g., Bigelow v. Virginia, 421 U.S. 809, 817–18 (1975) (declining to address facial overbreadth challenge where statutory amendment removed risk that statute “will chill the rights of others”); Law Students C.R. Rsch. Council, Inc. v. Wadmond, 401 U.S. 154, 167 (1971) (denying facial relief where “careful administration” of state regulatory scheme could avoid “chilling effects upon the exercise of constitutional freedoms”). The Court has also instructed that “the overbreadth doctrine is strong medicine” that should be employed “only as a last resort.” L.A. Police Dep’t v. United Reporting Pub. Corp., 528 U.S. 32, 39 (1999) (quotation omitted). And the overbreadth doctrine’s function “attenuates” as the regulated expression moves from “pure speech toward conduct.” Id. at 40 (quotation omitted). 2. The overbreadth doctrine does not apply to Section 7. That’s for three reasons. First, the primary concern of overbreadth doctrine is to avoid chilling speech. But Section 7 does not chill speech; instead, it chills censorship. So there can be no concern that declining to facially invalidate HB 20 will inhibit the marketplace of ideas or discourage commentary on matters of public concern. Perhaps as-applied challenges to speculative, now-hypothetical enforcement actions will delineate boundaries to the law. But in the meantime, HB 20’s prohibitions on censorship will cultivate rather than stifle the marketplace of ideas that justifies the overbreadth doctrine in the first place. 12 Case: 21-51178 Document: 00516474571 Page: 13 Date Filed: 09/16/2022 No. 21-51178 The Platforms, of course, argue that their censorship somehow should be construed as speech for First Amendment purposes. We deal with this contention at length in Parts III.B, III.C, III.D, and III.E, infra. But even stipulating arguendo that censorship can enjoy First Amendment protection, it’s a far cry from the “pure speech” that’s the core concern of the overbreadth doctrine. See United Reporting, 528 U.S. at 40. At most, the Platforms’ censorship is, in the district court’s words, a “way that online services express themselves and effectuate their community standards.” That is, censorship is at best a form of expressive conduct, for which the overbreadth doctrine provides only “attenuate[d]” protection. Ibid. (quotation omitted); see also Broadrick, 413 U.S. at 614 (“[O]verbreadth scrutiny has generally been somewhat less rigid in the context of statutes regulating conduct in the shadow of the First Amendment, but doing so in a neutral, noncensorial manner.”). Tellingly, the Platforms have pointed to no case applying the overbreadth doctrine to protect censorship rather than speech. To the contrary, the Platforms principally rely on three cases. See Miami Herald, 418 U.S. 241; PG&E v. Public Utilities Commission of California, 475 U.S. 1 (1986); and Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston, 515 U.S. 557 (1995). But all three involved challenges to concrete applications of an allegedly unconstitutional law, raised by a defendant in state court proceedings. So even if these cases supported the Platforms’ argument about their substantive First Amendment rights, they would provide no support for the Platforms’ attempt to use the First Amendment as a sword to facially invalidate a law before it has been applied to anyone under any circumstances. Second, overbreadth adjudication is meant to protect third parties who cannot “undertake the considerable burden” of as-applied litigation and whose speech is therefore likely to be chilled by an overbroad law. Hicks, 539 U.S. at 119; see also United States v. Sineneng-Smith, 140 S. Ct. 1575, 1586 13 Case: 21-51178 Document: 00516474571 Page: 14 Date Filed: 09/16/2022 No. 21-51178 (2020) (Thomas, J., concurring) (explaining that overbreadth doctrine “allow[s] individuals to challenge a statute based on a third party’s constitutional rights”). Courts have deemed this chilling effect on third parties particularly worrisome when the overbroad law imposes criminal sanctions. See, e.g., Gooding v. Wilson, 405 U.S. 518, 521 (1972). This rationale for overbreadth adjudication is wholly inapposite here. First of all, there are no third parties to chill. The plaintiff trade associations represent all the Platforms covered by HB 20. Additionally, unlike individual citizens potentially subject to criminal sanctions—the usual beneficiaries of overbreadth rulings—the entities subject to HB 20 are large, well-heeled corporations that have hired an armada of attorneys from some of the best law firms in the world to protect their censorship rights. And any fear of chilling is made even less credible by HB 20’s remedial scheme. Not only are criminal sanctions unavailable; damages are unavailable. It’s hard to see how the Platforms—which have already shown a willingness to stand on their rights—will be so chilled by the prospect of declaratory and injunctive relief that a facial remedy is justified. Third, the Platforms principally argue against HB 20 by speculating about the most extreme hypothetical applications of the law. Such whataboutisms further exemplify why it’s inappropriate to hold the law facially unconstitutional in a pre-enforcement posture. Texas enacted HB 20 to address “the Platforms’ evolution into internet censors.” Explaining the perceived need for the law, Texas and its amici cite numerous instances in which the Platforms have censored what Texas contends is pure political speech. For example, one amicus brief documents the Platforms’ censorship of fifteen prominent celebrities and political figures—including five holding federal elected office. See Brief for Amici Curiae The Babylon Bee, LLC, et al. at 26–38. Texas also points to the 14 Case: 21-51178 Document: 00516474571 Page: 15 Date Filed: 09/16/2022 No. 21-51178 Platforms’ “discriminat[ion] against Americans and in favor of foreign adversaries” and censorship of even a congressional hearing that featured disfavored viewpoints. The Platforms do not directly engage with any of these concerns. Instead, their primary contention—beginning on page 1 of their brief and repeated throughout and at oral argument—is that we should declare HB 20 facially invalid because it prohibits the Platforms from censoring “pro-Nazi speech, terrorist propaganda, [and] Holocaust denial[s].” Red Br. at 1. Far from justifying pre-enforcement facial invalidation, the Platforms’ obsession with terrorists and Nazis proves the opposite. The Supreme Court has instructed that “[i]n determining whether a law is facially invalid,” we should avoid “speculat[ing] about ‘hypothetical’ or ‘imaginary’ cases.” Wash. State Grange, 552 U.S. at 449–50. Overbreadth doctrine has a “tendency . . . to summon forth an endless stream of fanciful hypotheticals,” and this case is no exception. United States v. Williams, 553 U.S. 285, 301 (2008). But it’s improper to exercise the Article III judicial power based on “hypothetical cases thus imagined.” Raines, 362 U.S. at 22; cf. Sineneng- Smith, 140 S. Ct. at 1585–86 (Thomas, J., concurring) (explaining the tension between overbreadth adjudication and the constitutional limits on judicial power). If we focus instead on “the statute’s facial requirements,” Wash. State Grange, 552 U.S. at 450, its language renders implausible many of the Platforms’ extreme hypothesized applications of the law. HB 20 expressly permits the Platforms to censor any unlawful expression and certain speech that “incites criminal activity or consists of specific threats”—not to mention any content the Platforms are authorized to censor by federal law. Tex. Civ. Prac. & Rem. Code § 143A.006(a). So at a minimum, we should avoid “determin[ing] the constitutionality of [HB 20] in hypothetical 15 Case: 21-51178 Document: 00516474571 Page: 16 Date Filed: 09/16/2022 No. 21-51178 situations where it is not even clear the State itself would consider its law applicable.” Morales v. Trans World Airlines, Inc., 504 U.S. 374, 382 (1992). Or as one amicus puts it, the Platforms at this early stage may not use borderline hypotheticals involving vile expression to pretermit consideration of “what actually is at stake—namely, the suppression of domestic political, religious, and scientific dissent.” Brief of Amicus Curiae Prof. Philip Hamburger at 21. In short, Section 7 chills no speech whatsoever. To the extent it chills anything, it chills censorship. That is, Section 7 might make censors think twice before removing speech from the Platforms in a viewpoint- discriminatory manner. But we cannot find any cases, from any court, that suggest a would-be censor can bring a First Amendment overbreadth challenge because a regulation chills its efforts to prohibit others from speaking. B. We turn now to the merits of the Platforms’ First Amendment claim. As always, we start with the original public meaning of the Constitution’s text. We need not tarry long here because the Platforms—by pointing to no evidence whatsoever on this point—do not contend that the First Amendment’s history and original understanding provide any basis for invalidating Section 7. The First Amendment prevents the government from enacting laws “abridging the freedom of speech, or of the press.” U.S. Const. amend. I; see Gitlow v. New York, 268 U.S. 652 (1925) (incorporating this right against the States). At the Founding and “[f ]or most of our history, speech and press freedoms entailed two common-law rules—first, a prohibition on prior restraints and, second, a privilege of speaking in good faith on matters of public concern.” Jud Campbell, The Emergence of Neutrality, 131 Yale L.J. 16 Case: 21-51178 Document: 00516474571 Page: 17 Date Filed: 09/16/2022 No. 21-51178 861, 874–75 (2022). The first rule was central to the Speech Clause as originally understood, because the “core abuse against which it was directed was the scheme of licensing laws implemented by the monarch and Parliament to contain the ‘evils’ of the printing press in 16th- and 17-century England.” Thomas v. Chicago Park Dist., 534 U.S. 316, 320 (2002). For example, the Printing Act of 1662 required all printers to obtain a license and then “required that all works be submitted for approval to a government official, who wielded broad authority to suppress works that he found to be heretical, seditious, schismatical, or offensive.” Ibid. (quotation omitted). Licensing schemes like the Printing Act generated substantial opposition in both England and the American colonies. They disappeared in both places by the 1720s. See David S. Bogen, The Origins of Freedom of Speech and Press, 42 Md. L. Rev. 429, 443–44 (1983). Thus, Blackstone had this to say two decades before the First Amendment’s ratification: The liberty of the press is indeed essential to the nature of a free state: but this consists in laying no previous restraints upon publications, and not in freedom from censure for criminal matter when published. . . . To subject the press to the restrictive power of a licenser, as was formerly done, both before and since the revolution, is to subject all freedom of sentiment to the prejudices of one man, and make him the arbitrary and infallible judge of all controverted points in learning, religion, and government. 4 William Blackstone, Commentaries *151–52. Founding-era Americans similarly viewed the freedom from prior restraints as a central component of the freedoms of speech and the press. See Campbell, Emergence of Neutrality, supra, at 875–76; see also, e.g., 3 Joseph Story, Commentaries on the Constitution § 1874 (1833) (“It is plain, then, that the language of [the First A]mendment imports no more, than that 17 Case: 21-51178 Document: 00516474571 Page: 18 Date Filed: 09/16/2022 No. 21-51178 every man shall have a right to speak, write, and print his opinions upon any subject whatsoever, without any prior restraint . . . .”). As originally understood, the First Amendment’s Speech and Press Clauses also protected the freedom to make well-intentioned statements of one’s thoughts, particularly on matters of public concern. See generally Jud Campbell, Natural Rights and the First Amendment, 127 Yale L.J. 246, 280– 87 (2017). States recognized this freedom before the First Amendment’s ratification. 5 The Anti-Federalists worked to protect it in the federal Constitution. 6 And even the Federalists—who were generally less friendly to the freedom of speech—recognized that the First Amendment protected this right. See id. at 286; see also, e.g., 8 Annals of Cong. 2148 (1798) (statement of Rep. Harrison Gray Otis) (recognizing that the First Amendment protects “the liberty of writing, publishing, and speaking, one’s 5 For example, in 1788, Chief Justice McKean of the Supreme Court of Pennsylvania explained that “[t]he true liberty of the press is amply secured by permitting every man to publish his opinions; but it is due to the peace and dignity of society to enquire into the motives of such publications, and to distinguish between those which are meant for use and reformation, and with an eye solely to the public good, and those which are intended merely to delude and defame.” Respublica v. Oswald, 1 U.S. (1 Dall.) 319, 325 (Pa. 1788). This statement illustrates both facets of the First Amendment’s original public meaning. First, prior restraints were prohibited, full stop: “[E]very man [may] publish his opinions.” Ibid. Second, whether post-publication liability could be imposed depended on whether an opinion was “meant for use and reformation . . . [or] merely to delude and defame”—to use modern terminology, whether the statement was made in good faith. Ibid. 6 See, e.g., Centinel No. 1, in 2 The Complete Anti-Federalist 136, 136 (Herbert J. Storing ed., 1981) (urging the People to demand constitutional protection for “a right of freedom of speech”). Thomas Jefferson also wrote to James Madison—who later drafted the Bill of Rights—that he thought the Constitution should ensure “[t]he people shall not be deprived or abridged of their right to speak to write or otherwise to publish any thing but false facts affecting injuriously the life, property, or reputation of others or affecting the peace of the confederacy with foreign nations.” Letter from Thomas Jefferson to James Madison (Aug. 28, 1789), in 5 The Founders’ Constitution 129, 129–30 (Philip B. Kurland & Ralph Lerner eds., 1987). 18 Case: 21-51178 Document: 00516474571 Page: 19 Date Filed: 09/16/2022 No. 21-51178 thoughts, under the condition of being answerable . . . for false, malicious, and seditious expressions, whether spoken or written”). The Platforms neither challenge this understanding of the First Amendment’s original meaning nor suggest that Section 7 runs afoul of it. This apparent concession is unsurprising. First, Section 7 does not operate as a prior restraint on the Platforms’ speech—even if one accepts their characterization of censorship as speech. Recall Blackstone’s criticism of prior restraints: that they “subject all freedom of sentiment to the prejudices of one man, and make him the arbitrary and infallible judge of all controverted points in learning, religion, and government.” 4 Blackstone, supra, at *151–52. The Platforms operate “the modern public square,” Packingham, 137 S. Ct. at 1737, and it is they—not the government—who seek to defend viewpoint-based censorship in this litigation. Second, Section 7 does not prevent anyone from expressing their good-faith opinions on matters of public concern. Precisely the opposite: Section 7 protects Texans’ ability to freely express a diverse set of opinions through one of the most important communications mediums used in that State. And it leaves the Platforms free to similarly opine: They can still say whatever they want (or decline to say anything) about any post by any user. Moreover, Section 7’s exceptions—where viewpoint-based censorship is still permitted, like certain specific threats of violence—contemplate malicious, bad-faith speech not protected by the First Amendment as originally understood. See Campbell, Emergence of Neutrality, supra, at 878. So Section 7’s carveouts do nothing to impugn its constitutionality under the First Amendment’s original meaning. C. Rather than mount any challenge under the original public meaning of the First Amendment, the Platforms instead focus their attention on 19 Case: 21-51178 Document: 00516474571 Page: 20 Date Filed: 09/16/2022 No. 21-51178 Supreme Court doctrine. And under that doctrine, the Platforms contend, Section 7 somehow burdens their right to speak. How so, you might wonder? Section 7 does nothing to prohibit the Platforms from saying whatever they want to say in whatever way they want to say it. Well, the Platforms contend, when a user says something using one of the Platforms, the act of hosting (or rejecting) that speech is the Platforms’ own protected speech. Thus, the Platforms contend, Supreme Court doctrine affords them a sort of constitutional privilege to eliminate speech that offends the Platforms’ censors. We reject the Platforms’ efforts to reframe their censorship as speech. It is undisputed that the Platforms want to eliminate speech—not promote or protect it. And no amount of doctrinal gymnastics can turn the First Amendment’s protections for free speech into protections for free censoring. We (1) explain the relevant doctrine and Supreme Court precedent. Then we (2) hold this precedent forecloses the Platforms’ argument that Section 7 is unconstitutional. 1. Supreme Court precedent instructs that the freedom of speech includes “the right to refrain from speaking at all.” Wooley v. Maynard, 430 U.S. 705, 714 (1977); see also W. Va. State Bd. of Educ. v. Barnette, 319 U.S. 624, 642 (1943). So the State may not force a private speaker to speak someone’s else message. See Wooley, 430 U.S. at 714. But the State can regulate conduct in a way that requires private entities to host, transmit, or otherwise facilitate speech. Were it otherwise, no government could impose nondiscrimination requirements on, say, telephone companies or shipping services. But see 47 U.S.C. § 202(a) (prohibiting telecommunications common carriers from “mak[ing] any unjust or unreasonable discrimination in charges, practices, classifications, 20 Case: 21-51178 Document: 00516474571 Page: 21 Date Filed: 09/16/2022 No. 21-51178 regulations, facilities, or services”). Nor could a State create a right to distribute leaflets at local shopping malls. But see PruneYard Shopping Ctr. v. Robins, 447 U.S. 74, 88 (1980) (upholding a California law protecting the right to pamphleteer in privately owned shopping centers). So First Amendment doctrine permits regulating the conduct of an entity that hosts speech, but it generally forbids forcing the host itself to speak or interfering with the host’s own message. Five Supreme Court cases elucidate this distinction. The first is Miami Herald. It involved a Florida law providing that when a newspaper article criticizes the character or record of a political candidate, the newspaper must offer the candidate equal space in the paper to reply to the criticism. 418 U.S. at 244. The Court held that this “right-of-reply” law violated the First Amendment. Id. at 258. The Court explained that the law interfered with the newspaper’s speech by imposing a content-based penalty on it. See id. at 256 (“The Florida statute exacts a penalty on the basis of the content of a newspaper.”). If the newspaper chose to speak about most topics, there was no penalty— but if it spoke critically about a political candidate, it was penalized with the “cost in printing and composing time and materials” necessary to give the candidate a free and equally prominent response column. Ibid. Moreover, the reply would “tak[e] up space that could be devoted to other material the newspaper may have preferred to print.” Ibid. This interference would disincentivize the newspaper’s speech: Faced with these penalties, “editors might well conclude that the safe course is to avoid controversy” and reduce coverage of political candidates altogether. Id. at 257. The Court also concluded that the right-of-reply law impermissibly compelled the newspaper to speak messages it opposed. As the Court explained: 21 Case: 21-51178 Document: 00516474571 Page: 22 Date Filed: 09/16/2022 No. 21-51178 A newspaper is more than a passive receptacle or conduit for news, comment, and advertising. 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. Id. at 258. Because a newspaper prints a curated set of material selected by its editors, everything it publishes is, in a sense, the newspaper’s own speech. And the newspaper has a right to “editorial control and judgment” over its speech. Ibid. Newspapers thus cannot be compelled to “publish that which reason tells them should not be published.” Id. at 256 (quotation omitted). The second case is PruneYard. That case involved a group of high school students who sought to distribute pamphlets and solicit signatures at a local shopping mall. The California Supreme Court held that California law protected the right to “speech and petitioning, reasonably exercised, in shopping centers even when the centers are privately owned.” 447 U.S. at 78 (quotation omitted). The mall objected on First Amendment grounds, arguing that “a private property owner has a First Amendment right not to be forced by the State to use his property as a forum for the speech of others.” Id. at 85. The Supreme Court rejected the shopping mall’s challenge. It found the state law exacted no penalty on the basis of the mall’s speech, and the mall could “expressly disavow any connection with the [pamphleteers’] message by simply posting signs in the area where the speakers or handbillers st[oo]d.” Id. at 87–88. Nor did California law impermissibly compel the mall itself to speak. To the contrary, because the mall was open to anyone, “[t]he views expressed by members of the public in passing out pamphlets or seeking signatures . . . will not likely be identified with those of the owner.” Id. at 87. The Court also emphasized California’s neutrality among 22 Case: 21-51178 Document: 00516474571 Page: 23 Date Filed: 09/16/2022 No. 21-51178 viewpoints: Because “no specific message is dictated by the State to be displayed on appellants’ property,” there was “no danger of governmental discrimination for or against a particular message.” Ibid. The mall relied in part on Miami Herald, but the PruneYard Court easily found that case inapplicable. The Court stated that Miami Herald “rests on the principle that the State cannot tell a newspaper what it must print,” and it emphasized the “danger in [Miami Herald] that the statute would dampen the vigor and limit the variety of public debate by deterring editors from publishing controversial political statements that might trigger the application of the statute.” Id. at 88. Those concerns were “obviously . . . not present” in PruneYard. Ibid. The third case is PG&E. A utility company, PG&E, had a longstanding practice of including a monthly newsletter in its billing envelopes. 475 U.S. at 5 (plurality op.). “In appearance no different from a small newspaper,” the newsletter included political editorials and stories on matters of public interest alongside tips on energy conservation and information about utility services. Id. at 5, 8. Concerned that the expense of PG&E’s political speech was falling on customers, the California Public Utilities Commission (“Commission”) decided to apportion the billing envelopes’ “extra space”—that is, the space occupied by the company’s newsletter—and permit a third-party group representing PG&E ratepayers to use that space for its opposing messages four months per year. Id. at 5–6. PG&E objected, arguing that the First Amendment prevented the Commission from forcing it to include an adverse party’s speech in its billing envelopes. The Supreme Court ruled for PG&E. A plurality held that the Commission’s order both interfered with PG&E’s own speech and impermissibly forced it to associate with the views of other speakers. As in Miami Herald, the “one-sidedness” of the Commission’s order penalized 23 Case: 21-51178 Document: 00516474571 Page: 24 Date Filed: 09/16/2022 No. 21-51178 and disincentivized PG&E’s expression by awarding space only to those who disagreed with PG&E’s speech: [B]ecause access is awarded only to those who disagree with appellant’s [PG&E’s] views and who are hostile to appellant’s interests, appellant must contend with the fact that whenever it speaks out on a given issue, it may be forced—at [a third- party’s] discretion—to help disseminate hostile views. Appellant “might well conclude” that, under these circumstances, “the safe course is to avoid controversy,” thereby reducing the free flow of information and ideas that the First Amendment seeks to promote. Id. at 14 (quoting Miami Herald, 418 U.S. at 257). The plurality also found that the Commission’s order impermissibly “require[d] [PG&E] to associate with speech with which [it] may disagree.” Id. at 15. Because the third party could “use the billing envelopes to discuss any issues it chooses,” PG&E “may be forced either to appear to agree . . . or to respond.” Ibid. “That kind of forced response is antithetical to the free discussion that the First Amendment seeks to foster.” Id. at 16. Finally, the PG&E plurality found PruneYard distinguishable for two reasons. First, PruneYard did not involve a concern that the challenged law “might affect the shopping center owner’s exercise of his own right to speak.” Id. at 12. Second, the right of access at issue in PruneYard was not content-based. Ibid. 7 7 Justice Marshall provided the fifth vote to invalidate the Commission’s order. See PG&E, 475 U.S. at 21 (Marshall, J., concurring in the judgment). He emphasized two ways in which the Commission’s order was different from the law upheld in PruneYard. First, the right of access created by the Commission was more intrusive than the one upheld in PruneYard. That’s because the shopping mall owner in PruneYard had 24 Case: 21-51178 Document: 00516474571 Page: 25 Date Filed: 09/16/2022 No. 21-51178 The fourth case is Hurley. GLIB, an organization of Irish-American gay, lesbian, and bisexual individuals, sought to march in a St. Patrick’s Day parade in Boston. 515 U.S. at 561. The parade was organized by a private group, the South Boston Allied War Veterans Council (“Council”). Id. at 560. The Council refused to admit GLIB, citing “traditional religious and social values.” Id. at 562 (quotation omitted). But the Supreme Judicial Court of Massachusetts held that the parade was a public accommodation under state law, so the Council had to let GLIB participate. Id. at 564. The Council argued that this application of Massachusetts’s public accommodation law violated the First Amendment, and the Supreme Court agreed. Id. at 566. The Court concluded that the parade was a “form of expression” that receives First Amendment protection. Id. at 568. That’s because “[r]ather like a composer, the Council selects the expressive units of the parade from potential participants, and though the score may not produce a particularized message, each contingent’s expression in the Council’s eyes comports with what merits celebration on that day.” Id. at 574. And it didn’t matter that the Council was “rather lenient in admitting participants,” because “a private speaker does not forfeit constitutional protection simply by combining multifarious voices, or by failing to edit their themes to isolate an exact message as the exclusive subject matter of the speech.” Id. at 569–70. voluntarily opened his property up to the public, whereas PG&E “has never opened up its billing envelope to the use of the public.” Id. at 22. Second, in PruneYard, the speech of the shopping mall owner was not “hindered in the slightest” by the public’s pamphleteering right. Id. at 24. PG&E, by contrast, involved “a forum of inherently limited scope,” such that the State’s appropriation of that forum for a third party’s use necessarily curtailed PG&E’s ability to speak in that forum. Ibid. And this interference with PG&E’s speech could not be justified by the State’s goal of “subsidiz[ing] . . . another speaker chosen by the State.” Ibid. 25 Case: 21-51178 Document: 00516474571 Page: 26 Date Filed: 09/16/2022 No. 21-51178 The cornerstone of the Court’s reasoning was that the parade sponsors were “intimately connected” to the message communicated by the parade. Id. at 576. This intimate connection was crucial, the Court held, because forcing the sponsors to include a particular float was tantamount to forcing the sponsors to speak: “[W]hen dissemination of a view contrary to one’s own is forced upon a speaker intimately connected with the communication advanced, the speaker’s right to autonomy over the message is compromised.” Id. at 576; see also id. at 573 (emphasizing that “a speaker has the autonomy to choose the content of his own message,” including by “decid[ing] what not to say”) (quotation omitted). The final case that’s particularly relevant to our discussion is Rumsfeld v. Forum for Academic and Institutional Rights, Inc., 547 U.S. 47 (2006). Certain law schools sought to restrict military recruiting on their campuses because of the military’s policies on sexual orientation. Id. at 51. Congress responded by enacting the Solomon Amendment, which denied federal funding to schools that did not give military recruiters “access to students that is at least equal in quality and scope to the access provided other potential employers.” Id. at 54 (quotation omitted). An organization of law schools sued, arguing that the Solomon Amendment violated the First Amendment. The Supreme Court disagreed. It unanimously held that “the First Amendment would not prevent Congress from directly imposing the Solomon Amendment’s access requirement,” and the statute thus did not place an unconstitutional condition on the receipt of federal funds. Id. at 60. The Court first held that the Solomon Amendment did not impermissibly force the law schools to speak. Id. at 61–62. The Court recognized that “recruiting assistance provided by the schools often includes elements of speech”—like sending emails or posting bulletin board notices on the recruiter’s behalf. Id. at 61. But the Court determined that this speech was “plainly incidental to the Solomon Amendment’s regulation of 26 Case: 21-51178 Document: 00516474571 Page: 27 Date Filed: 09/16/2022 No. 21-51178 conduct” and was nothing like a “Government-mandated pledge or motto” as in Barnette and Wooley. Id. at 62. Congress could therefore compel this “incidental” speech without violating the First Amendment. Ibid. The Court then held that the Solomon Amendment did not impermissibly interfere with the schools’ own speech, distinguishing Miami Herald, PG&E, and Hurley. Id. at 63–65. It acknowledged that those three cases “limited the government’s ability to force one speaker to host or accommodate another speaker’s message.” Id. at 63. But it then explained that these “compelled-speech violation[s] . . . resulted from the fact that the complaining speaker’s own message was affected by the speech it was forced to accommodate.” Ibid.; see also id. at 63–64 (explaining how the challenged laws “interfere[d] with a speaker’s desired message” in Miami Herald, PG&E, and Hurley). In Rumsfeld, by contrast, “accommodating the military’s message [did] not affect the law schools’ speech, because the schools [were] not speaking when they host interviews and recruiting receptions.” Id. at 64. That was true despite the risk that students might mistakenly interpret the law schools’ conduct as sending the message that they see nothing wrong with the military’s policies. Id. at 64–65. In sum, even though it required law schools to host and accommodate others’ speech, the Solomon Amendment was constitutional because it “neither limit[ed] what law schools may say nor require[d] them to say anything.” Id. at 60. 2. Under these precedents, a speech host must make one of two showings to mount a First Amendment challenge. It must show that the challenged law either (a) compels the host to speak or (b) restricts the host’s own speech. The Platforms cannot make either showing. And (c) the Platforms’ counterarguments are unpersuasive. 27 Case: 21-51178 Document: 00516474571 Page: 28 Date Filed: 09/16/2022 No. 21-51178 a. Let’s start with compelled speech. In Miami Herald, the Supreme Court held that Florida’s right-of-reply law was unconstitutional because it compelled newspapers to speak. Crucially, the Court emphasized that “[a] newspaper is more than a passive receptable or conduit for news, comment, or advertising.” Miami Herald, 418 U.S. at 258. Rather, a newspaper curates and publishes a narrow “choice of material” in accordance with the “editorial control and judgment” of its editors. Ibid. Thus, when a newspaper affirmatively chooses to publish something, it says that particular speech— at the very least—should be heard and discussed. So forcing a newspaper to run this or that column is tantamount to forcing the newspaper to speak. The Platforms are nothing like the newspaper in Miami Herald. Unlike newspapers, the Platforms exercise virtually no editorial control or judgment. The Platforms use algorithms to screen out certain obscene and spam-related content. 8 And then virtually everything else is just posted to the Platform with zero editorial control or judgment. “Something well north of 99% of th[is] content . . . never gets reviewed further. The content on a site is, to that extent, invisible to the [Platform].” NetChoice, LLC v. Moody, 546 F. Supp. 3d 1082, 1092 (N.D. Fla. 2021). Thus the Platforms, unlike newspapers, are primarily “conduit[s] for news, comment, and advertising.” Miami Herald, 418 U.S. at 258. And that’s why the Supreme Court has described them as “the modern public square.” Packingham, 137 S. Ct. at 1737; see also Biden v. Knight First Amend. Inst., 141 S. Ct. 1220, 1224 (2021) (Thomas, J., concurring) (noting Platforms are also “unlike newspapers” in that they 8 The Platforms have disclosed little about their algorithms in this appeal, other than suggesting that they “often moderate certain policy-violating content before users see it.” The Platforms never suggest their algorithms somehow exercise substantive, discretionary review akin to newspaper editors. 28 Case: 21-51178 Document: 00516474571 Page: 29 Date Filed: 09/16/2022 No. 21-51178 “hold themselves out as organizations that focus on distributing the speech of the broader public”). The Platforms’ own representations confirm this. 9 They’ve told their users: “We try to explicitly view ourselves as not editors. . . . We don’t want to have editorial judgment over the content that’s in your feed.” 10 They’ve told the public that they “may not monitor,” “do not endorse,” and “cannot take responsibility for” the content on their Platforms. 11 They’ve told Congress that their “goal is to offer a platform for all ideas.” 12 And they’ve told courts—over and over again—that they simply “serv[e] as conduits for other parties’ speech.” 13 9 To the extent that these representations vary between Platforms, that further cuts against the propriety of this facial, pre-enforcement challenge. Cf. supra Part III.A. To establish associational standing, the plaintiff trade associations asserted in the district court that this suit “does not require individualized facts about any particular covered social media platform.” ROA.645; see also Tex. Ass’n of Mfrs. v. U.S. Consumer Prod. Safety Comm’n, 989 F.3d 368, 377 (5th Cir. 2021) (stating the relevant rule). So the Platforms may not now rely on individualized facts to claim that, for example, one Platform operates like a newspaper even if the others don’t. 10 Ravi Somaiya, How Facebook Is Changing the Way Its Users Consume Journalism, N.Y. Times, Oct. 26, 2014, https://nyti.ms/3ommZXb. 11 Twitter, Terms of Service § 3, https://twitter.com/en/tos (last visited Aug. 6, 2022) [hereinafter Twitter Terms]; see also Facebook, Terms of Service § 4.3, https://www.facebook.com/terms.php (last visited Aug. 6, 2022) [hereinafter Facebook Terms] (“We are not responsible for [users’] actions or conduct . . . or any content they share.”); YouTube, Terms of Service, https://www.youtube.com/static?template=terms (last visited Aug. 6, 2022) (“Content is the responsibility of the person or entity that provides it to [YouTube].”). 12 Online Platforms and Market Power, Part 6: Hearing Before the Subcomm. on Antitrust, Com. and Admin. Law of the H. Comm. on the Judiciary, 116th Cong. 33 (2020) (testimony of Mark Zuckerberg, CEO, Facebook, Inc.). 13 Brief for Appellees at 1, Klayman v. Zuckerberg, No. 13-7017 (D.C. Cir. Oct. 25, 2013); see also, e.g., Notice of Motion and Motion to Dismiss at 10 n.5, Fields v. Twitter, Inc., 29 Case: 21-51178 Document: 00516474571 Page: 30 Date Filed: 09/16/2022 No. 21-51178 It is no answer to say, as the Platforms do, that an observer might construe the act of hosting speech as an expression of support for its message. That was the precise contention the Court rejected in both PruneYard and Rumsfeld: Neither the shopping mall nor the law schools wanted to endorse the hosted speech. The Rumsfeld Court dismissed that concern out of hand because even schoolchildren know the difference between sponsoring speech and allowing it. See 547 U.S. at 65 (citing Bd. of Educ. of Westside Cmty. Schs. (Dist. 66) v. Mergens, 496 U.S. 226, 250 (1990)). That’s precisely why even the Platforms concede that “objective observer[s]” would not “conclude that [Platforms] intended . . . to promote terrorism” when they host terrorist content. Motion to Dismiss at 23, Gonzalez v. Twitter, Inc., No. 4:16-cv-03282 (N.D. Cal. Jan. 13, 2017). Recognizing that their compelled-speech analogy to newspapers is a stretch, the Platforms turn to parades and the Hurley case. The Platforms contend that Section 7 forces them to host speech that’s inconsistent with their corporate “values.” But of course, the Platforms do not contend that they carefully curate users’ speech the way a parade sponsor or composer “selects . . . expressive units . . . from potential participants.” Hurley, 515 U.S. at 568. Nor do they suggest that they are “intimately connected with the communication” Section 7 requires them to host. Id. at 576. The Platforms instead contend that their censorship is protected because Hurley creates a freewheeling right for speech hosts to discriminate against messages they don’t like. Hurley said nothing of the sort. The Court instead carefully limited its holding to a speech host (like a parade organizer or composer) who is “intimately connected” with the hosted speech (like a parade or a No. 3:16-cv-00213 (N.D. Cal. Apr. 6, 2016) (stating Twitter is “a service provider acting as a conduit for huge quantities of third-party speech”). 30 Case: 21-51178 Document: 00516474571 Page: 31 Date Filed: 09/16/2022 No. 21-51178 symphony). Ibid. And the Platforms are nothing like such hosts. They don’t pick content to “mak[e] some sort of collective point,” even an abstract one like “what merits celebration on [St. Patrick’s] day.” Id. at 568, 574. Rather, the Platforms permit any user who agrees to their boilerplate terms of service to communicate on any topic, at any time, and for any reason. And as noted above, virtually none of this content is meaningfully reviewed or edited in any way. Nor can the Platforms point to the content they do censor and claim that makes them akin to parade organizers. In Rumsfeld, for example, the law schools argued that their denial of access to military recruiters was protected expressive conduct because it “expressed” the schools’ disagreement with the military. 547 U.S. at 66. But the Court held that the denial of access was not inherently expressive, because such conduct would only be understood as expressive in light of the law schools’ speech explaining it. See ibid. Otherwise, observers wouldn’t know that the denial of access stemmed from an ideological disagreement—they might instead conclude, for example, that “the military recruiters decided for reasons of their own that they would rather interview someplace else.” Ibid. The same reasoning applies here. 14 If a Platform censors a user’s post, the expressive quality of that censorship arises only from the Platform’s 14 To be clear, unlike in Rumsfeld, the Platforms in this case never argue that their acts of censorship constitute “expressive conduct.” Cf., e.g., Texas v. Johnson, 491 U.S. 397, 404 (1989) (noting that expressive conduct may be protected by the First Amendment if the actor both has “an intent to convey a particularized message” and “the likelihood [is] great that the message would be understood by those who viewed it” (quotation omitted)). In fact, the phrase “expressive conduct” never even appears in their brief before our court. Compare infra at 82 n.41 (noting that the Platforms made such an argument before the Eleventh Circuit). But to the extent any such argument is latent in their reliance on Hurley or their claim of protected “editorial discretion,” it’s plainly foreclosed by the 31 Case: 21-51178 Document: 00516474571 Page: 32 Date Filed: 09/16/2022 No. 21-51178 speech (whether on an individualized basis or in its terms of service) stating that the Platform chose to censor the speech and explaining how the censorship expresses the Platform’s views. Otherwise, as in Rumsfeld, an observer might just as easily infer that the user himself deleted the post and chose to speak elsewhere. In terms of the conduct’s inherent expressiveness, there is simply no plausible way to distinguish the targeted denial of access to only military recruiters in Rumsfeld from the viewpoint-based censorship regulated by HB 20. Section 7 does not compel the Platforms to speak. b. Nor does it do anything to prohibit the Platforms from speaking. That’s for three independent reasons. First, the Platforms have virtually unlimited space for speech, so Section 7’s hosting requirement does nothing to prohibit the Platforms from saying what they want to say. Contrariwise, both Miami Herald and PG&E involved “forum[s] of inherently limited scope”—a newspaper and newsletter with significant space constraints. PG&E, 475 U.S. at 24 (Marshall, J., concurring in the judgment). So when the State appropriated space in the newspaper or newsletter for a third party’s use, it necessarily curtailed the owner’s ability to speak in its own forum. See Miami Herald, 418 U.S. at 256 (“[T]he compelled printing . . . tak[es] up space that could be devoted to other material the newspaper may have preferred to print.”); see also Rumsfeld, 547 U.S. at 64 (explaining the results in Miami Herald and PG&E in these terms). Accordingly, when a “speaker’s own message [is] affected by the speech it [is] forced to accommodate,” the speaker may invoke the First Amendment to protect their own ability to speak. Rumsfeld, Supreme Court’s reasoning in Rumsfeld. Moreover, the Platforms never suggest that their censorship could “convey a particularized message.” See Johnson, 491 U.S. at 404. 32 Case: 21-51178 Document: 00516474571 Page: 33 Date Filed: 09/16/2022 No. 21-51178 547 U.S. at 63. By contrast, “space constraints on digital platforms are practically nonexistent”—unlike with newspapers, cable companies, and many of the other entities the Platforms invoke by analogy. Knight, 141 S. Ct. at 1226 (Thomas, J., concurring). For this reason, the Platforms can host users’ speech without giving up their power or their right to speak their own message(s). Second, the Platforms are free to say whatever they want to distance themselves from the speech they host. The Supreme Court has been very careful to limit forced-affiliation claims by speech hosts. After all, any speech host could always object that its accommodation for speech might be confused for a coerced endorsement of it. But the Court rejected that forced- affiliation argument in PruneYard, where the shopping mall owner was not required to affirm the pamphleteers’ expression in any way, and was “free to publicly dissociate [himself] from the views of the speakers or handbillers.” 447 U.S. at 88. Similarly, in Rumsfeld, the law schools argued “that if they treat military and nonmilitary recruiters alike . . . they could be viewed as sending the message that they see nothing wrong with the military’s policies.” 547 U.S. at 64–65. But the Supreme Court easily rejected this argument, because “[n]othing about recruiting suggests that law schools agree with any speech by recruiters, and nothing in the Solomon Amendment restricts what the law schools may say about the military’s policies.” Id. at 65. Rather, to win a forced-affiliation claim, the speech host must show that it’s “intimately connected with the communication” and hence cannot dissociate itself from it. Hurley, 515 U.S. at 576. Here, the Platforms remain free to expressly disavow, distance themselves from, or say whatever they want about any expression they host. For example, Platforms can add addenda or disclaimers—containing their own speech—to users’ posts. And many of them already do this, thus dramatically underscoring that Section 7 prohibits none of their speech. 33 Case: 21-51178 Document: 00516474571 Page: 34 Date Filed: 09/16/2022 No. 21-51178 Third, Section 7 does not impose a content-based penalty on the Platforms’ speech. Recall that the right-of-reply law in Miami Herald burdened newspapers with the duty to publish a response column if they published an article questioning the character or record of a political candidate. 418 U.S. at 244. As the PG&E plurality explained, this imposed a content-based penalty on the newspaper’s speech in two distinct senses: First, the penalty was “triggered by a particular category of newspaper speech”; and second, access “was awarded only to those who disagreed with the newspaper’s views.” 475 U.S. at 13; see also id. at 14 (explaining that the Commission’s order in PG&E was content-based in the second sense). Here, by contrast, no category of Platform speech can trigger any additional duty— or obviate an existing duty—under Section 7. And Section 7 does not create a special privilege for those who disagree with the Platforms’ views. Cf. id. at 14 (billing envelope space was awarded only to a single entity formed to oppose PG&E’s views). Rather, it gives the exact same protection to all Platform users regardless of their viewpoint. c. The Platforms do not seriously dispute any of this. Instead, they argue that Section 7 interferes with their speech by infringing their “right to exercise editorial discretion.” They reason as follows. Premise one is that “editorial discretion” is a separate, freestanding category of First- Amendment-protected expression. Premise two is that the Platforms’ censorship efforts constitute “editorial discretion.” Conclusion: Section 7 burdens the Platforms’ First Amendment rights by obstructing their censorship efforts. Both premises in that syllogism are flawed. Premise one is faulty because the Supreme Court’s cases do not carve out “editorial discretion” as a special category of First-Amendment-protected expression. Instead, the 34 Case: 21-51178 Document: 00516474571 Page: 35 Date Filed: 09/16/2022 No. 21-51178 Court considers editorial discretion as one relevant consideration when deciding whether a challenged regulation impermissibly compels or restricts protected speech. Take, for example, Turner Broadcasting System, Inc. v. FCC, 512 U.S. 622 (1994) (“Turner I”). There the Court noted a cable operator “exercis[es] editorial discretion over which stations or programs to include in its repertoire.” Id. at 636 (quotation omitted). For this reason, among others, the Court concluded that selecting a limited repertoire of cable channels to transmit constitutes First-Amendment-protected speech. See id. at 636–37. Similarly, Miami Herald emphasized newspapers’ “exercise of editorial control and judgment” to support its holding that their close affiliation with the speech they publish gives them the right not to publish “that which reason tells them should not be published.” 418 U.S. at 256, 258 (quotation omitted). But both cases treated editorial discretion as a relevant consideration supporting their legal conclusions about the presence or absence of protected speech. Neither case implied that editorial discretion is itself a freestanding category of constitutionally protected expression. 15 Accordingly, the Platforms cannot invoke “editorial discretion” as if uttering some sort of First Amendment talisman to protect their censorship. Were it otherwise, the shopping mall in PruneYard and law schools in Rumsfeld could have changed the outcomes of those cases by simply asserting a desire to exercise “editorial discretion” over the speech in their forums. Instead, the Platforms must show that Section 7 either coerces them to speak 15 The Platforms’ other cases ostensibly supporting premise one are even farther afield. Manhattan Community Access Corp. v. Halleck, 139 S. Ct. 1921 (2019), discussed the constitutional limits on editorial discretion in public forums and described the issue in this case as “[a] distinct question not raised here.” Id. at 1931 & n.2. And Arkansas Educational Television Commission v. Forbes, 523 U.S. 666 (1998), simply reiterated Turner I’s conclusion that cable operators’ selection and presentation of programming is speech for First Amendment purposes. Id. at 674. 35 Case: 21-51178 Document: 00516474571 Page: 36 Date Filed: 09/16/2022 No. 21-51178 or interferes with their speech. Of course, how the Platforms do or don’t exercise editorial control is relevant to this inquiry, as it was in Miami Herald and Turner I. But the Platforms can’t just shout “editorial discretion!” and declare victory. 16 Premise two of the Platforms’ syllogism is also faulty. Even assuming “editorial discretion” is a freestanding category of First-Amendment- protected expression, the Platforms’ censorship doesn’t qualify. Curiously, the Platforms never define what they mean by “editorial discretion.” (Perhaps this casts further doubt on the wisdom of recognizing editorial discretion as a separate category of First-Amendment-protected expression.) Instead, they simply assert that they exercise protected editorial discretion because they censor some of the content posted to their Platforms and use sophisticated algorithms to arrange and present the rest of it. But whatever the outer bounds of any protected editorial discretion might be, the Platforms’ censorship falls outside it. That’s for two independent reasons. First, an entity that exercises “editorial discretion” accepts reputational and legal responsibility for the content it edits. In the newspaper context, for instance, the Court has explained that the role of “editors and 16 Our esteemed colleague in dissent makes a similar argument with a different label. The dissent reads Miami Herald to protect “two levels of publisher speech”: the published speech itself as well as “the selection process” (or “publishing process”) used to choose that speech. Post, at 5–6, 11. And it concludes that Section 7 impermissibly interferes with the Platforms’ publishing process. Id. at 11. It’s of course true that the right to speak generally entails the right to select what to speak. But asserting that Section 7 obstructs the Platforms’ “selection process” begs the question whether the Platforms’ censorship is protected speech at all. If it’s not, then there’s no First Amendment right for censors to select their targets—just as there’s no First Amendment right for law schools to select their recruiters, no First Amendment right for shopping malls to select their pamphleteers, and no First Amendment right for telephone companies to select which calls to drop. 36 Case: 21-51178 Document: 00516474571 Page: 37 Date Filed: 09/16/2022 No. 21-51178 editorial employees” generally includes “determin[ing] the news value of items received” and taking responsibility for the accuracy of the items transmitted. Associated Press v. NLRB, 301 U.S. 103, 127 (1937). And editorial discretion generally comes with concomitant legal responsibility. For example, because of “a newspaper’s editorial judgments in connection with an advertisement,” it may be held liable “when with actual malice it publishes a falsely defamatory” statement in an ad. Pittsburgh Press Co. v. Pittsburgh Comm’n on Human Rels., 413 U.S. 376, 386 (1973). But the Platforms strenuously disclaim any reputational or legal responsibility for the content they host. See supra Part III.C.2.a (quoting the Platforms’ adamant protestations that they have no responsibility for the speech they host); infra Part III.D (discussing the Platforms’ representations pertaining to 47 U.S.C. § 230). Second, editorial discretion involves “selection and presentation” of content before that content is hosted, published, or disseminated. See Ark. Educ. Television Comm’n v. Forbes, 523 U.S. 666, 674 (1998); see also Miami Herald, 418 U.S. at 258 (a newspaper exercises editorial discretion when selecting the “choice of material” to print). The Platforms do not choose or select material before transmitting it: They engage in viewpoint-based censorship with respect to a tiny fraction of the expression they have already disseminated. The Platforms offer no Supreme Court case even remotely suggesting that ex post censorship constitutes editorial discretion akin to ex ante selection. 17 They instead baldly assert that “it is constitutionally irrelevant at what point in time platforms exercise editorial discretion.” Red 17 The Platforms claim Horton v. City of Houston, 179 F.3d 188 (5th Cir. 1999), recognized First Amendment rights for organizations that “do not pre-screen submitted programs.” Id. at 190. Horton is wholly irrelevant. It involved a public forum—a public access cable channel—and concerned the First Amendment rights of a different party seeking access to the forum. See id. at 190–91. 37 Case: 21-51178 Document: 00516474571 Page: 38 Date Filed: 09/16/2022 No. 21-51178 Br. at 25. Not only is this assertion unsupported by any authority, but it also illogically equates the Platforms’ ex post censorship with the substantive, discretionary, ex ante review that typifies “editorial discretion” in every other context. 18 In sum, even if “editorial discretion” is a protected legal category, it’s far from clear why (even viewpoint-agnostic) content arrangement and (even infrequent and ex post) censorship should be the criteria for qualification. And in any event, the Supreme Court has never recognized “editorial discretion” as a freestanding category of First-Amendment-protected expression. Rather, the applicable inquiry is whether Section 7 forces the Platforms to speak or interferes with their speech. Section 7 does neither of those things. It therefore passes constitutional muster. 18 Our esteemed colleague in dissent suggests that the timing of the Platforms’ censorship doesn’t matter because censorship decisions “can only be made, as a practical matter, after the appearance of the content on the Platform.” Post, at 13. The dissent’s factual premise is incorrect: Online platforms can and do moderate submissions before transmitting them. For example, the New York Times moderates online comments on its articles before posting them. See The Comments Section, N.Y. Times, https://help.nytimes.com/hc/en-us/articles/115014792387-The-Comments-Section (last visited Aug. 6, 2022). That’s arguably the same form of ex ante curation that newspapers use for other material they publish and that enjoys constitutional protection under Miami Herald. If the Platforms wanted the same protections, they could’ve used the same ex ante curation process. Early online forums and message boards often preapproved all submissions before transmission. See, e.g., Stratton Oakmont, Inc. v. Prodigy Servs. Co., 1995 WL 323710, at *3 (N.Y. Sup. Ct. May 24, 1995) (noting Prodigy’s early policy of “manually reviewing all messages prior to posting”). Later on, the Platforms made a judgment that jettisoning editorial discretion to allow instantaneous transmission would make their Platforms more popular, scalable, and commercially successful. The Platforms thus disclaimed ex ante curation—precisely because they wanted users to speak without editorial interference. That decision has consequences. And it reinforces that the users are speaking, not the Platforms. 38 Case: 21-51178 Document: 00516474571 Page: 39 Date Filed: 09/16/2022 No. 21-51178 D. We have no doubts that Section 7 is constitutional. But even if some were to remain, 47 U.S.C. § 230 would extinguish them. Section 230 provides that the Platforms “shall [not] be treated as the publisher or speaker” of content developed by other users. Id. § 230(c)(1). Section 230 reflects Congress’s judgment that the Platforms do not operate like traditional publishers and are not “speak[ing]” when they host user- submitted content. Congress’s judgment reinforces our conclusion that the Platforms’ censorship is not speech under the First Amendment. Congress enacted Section 230 in 1996 to ease uncertainty regarding online platforms’ exposure to defamation liability for the content they host. One leading case, Cubby, Inc. v. CompuServe, Inc., 776 F. Supp. 135 (S.D.N.Y. 1991), held that an online platform could not be liable absent knowledge of the defamatory statements, because it was a distributor that did not exercise meaningful editorial control. See id. at 139–40. But then a different case, Stratton Oakmont, Inc. v. Prodigy Services Co., 1995 WL 323710 (N.Y. Sup. Ct. May 24, 1995), accepted an argument very similar to the Platforms’ argument here. It noted that Prodigy’s online platform had “content guidelines” prohibiting certain obscene and offensive content. Id. at *2. And Prodigy used an “automatic software screening program” as well as manual review “to delete notes from its computer bulletin boards” that violated the guidelines. Id. at *4. The court held that this conduct “constitute[d] editorial control” over the platform, so the platform was akin to a newspaper and Prodigy could be held liable for defamation on that basis. Ibid. Congress disagreed with Stratton Oakmont and abrogated it by enacting § 230. See H.R. Rep. No. 104-458, at 194 (1996) (“One of the specific purposes of [§ 230] is to overrule Stratton-Oakmont v. Prodigy and any other similar decisions which have treated such providers and users as 39 Case: 21-51178 Document: 00516474571 Page: 40 Date Filed: 09/16/2022 No. 21-51178 publishers or speakers of content that is not their own because they have restricted access to objectionable material.”). Congress instructed that “No provider or user of an interactive computer service [i.e., online platform] shall be treated as the publisher or speaker of any information provided by another information content provider.” 47 U.S.C. § 230(c)(1). Online platforms are thus immune from defamation liability for the content they host, unless they play a part in the “creation or development” of that content. See id. § 230(f )(3). And this is true even if the online platforms act “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.” Id. § 230(c)(2). Section 230 undercuts both of the Platforms’ arguments for holding that their censorship of users is protected speech. Recall that they rely on two key arguments: first, they suggest the user-submitted content they host is their speech; and second, they argue they are publishers akin to a newspaper. Section 230, however, instructs courts not to treat the Platforms as “the publisher or speaker” of the user-submitted content they host. Id. § 230(c)(1). And those are the exact two categories the Platforms invoke to support their First Amendment argument. So if § 230(c)(1) is constitutional, how can a court recognize the Platforms as First-Amendment-protected speakers or publishers of the content they host? The Platforms respond that they in fact are speakers and publishers, and Congress simply instructed courts to pretend they aren’t for purposes of publishing-related liability. Moreover, the legislature can’t define what constitutes “speech” under the First Amendment—otherwise, for example, it could abrogate Miami Herald by simply defining newspapers as “not publishers.” Because the legislature may not define what constitutes First- Amendment-protected speech, the Platforms argue § 230 has no bearing on the constitutional questions in this case. 40 Case: 21-51178 Document: 00516474571 Page: 41 Date Filed: 09/16/2022 No. 21-51178 It’s obviously true that a legislature can’t define what speech is or is not protected by the First Amendment. Cf. Marbury, 5 U.S. at 177. It’s also irrelevant because that’s not what § 230 purports to do. The First Amendment generally precludes liability based on the content of someone’s speech or expression. E.g., Cohen v. California, 403 U.S. 15 (1971). Defamation liability for publishers is one of the several exceptions to this rule. See generally New York Times Co. v. Sullivan, 376 U.S. 254 (1964) (relationship between defamation liability and the First Amendment). But § 230 creates an exemption from that exception for the “interactive computer services” that fall within its scope, including the Platforms. And it does so by stating that they should not be treated as publishers. Thus, § 230 is nothing more (or less) than a statutory patch to a gap in the First Amendment’s free speech guarantee. Given that context, it’s strange to pretend that § 230’s declaration that Platforms “shall [not] be treated as . . . publisher[s]” has no relevance in the First Amendment context. Moreover, Congress’s factual determinations do carry weight in constitutional adjudication. As the Supreme Court has explained, Congress’s findings on “essentially factual issues . . . are of course entitled to a great deal of deference.” Walters v. Nat’l Ass’n of Radiation Survivors, 473 U.S. 305, 330 n.12 (1985); see also, e.g., Turner Broad. Sys., Inc. v. FCC, 520 U.S. 180, 195–96 (1997) (“Turner II”). And § 230 reflects Congress’s factual determination that the Platforms are not “publishers.” Deference to Congress’s judgment is particularly appropriate here because the Platforms themselves have extensively affirmed, defended, and relied on that judgment. For example, they’ve asserted that § 230 “promotes the free exchange of information and ideas over the Internet and prevents the inevitable chill of speech that would occur if interactive computer services could be held liable merely for serving as conduits for other parties’ 41 Case: 21-51178 Document: 00516474571 Page: 42 Date Filed: 09/16/2022 No. 21-51178 speech.” 19 Consistent with Congress’s judgment, they’ve told courts repeatedly that they merely serve as “conduits” for other parties’ speech and use “neutral tools” to conduct any processing, filtering, or arranging that’s necessary to transmit content to users. 20 They’ve also repeatedly defended the wisdom of Congress’s judgment, arguing that § 230 “made it possible for every major internet service to be built and ensured important values like free expression and openness were part of how platforms operate.” 21 The Platforms’ position in this case is a marked shift from their past claims that they are simple conduits for user speech and that whatever might look like editorial control is in fact the blind operation of “neutral tools.” Two amici argue that the Platforms are therefore judicially estopped from asserting that their censorship is First-Amendment-protected editorial discretion. 22 See In re Superior Crewboats, Inc., 374 F.3d 330, 334 (5th Cir. 2004) (“Judicial estoppel is a common law doctrine that prevents a party from assuming inconsistent positions in litigation.”). That’s a fair point. But in any event, the Platforms’ frequent affirmation of Congress’s factual judgment underlying § 230 makes us even more skeptical of their radical 19 Brief for Appellees at 1, Klayman v. Zuckerberg, No. 13-7017 (D.C. Cir. Oct. 25, 2013). 20 E.g., Notice of Motion and Motion to Dismiss at 10 n.5, Fields v. Twitter, Inc., No. 3:16-cv-00213 (N.D. Cal. Apr. 6, 2016) (“conduit”); Motion to Dismiss at 10, Doe v. Twitter, Inc., No. 3:21-cv-00485 (N.D. Cal. Mar. 10, 2021) (“neutral tools”); Brief for Defendants-Appellants at 50, Colon v. Twitter, Inc., No. 21-11283 (11th Cir. Aug. 10, 2020) (“neutral tools”). 21 Does Section 230’s Sweeping Immunity Enable Big Tech Bad Behavior? Hearing Before the S. Comm. on Com., Sci., & Transp., 116th Cong. 2 (2020) [hereinafter Senate Hearings] (statement of Mark Zuckerberg, CEO, Facebook, Inc.); see also id. at 1 (statement of Jack Dorsey, CEO, Twitter, Inc.) (arguing that “Section 230 is the internet’s most important law for free speech and safety”). 22 See Brief for Amici Curiae Heartland Inst. & Am. Principles Project at 12. 42 Case: 21-51178 Document: 00516474571 Page: 43 Date Filed: 09/16/2022 No. 21-51178 switcheroo that, in this case, they are publishers. Cf. ibid. (doctrine of judicial estoppel “protect[s] the integrity of the judicial process by preventing parties from playing fast and loose with the courts to suit the exigencies of self interest” (quotation omitted)). The Platforms’ only response is that in passing § 230, Congress sought to give them an unqualified right to control the content they host— including through viewpoint-based censorship. They base this argument on § 230(c)(2), which clarifies that the Platforms are immune from defamation liability even if they remove certain categories of “objectionable” content. But the Platforms’ argument finds no support in § 230(c)(2)’s text or context. First, § 230(c)(2) only considers the removal of limited categories of content, like obscene, excessively violent, and similarly objectionable expression. 23 It says nothing about viewpoint-based or geography-based censorship. Second, read in context, § 230(c)(2) neither confers nor contemplates a freestanding right to censor. Instead, it clarifies that censoring limited categories of content does not remove the immunity conferred by § 230(c)(1). So rather than helping the Platforms’ case, § 230(c)(2) further undermines the Platforms’ claim that they are akin to newspapers for First Amendment purposes. That’s because it articulates 23 Section 230(c)(2) refers to “obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable” material. To the extent the Platforms try to extract an unqualified censorship right from the phrase “otherwise objectionable” in isolation, that’s foreclosed by the Supreme Court’s repeated instruction that “where general words follow specific words in a statutory enumeration, the general words are construed to embrace only objects similar in nature to those objects enumerated by the preceding specific words.” Wash. State Dep’t of Soc. & Health Servs. v. Guardianship Estate of Keffeler, 537 U.S. 371, 384 (2003) (quotation omitted); see also, e.g., Yates v. United States, 574 U.S. 528, 545 (2015) (plurality op.). 43 Case: 21-51178 Document: 00516474571 Page: 44 Date Filed: 09/16/2022 No. 21-51178 Congress’s judgment that the Platforms are not like publishers even when they engage in censorship. 24 In sum, § 230 reflects Congress’s judgment that the Platforms are not acting as speakers or publishers when they host user-submitted content. While a statute may not abrogate constitutional rights, Congress’s factual judgment about the role of online platforms counsels against finding that the Platforms “publish” (and hence speak) the content that other users post. And that’s particularly true here, because the Platforms have long relied on and vigorously defended that judgment—only to make a stark about-face for this litigation. Section 230 thus reinforces our conclusion that the Platforms’ censorship is not protected speech under the First Amendment. E. The common carrier doctrine is a body of common law dating back long before our Founding. It vests States with the power to impose nondiscrimination obligations on communication and transportation providers that hold themselves out to serve all members of the public without individualized bargaining. The Platforms are communications firms of tremendous public importance that hold themselves out to serve the public without individualized bargaining. And Section 7 of HB 20 imposes a basic 24 The Platforms also suggest, in a single sentence of their brief, that HB 20 is preempted by § 230(c)(2). The district court did not address this argument, so we are reluctant to pass on it. Cf. Cutter v. Wilkinson, 544 U.S. 709, 718 n.7 (2005) (“[W]e are a court of review, not of first view.”). Of course, an appellee may urge any ground properly raised below as an alternative basis for affirmance. See United States v. Am. Ry. Express Co., 265 U.S. 425, 435–36 (1924). But one sentence is insufficient to adequately brief a claim. See, e.g., United States v. Williams, 620 F.3d 483, 496 (5th Cir. 2010); Ragas v. Tenn. Gas Pipeline Co., 136 F.3d 455, 458 (5th Cir. 1998). We therefore hold that the Platforms have forfeited their preemption argument. 44 Case: 21-51178 Document: 00516474571 Page: 45 Date Filed: 09/16/2022 No. 21-51178 nondiscrimination requirement that falls comfortably within the historical ambit of permissible common carrier regulation. For this reason, to facially invalidate Texas’s nondiscrimination rule would be a remarkable derogation of core principles of federalism. American courts have recognized these principles since the Founding and only briefly abjured them to serve two unfortunate causes: imposing racial segregation and enforcing a discredited Lochner-era vision of property rights. Accepting the Platforms’ theory would represent the first time since those ignominious years that federal courts have prevented a State from requiring interstate transportation and communications firms to serve customers without discrimination. Given the firm rooting of common carrier regulation in our Nation’s constitutional tradition, any interpretation of the First Amendment that would make Section 7 facially unconstitutional would be highly incongruous. Common carrier doctrine thus reinforces our conclusion that Section 7 comports with the First Amendment. This section (1) begins with a brief primer on the history of common carrier doctrine. Then it (2) explains why common carrier doctrine permits Texas to impose Section 7’s nondiscrimination requirement on the Platforms. And this (3) supports our constitutional holding that the Platforms’ viewpoint-based censorship is not First-Amendment-protected speech. 1. The doctrine’s roots lie in the notion that persons engaged in “common callings” have a “duty to serve.” This principle has been part of Anglo-American law for more than half a millennium. For early English courts, this principle meant that private enterprises providing essential public services must serve the public, do so without discrimination, and charge a reasonable rate. The first “carriers” to which this principle was 45 Case: 21-51178 Document: 00516474571 Page: 46 Date Filed: 09/16/2022 No. 21-51178 applied were ferries. As Justice Newton of the Court of Common Pleas recounted, a ferry operator is “required to maintain the ferry and to operate it and repair it for the convenience of the common people.” Trespass on the Case in Regard to Certain Mills, YB 22 Hen. VI, fol. 14 (C.P. 1444). By the time of the American Founding, the duty to serve had crystallized into a key tenet of the common law. English courts applied this principle to numerous “common callings,” like stagecoaches, barges, gristmills, and innkeepers. See 3 Blackstone, supra, at *164 (discussing the duties of innkeepers, bargemasters, and farriers). For example, Blackstone explained that a public innkeeper offers “an implied engagement to entertain all persons who travel that way; and upon this universal assumpsit an action on the case will lie against him for damages, if he without good reason refuses to admit a traveler.” Ibid. Or as Sir Matthew Hale explained regarding wharves, when a private person builds the only wharf in a port, “the wharf and crane and other conveniences are affected with a public interest, and they cease to be juris privati only.” Matthew Hale, De Portibus Maris, in A Collection of Tracts Relative to the Law of England 77–78 (Francis Hargrave ed., 1787). The common law thus required the wharf owner to serve the public and not to impose discriminatory or unreasonable rates. See id. at 77 (wharf owner may not take “arbitrary and excessive duties for cranage”). The common carrier’s duty to serve without discrimination was transplanted to America along with the rest of the common law. See Charles M. Haar & Daniel Wm. Fessler, The Wrong Side of the Tracks: A Revolutionary Rediscovery of the Common Law Tradition of Fairness in the Struggle Against Inequality 109–40 (1986) [hereinafter Haar & Fessler]. It got its first real test with the rise of railroad empires in the second half of the nineteenth century. Rail companies became notorious for using rate 46 Case: 21-51178 Document: 00516474571 Page: 47 Date Filed: 09/16/2022 No. 21-51178 differentials and exclusive contracts to control industries dependent on cross-country shipping, often structuring contracts to give allies (like the Standard Oil Company) impenetrable monopolies. See id. at 112–15, 129. American courts, however, often found that these discriminatory practices violated the railroads’ common carrier obligations. See, e.g., Messenger v. Pa. R.R. Co., 37 N.J.L. 531, 534 (1874) (refusing to enforce rate differentials because “the carrier cannot discriminate between individuals for whom he will render the service”); New England Express Co. v. Me. Cent. R.R. Co., 57 Me. 188, 196 (1869) (rejecting exclusive contract because “[t]he very definition of a common carrier excludes the idea of the right to grant monopolies or to give special and unequal preferences”). And even when courts did not impose common carrier duties, they reaffirmed that state legislatures were vested with the power to do so by statute, as England did with the Railway and Canal Act of 1854. See Haar & Fessler, supra, at 115–23; see also, e.g., Fitchburg R.R. Co. v. Gage, 78 Mass. (12 Gray) 393, 398 (1859) (because railroads are common carriers, unequal rates are “very fully, and reasonably, subjected to legislative supervision and control”). The telegraph was the first communications industry subjected to common carrier laws in the United States. See Genevieve Lakier, The Non- First Amendment Law of Freedom of Speech, 134 Harv. L. Rev. 2299, 2320– 24 (2021). Invented in 1838, the telegraph revolutionized how people engaged with the media and communicated with each other over the next half century. But by the end of the nineteenth century, legislators grew “concern[ed] about the possibility that the private entities that controlled this amazing new technology would use that power to manipulate the flow of information to the public when doing so served their economic or political self-interest.” Id. at 2321. These fears proved well-founded. For example, Western Union, the largest telegraph company, sometimes refused to carry messages from journalists that competed with its ally, the Associated Press— 47 Case: 21-51178 Document: 00516474571 Page: 48 Date Filed: 09/16/2022 No. 21-51178 or charged them exorbitant rates. See id. at 2321–22. And the Associated Press in turn denied its valuable news digests to newspapers that criticized Western Union. See ibid. Western Union also discriminated against certain political speech, like strike-related telegraphs. See id. at 2322. And it was widely believed that Western Union and the Associated Press “influenc[ed] the reporting of political elections in an effort to promote the election of candidates their directors favored.” Ibid.; see, e.g., The Blaine Men Bluffing, N.Y. Times, Nov. 6, 1884, at 5 (accusing them of trying to influence the close presidential election of 1884 by misreporting and delaying the transmission of election returns). In response, States enacted common carrier laws to limit discrimination in the transmission of telegraph messages. The first such law, passed by New York, required telegraph companies to “receive d[i]spatches from and for . . . any individual, and on payment of their usual charges . . . to transmit the same with impartiality and good faith.” Act of April 12, 1848, ch. 265, § 11, 1848 N.Y. Laws 392, 395. New York further required such companies to “transmit all d[i]spatches in the order in which they [we]re received.” Id. § 12. Many States eventually passed similar laws, see Lakier, supra, at 2320, 2322, and Congress ultimately mandated that telegraph companies “operate their respective telegraph lines as to afford equal facilities to all, without discrimination in favor of or against any person, company, or corporation whatever.” Telegraph Lines Act, ch. 772, § 2, 25 Stat. 382, 383 (1888). Courts considering challenges to these laws—or requests to impose common carrier duties even in their absence—had to grapple with deciding whether and to what extent the common carrier doctrine applied to new innovations and technologies. For transportation and communications firms, courts focused on two things. First, did the carrier hold itself out to serve any member of the public without individualized bargaining? As Justice Story had 48 Case: 21-51178 Document: 00516474571 Page: 49 Date Filed: 09/16/2022 No. 21-51178 explained in the transportation context, “[t]o bring a person within the description of a common carrier, he must exercise it as a public employment; he must undertake to carry goods for persons generally; and he must hold himself out as ready to engage in the transportation of goods for hire as a business, not as a casual occupation.” Joseph Story, Commentaries on the Law of Bailments § 495 (9th ed. 1878). Courts applied this same holding-out test to novel communications enterprises. For example, in State ex rel. Webster v. Nebraska Telephone Co., 22 N.W. 237 (Neb. 1885), a Nebraska lawyer sought a writ of mandamus to compel a telephone company to put a telephone in his office. The Supreme Court of Nebraska granted the writ, explaining that the company “ha[d] undertaken with the public to send messages from its instruments, one of which it propose[d] to supply to each person or interest requiring it.” Id. at 239. Because the company had “so assumed and undertaken to the public,” it could not arbitrarily deny the lawyer a telephone. Ibid. Other courts agreed and clarified that telephone companies owed this common carrier obligation even though they also imposed “reasonable rules and regulations” upon their customers. Chesapeake & Potomac Tel. Co. v. Balt. & Ohio Tel. Co., 7 A. 809, 811 (Md. 1887); see also, e.g., Walls v. Strickland, 93 S.E. 857, 858 (N.C. 1917) (describing this rule as “well settled” by “numerous cases”). Second, drawing on Hale’s influential seventeenth-century formulation, courts considered whether the transportation or communications firm was “affected with a public interest.” This test might appear unhelpful, but it was “quickened into life by interpretation” over centuries of common law decisions. See Walton H. Hamilton, Affectation with Public Interest, 39 Yale L.J. 1089, 1090 (1930). Courts applying this test looked to whether a firm’s service played a central economic and social role in society. This discussion by the Supreme Court of Indiana is an instructive example: 49 Case: 21-51178 Document: 00516474571 Page: 50 Date Filed: 09/16/2022 No. 21-51178 The telephone is one of the remarkable productions of the present century, and, although its discovery is of recent date, it has been in use long enough to have attained well-defined relations to the general public. It has become as much a matter of public convenience and of public necessity as were the stage-coach and sailing vessel a hundred years ago, or as the steam-boat, the railroad, and the telegraph have become in later years. It has already become an important instrument of commerce. No other known device can supply the extraordinary facilities which it affords. It may therefore be regarded, when relatively considered, as an indispensable instrument of commerce. The relations which it has assumed towards the public make it a common carrier of news—a common carrier in the sense in which the telegraph is a common carrier—and impose upon it certain well-defined obligations of a public character. Hockett v. Indiana, 5 N.E. 178, 182 (Ind. 1886); see also, e.g., Webster, 22 N.W. at 239 (“That the telephone, by the necessities of commerce and public use, has become a public servant, a factor in the commerce of the nation, and of a great portion of the civilized world, cannot be questioned.”). In determining whether a communications firm was “affected with a public interest,” courts also considered the firm’s market share and the relevant market dynamics. In Hale’s original formulation, if a wharf owner operated the “only [wharf ] licensed by the queen” or if “there [wa]s no other wharf in that port,” then the wharf was “affected with a public interest,” and the owner acquired a duty to serve without discrimination. Hale, supra, at 77– 78. Similarly, a railroad, telegraph, or telephone company’s status as the only provider in a region heavily suggested it was affected with the public interest. See, e.g., Webster, 22 N.W. at 238 (“While there is no law giving [the phone company] a monopoly[,] . . . the mere fact of this territory being covered by the ‘plant’ of [the company], from the very nature and character of its business, gives it a monopoly of the business which it transacts.”). 50 Case: 21-51178 Document: 00516474571 Page: 51 Date Filed: 09/16/2022 No. 21-51178 When state legislatures or state courts imposed new common carrier requirements, affected firms often sought to evade them by bringing constitutional claims in federal court. The landmark case is Munn v. Illinois, 94 U.S. 113 (1876). Illinois passed a statute regulating railroads and grain elevators. Among other things, the statute regulated grain elevators’ rates and prohibited rate discrimination. See id. at 117. Munn & Scott, proprietors of a Chicago grain elevator, brought a litany of constitutional challenges to Illinois’s law, arguing that it violated the Commerce and Port Preference Clauses of Article I, as well as the Due Process and Equal Protection Clauses of the Fourteenth Amendment. See id. at 119–20. The thrust of the challenge was that Illinois’s law subverted private property rights without compensation and without sufficient justification. See, e.g., id. at 133. The Supreme Court rejected Munn & Scott’s claims and held that state legislatures may constitutionally regulate private firms if the service they provide is “affected with a public interest.” Id. at 130. The Court expounded at length “the doctrine which Lord Hale has so forcibly stated,” approving Hale’s formulation and tracing its adoption and development in American common law. See id. at 126–30. It then explained that the Illinois legislature could have reasonably determined that grain elevators were affected with a public interest. That’s because they were enormously important to the agriculture and shipping industries: They stood in the “gateway of commerce” and provided an indispensable link between western grain and eastern markets. Id. at 132. And while there were fourteen grain elevators in Chicago, controlled by nine firms, the market was small and interconnected enough to be ripe for abuse if state regulation was wholly prohibited. See id. at 131. After Munn, the Supreme Court repeatedly upheld common carrier regulations against constitutional challenges. The same year, for example, it easily rejected a railroad’s challenge to rate regulation and nondiscrimination 51 Case: 21-51178 Document: 00516474571 Page: 52 Date Filed: 09/16/2022 No. 21-51178 requirements imposed by the Iowa legislature. See Chi., Burlington & Quincy R.R. Co. v. Iowa, 94 U.S. 155, 161 (1876) (holding that railroads are “engaged in a public employment affecting the public interest, and, under [Munn v. Illinois, are] subject to legislative control as to their rates of fare and freight, unless protected by their charters”). It similarly rejected a constitutional challenge to a state legislature’s imposition of a duty on telegraph companies to deliver messages with “impartiality and good faith.” W. Union Tel. Co. v. James, 162 U.S. 650, 651 (1896). The Court deviated from this path only briefly and only during an ignominious period of history marked by racism and the now-discredited theory of Lochner v. New York, 198 U.S. 45 (1905). Shortly after Munn, for example, the Court considered a Louisiana law that required common carriers operating steamboats, railroads, and other vehicles to admit persons equally, without segregating on the basis of race. Hall v. De Cuir, 95 U.S. 485, 486 (1877). The Court sustained a constitutional challenge to the law on the ground that it regulated interstate commerce and violated the (negative or dormant) Commerce Clause. See id. at 490. Although the law only applied in Louisiana, the Court found it “impose[d] a direct burden upon inter-state commerce” because “[a] passenger in the cabin set apart for the use of whites without the State must, when the boat comes within, share the accommodations of that cabin with such colored persons as may come on board afterwards, if the law is enforced.” Id. at 488–89. Moreover, during the heyday of Lochner’s substantive due process misadventure, the Court repeatedly rejected States’ arguments that various industries were “affected with a public interest” and often invalidated state laws that included nondiscrimination rules. E.g., Chas. Wolff Packing Co. v. Ct. of Indus. Rels., 262 U.S. 522, 544 (1923) (invalidating state law regulating wages in the meat-packing industry); cf. Ray A. Brown, Due Process of Law, Police Power, and the Supreme Court, 40 Harv. L. Rev. 943, 944 (1927) 52 Case: 21-51178 Document: 00516474571 Page: 53 Date Filed: 09/16/2022 No. 21-51178 (noting that the Supreme Court declared more economic and social regulations unconstitutional between 1920 and 1927 than during the preceding 52 years). The Court has obviously rejected both Lochner and the odious racism that infected its decisions in the era of Hall and Plessy v. Ferguson, 163 U.S. 537 (1896). The Platforms have pointed to no case since that time—and we are not aware of any—sustaining a constitutional challenge to a state law imposing nondiscrimination obligations on a common carrier. 2. Texas permissibly determined that the Platforms are common carriers subject to nondiscrimination regulation. That’s because the Platforms are communications firms, hold themselves out to serve the public without individualized bargaining, and are affected with a public interest. To state the obvious, the Platforms are communications firms. The Platforms halfheartedly suggest that they are not “members of the ‘communications industry’” because their mode of transmitting expression differs from what other industry members do. But that’s wrong. The whole purpose of a social media platform—as aptly captured in HB 20’s definitional provisions—is to “enable[] users to communicate with other users.” Tex. Bus. & Com. Code § 120.001(1). The Platforms’ own representations confirm this—for example, Facebook’s Terms of Service indicates its purpose is to enable users to “communicate with friends, family, and others.” 25 In that sense, the Platforms are no different than Verizon or AT&T. 25 Facebook Terms, § 1; see also Twitter Terms, § 3 (purpose of Twitter is to host “Content” and “communications”). 53 Case: 21-51178 Document: 00516474571 Page: 54 Date Filed: 09/16/2022 No. 21-51178 The Platforms also hold themselves out to serve the public. 26 They permit any adult to make an account and transmit expression after agreeing to the same boilerplate terms of service. They’ve thus represented a “willingness to carry [anyone] on the same terms and conditions.” Semon v. Royal Indem. Co., 279 F.2d 737, 739 (5th Cir. 1960). The Platforms resist this conclusion, arguing that they have not held themselves out to serve the public equally. That’s so, they contend, because they are only willing to do business with users who agree to their terms of service. But requiring “compliance with their reasonable rules and regulations” has never permitted a communications firm to avoid common carrier obligations. Chesapeake, 7 A. at 811. The relevant inquiry isn’t whether a company has terms and conditions; it’s whether it offers the “same terms and conditions [to] any and all groups.” Semon, 279 F.2d at 739 (emphasis added). Put differently, the test is whether the company “make[s] individualized decisions, in particular cases, whether and on what terms to deal.” FCC v. Midwest Video Corp., 440 U.S. 689, 701 (1979) (quotation omitted). Here, it’s undisputed the Platforms apply the same terms and conditions to all existing and prospective users. The Platforms also contend they are not open to the public generally because they censor and otherwise discriminate against certain users and expression. To the extent the Platforms are arguing that they are not common carriers because they filter some obscene, vile, and spam-related expression, this argument lacks any historical or doctrinal support. For example, phone companies are privileged by law to filter obscene or harassing expression, and they often do so. 47 U.S.C. § 223; see, e.g., Carlin Commc’ns, Inc. v. Mountain States Tel. & Tel. Co., 827 F.2d 1291, 1292 (9th Cir. 1987). Yet they’re still 26 Indeed, one Platform has described its purpose as “to serve the public conversation.” Senate Hearings, supra, at 1 (statement of Jack Dorsey, CEO, Twitter, Inc.). 54 Case: 21-51178 Document: 00516474571 Page: 55 Date Filed: 09/16/2022 No. 21-51178 regulated as common carriers. Similarly, transportation providers may eject vulgar or disorderly passengers, yet States may nonetheless impose common carrier regulations prohibiting discrimination on more invidious grounds. E.g., Williams v. Trans World Airlines, 509 F.2d 942, 948 (2d Cir. 1975). The Platforms nonetheless contend that they cannot be regulated as common carriers because they engage in viewpoint-based censorship—the very conduct common carrier regulation would forbid. This contention is upside down. The Platforms appear to believe that any enterprise can avoid common carrier obligations by violating those same obligations. That is obviously wrong and would rob the common carrier doctrine of any content. The Platforms’ contention also involves a fair bit of historical amnesia. As discussed earlier, telegraph companies once engaged in extensive viewpoint-based discrimination, but that did not immunize them from common carrier regulation. Rather, for most legislators and courts, it made such regulation all the more urgent. See Lakier, supra, at 2322–23. And nearly every other industry historically subjected to common carrier regulation initially discriminated against their customers and sought the right to continue to do so. See, e.g., Messenger, 37 N.J.L. at 532–33 (railroad); Munn, 94 U.S. at 119–20 (grain elevators); Webster, 22 N.W. at 238 (telephone); Portland Nat. Gas & Oil Co. v. State ex rel. Kern, 34 N.E. 818, 818 (Ind. 1893) (gas); City of Danville v. Danville Water Co., 53 N.E. 118, 121 (Ill. 1899) (water). The Platforms offer no reason to adopt an ahistorical approach under which a firm’s existing desire to discriminate against its customers somehow gives it a permanent immunity from common carrier nondiscrimination obligations. Texas also reasonably determined that the Platforms are “affected with a public interest.” Numerous members of the public depend on social media platforms to communicate about civic life, art, culture, religion, 55 Case: 21-51178 Document: 00516474571 Page: 56 Date Filed: 09/16/2022 No. 21-51178 science, politics, school, family, and business. The Supreme Court in 2017 recognized that social media platforms “for many 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.” Packingham, 137 S. Ct. at 1737. The Court’s “modern public square” label reflects the fact that in-person social interactions, cultural experiences, and economic undertakings are increasingly being replaced by interactions and transactions hosted or facilitated by the Platforms. And if anything, the Platforms’ position as the modern public square has only become more entrenched in the four years between Packingham and the Texas legislature’s finding, as the public’s usage of and dependance on the Platforms has continued to increase. 27 The centrality of the Platforms to public discourse is perhaps most vividly illustrated by multiple federal court of appeals decisions holding that the replies to a public official’s Twitter feed constitute a government “public forum” for First Amendment purposes. See Knight First Amend. Inst. v. Trump, 928 F.3d 226, 237 (2d Cir. 2019), vacated, 141 S. Ct. 1220 (2021) (mem.); Garnier v. O’Connor-Ratcliff, --- F.4th ----, 2022 WL 2963453, at *15 (9th Cir. July 27, 2022). 28 These decisions reflect the modern intuition that the Platforms are the forum for political discussion and debate, and exclusion 27 See Brooke Auxier & Monica Anderson, Social Media Use in 2021, Pew Research Ctr. (Apr. 7, 2021), https://perma.cc/TR42-LDDT; see also Daily Time Spent on Social Networking by Internet Users Worldwide from 2012 to 2022, Statista, https://www.statista.com/statistics/433871/daily-social-media-usage-worldwide/ (last visited Aug. 6, 2022) (stating that in 2022, the average American spends 123 minutes per day on social media). 28 See also Knight, 141 S. Ct. at 1221 (Thomas, J., concurring) (noting the tension between holding a Platform account to be a government public forum and the notion that the Platforms have no nondiscrimination obligations and may censor a user “at any time for any reason or no reason”). 56 Case: 21-51178 Document: 00516474571 Page: 57 Date Filed: 09/16/2022 No. 21-51178 from the Platforms amounts to exclusion from the public discourse. And for many, the Platforms are also no less central to quotidian discussions about matters like school, family, and business, than they are to debates about politics, science, and religion. In addition to their social importance, the Platforms play a central role in American economic life. For those who traffic in information—journalists, academics, pundits, and the like—access to the Platforms can be indispensable to vocational success. That’s because in the modern economy, the Platforms provide the most effective way to disseminate news, commentary, and other information. The same is true for all sorts of cultural figures, entertainers, and educators, a growing number of whom rely for much or all of their income on monetizing expression posted to the Platforms. Finally, even people and companies who traffic in physical goods often lean heavily on the Platforms to build their brand and market their products to consumers. That’s why the Platforms, which earn almost all their revenue through advertising, are among the world’s most valuable corporations. Thus, just like the telephone a century ago, the Platforms have become a key “factor in the commerce of the nation, and of a great portion of the civilized world.” Webster, 22 N.W. at 239. Or at the very least, one cannot say the Texas legislature’s judgment to that effect was unreasonable. It’s also true that each Platform has an effective monopoly over its particular niche of online discourse. Many early telephone companies did not have legal monopolies, but as a practical matter, they monopolized their geographic area due to the nature of the telephone business. See id. at 238. Likewise with the Platforms: While no law gives them a monopoly, “network effects entrench these companies” because it’s difficult or impossible for a competitor to reproduce the network that makes an established Platform useful to its users. Knight, 141 S. Ct. at 1224 (Thomas, J., concurring). 57 Case: 21-51178 Document: 00516474571 Page: 58 Date Filed: 09/16/2022 No. 21-51178 Academics have explored this concept in depth, 29 but to those familiar with the Platforms, a few concrete examples can easily demonstrate the point. To effectively monetize, say, carpet cleaning instructional videos (a real niche), one needs access to YouTube. Alternatively, sports “influencers” need access to Instagram. And political pundits need access to Twitter. It’s thus no answer to tell the censored athlete, as the Platforms do, that she can just post from a different platform. As Justice Thomas has aptly pointed out, that’s like telling a man kicked off the train that he can still “hike the Oregon Trail.” Id. at 1225. The Platforms’ entrenched market power thus further supports the reasonableness of Texas’s determination that the Platforms are affected with a public interest. Cf. Munn, 94 U.S. at 131 (market dynamics supported state legislature’s affectation finding when nine firms controlled the fourteen major grain elevators serving Chicago). The Platforms and their amici make three counterarguments that merit additional responses. First, they suggest that common carrier regulations are impermissible—or at least disfavored—unless the government has contributed to a carrier’s monopoly, such as by licensing a legal monopoly or acquiring property for the carrier through eminent domain. That’s obviously wrong. Recall that in Hale’s original formulation, common carrier treatment was appropriate if a proprietor operated the “only [wharf] licensed by the queen” or if there was simply “no other wharf in that port.” Hale, supra, at 77. American courts followed this formulation and did not require a government-conferred monopoly. E.g., Webster, 22 N.W. at 238. 29 See James Alleman, Edmond Baranes & Paul Rappoport, Multisided Markets and Platform Dominance, in Applied Economics in the Digital Era (James Alleman et al. eds. 2020); Kenneth A. Bamberger & Orly Lobel, Platform Market Power, 32 Berkeley Tech. L.J. 1051 (2017). 58 Case: 21-51178 Document: 00516474571 Page: 59 Date Filed: 09/16/2022 No. 21-51178 Even if the Platforms were right, however, the government has conferred a major benefit on the Platforms by enacting § 230. See supra Part III.D. As the Platforms have acknowledged, “Section 230 made it possible for every major internet service to be built.” 30 By their own admission, the Platforms are just as dependent on § 230’s liability shield as the old railroad companies were on the ability to traverse land acquired via eminent domain. Accordingly, the Texas legislature reasonably determined that the Platforms “have enjoyed governmental support in the United States” and that this supports common carrier regulation. 31 Second, the Platforms rely on a handful of modern precedents. Chief among them is U.S. Telecomm. Ass’n v. FCC, 855 F.3d 381 (D.C. Cir. 2017). There, Judge Srinivasan and then-Judge Kavanaugh sparred over the validity of the FCC’s net neutrality rule, which purported to use the FCC’s authority under the Telecommunications Act of 1996 to impose common carrier obligations on internet service providers. See id. at 418 (Kavanaugh, J., dissenting from the denial of rehearing en banc). Because the primary question was one of a federal agency’s regulatory authority, see id. at 418–26, the case has little relevance to a State’s invocation of the deeply rooted 30 Senate Hearings, supra, at 2 (statement of Mark Zuckerberg, CEO, Facebook, Inc.). 31 Amicus TechFreedom argues that if common carrier regulations are based on this sort of quid pro quo relationship, and § 230 is the quid, then a state government shouldn’t be able to exact the quo. Even apart from the fact that the common carrier doctrine does not require a quid pro quo arrangement, the argument that the quid and the quo must come from the same government fails on historical terms. For example, nineteenth-century railroads were chartered (the quid) by state governments, yet comprehensive common carrier regulations (the quo) were imposed by the federal government through the Interstate Commerce Act of 1887 and the Hepburn Amendments of 1906. See Haar & Fessler, supra, at 137–40. 59 Case: 21-51178 Document: 00516474571 Page: 60 Date Filed: 09/16/2022 No. 21-51178 common carrier doctrine. 32 And while it’s true that then-Judge Kavanaugh also argued that the net neutrality rule violated the First Amendment, that was because “the FCC ha[d] not even tried to make a market power showing.” See id. at 418; see also id. at 435 (rule would be constitutional upon showing of market power). Here, the Texas legislature found that “social media platforms with the largest number of users are common carriers by virtue of their market dominance,” and this finding is reasonable. See supra at 57–58. At any rate, Turner I is the closest Supreme Court case from the modern era, and it provides no help to the Platforms. There the Court, by a 5-4 vote, refused to hold unlawful federal regulations requiring cable operators to set aside certain channels for commercial broadcast stations. See 512 U.S. at 661–68. Most significant for our purposes, even the four dissenting Justices believed Congress could have permissibly imposed more modest common carrier regulations, rather than singling out broadcasters for preferential treatment as it had done. Id. at 684 (O’Connor, J., concurring in part and dissenting in part); see also ibid. (“[I]t stands to reason that if Congress may demand that telephone companies operate as common carriers, it can ask the same of cable companies; such an approach would not suffer from the defect of preferring one speaker to another.”). Third, the Platforms and their amici argue that they are not engaged in “carriage.” They claim that “at its core,” the common carrier doctrine is 32 The Platforms’ contention that federal law does not treat them as common carriers is similarly beside the point. See 47 U.S.C. § 223(e)(6) (clarifying that certain provisions of federal law should not “be construed to treat interactive computer services as common carriers”). No party is arguing that the Platforms’ common carrier obligations stem from federal law. The question is whether the State of Texas can impose common carrier obligations on the Platforms. And no party has argued that § 223(e)(6) preempts state common carrier regulation. 60 Case: 21-51178 Document: 00516474571 Page: 61 Date Filed: 09/16/2022 No. 21-51178 about “the transportation of property”—that is, carrying literal things. But rather than transport some physical thing, the Platforms “process” and “manipulate” data in their users’ newsfeeds. They claim this distinction between “processing” and “carriage” puts them outside the realm of the common carrier doctrine. There is no basis for the Platforms’ wooden metaphysical literalism. A distinction between literal “carriage” and the processing of data obviously would not fit the doctrine. Were that the case, the telephone and telegraph could never have been regulated as common carriers. So to make the purported distinction work, the Platforms and their amici ask us to conceive of telegraphy and telephony as conveying a “widget of private information” as a discrete “commodity product.” Brief for Amicus Curiae TechFreedom at 7–8. This wordgame defies both law and logic. First, it has no doctrinal support. The Platforms and their amici cite one case asserting that “[t]he transportation of property [is the] business of common carriers,” German All. Ins. Co. v. Lewis, 233 U.S. 389, 406 (1914), but they offer no support whatsoever for the proposition that property transportation is the only thing that defines common carriers. Second, because the Platforms, telephones, and telegraphs all process data at some level, the Platforms’ purported standard collapses into a distinction between “more complicated communications processing” (e.g., social media) and “less complicated communications processing” (e.g., telephony). There’s no logical or historical basis to adopt this framework. After all, it would have prevented the common carrier doctrine from ever being applied to a more sophisticated communications medium than the one it began with. Relatedly, the Platforms argue that even if they can be regulated as common carriers, Section 7 goes beyond the permissible scope of the 61 Case: 21-51178 Document: 00516474571 Page: 62 Date Filed: 09/16/2022 No. 21-51178 common carrier doctrine. That’s because it requires more than simple “carriage,” or hosting. It also prohibits censorship that “den[ies] equal access or visibility to, or otherwise discriminate[s] against expression.” Tex. Civ. Prac. & Rem. Code § 143A.001(1). The Platforms claim this will interfere with how they process the communications they host and transmit. This is simply another version of the argument that social media is too complicated a medium to bear common carrier nondiscrimination obligations. Common carriers have not normally been able to discharge their duties by hosting or transmitting communications per se. Rather, they’ve been required to do so without discriminating—with “impartiality and good faith,” as required by many state laws concerning telegraph transmission. W. Union, 162 U.S. at 651. States could even require telegraph companies to “transmit all d[i]spatches in the order in which they are received.” Act of April 12, 1848, ch. 265, § 12, 1848 N.Y. Laws 392, 395. Section 7 thus imposes ordinary common carrier nondiscrimination obligations, drafted to fit the particularities of the Platforms’ medium. At bottom, the Platforms ask us to hold that in the long technological march from ferries and bakeries, to barges and gristmills, to steamboats and stagecoaches, to railroads and grain elevators, to water and gas lines, to telegraph and telephone lines, to social media platforms—that social media marks the point where the underlying technology is finally so complicated that the government may no longer regulate it to prevent invidious discrimination. But we may not inter this venerable and centuries-old doctrine just because Twitter’s censorship tools are more sophisticated than Western Union’s. Cf. Brown v. Ent. Merch. Ass’n, 564 U.S. 786, 790 (2011) (“[B]asic 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.” (quotation omitted)). 62 Case: 21-51178 Document: 00516474571 Page: 63 Date Filed: 09/16/2022 No. 21-51178 3. The Platforms next argue that even if Section 7 is a valid common carrier regulation, it’s still unconstitutional. That’s wrong for two reasons. First, it’s instructive that federal courts have been generally skeptical of constitutional challenges to States’ common carrier nondiscrimination rules. Indeed, it appears that federal courts have only ever sustained such challenges for the now-discredited purposes of imposing racial segregation and enforcing a Lochner-era conception of private property rights. See supra Part III.E.1. Significantly, the Platforms rely on the dissenting opinion in Nebbia v. New York, 291 U.S. 502 (1934), a case in which a majority of the Court began to repudiate Lochner. They cite Justice McReynolds’s dissent for the proposition that “a state may not by legislative fiat convert a private business into a public utility.” Red Br. at 36 (quoting Nebbia, 291 U.S. at 555 (McReynolds, J., dissenting)). 33 Section 7 imposes a nondiscrimination requirement that comes nowhere close to making the Platforms public utilities. But more importantly, the Supreme Court has rejected Lochner and Justice McReynolds’s position. See, e.g., Ferguson v. Skrupa, 372 U.S. 726, 729 (1963). This court may not resurrect it, and the Platforms’ arguments provide little reassurance that we could hold Section 7 unconstitutional without doing so. 33 This and other frequent invocations of private property rights suggest the Platforms’ real complaint is with the Texas legislature meddling in their right to control their own business. But the Platforms have not brought a regulatory takings claim. Cf. Pa. Coal Co. v. Mahon, 260 U.S. 393, 415 (1922). Instead, they’ve asked for the more drastic remedy of invalidation of an economic regulation—a remedy the federal courts have not been in the business of providing since the Lochner era. Given the courts’ deference to state economic regulations for the last eight decades, “it would be freakish to single out” this historically grounded nondiscrimination requirement “for special treatment.” Cf. Gundy v. United States, 139 S. Ct. 2116, 2131 (2019) (Alito, J., concurring). 63 Case: 21-51178 Document: 00516474571 Page: 64 Date Filed: 09/16/2022 No. 21-51178 Second, the fact that the Platforms fall within the historical scope of the common carrier doctrine further undermines their attempt to characterize their censorship as “speech.” As discussed at length earlier, the Platforms’ primary constitutional argument is that they so closely oversee the speech on their Platforms that they exercise “editorial discretion” akin to a newspaper. But the same characteristics that make the Platforms common carriers—first, holding out their communications medium for the public to use on equal terms; and second, their well-understood social and economic role as facilitators of other people’s speech—render them not newspapers but instead indispensable conduits for transporting information. Put differently, it’s bizarre to posit that the Platforms provide much of the key communications infrastructure on which the social and economic life of this Nation depends, and yet conclude each and every communication transmitted through that infrastructure still somehow implicates the Platforms’ own speech for First Amendment purposes. F. Suppose Section 7 did implicate the Platforms’ First Amendment rights. The Platforms would still not be entitled to facial pre-enforcement relief. That’s because (1) it’s a content- and viewpoint-neutral law and is therefore subject to intermediate scrutiny at most. And (2) Texas’s interests undergirding Section 7 are sufficient to satisfy that standard. 1. Even if Section 7 burdens the Platforms’ First Amendment rights, it does so in a content-neutral way. Such “regulations that are unrelated to the content of speech are subject to an intermediate level of scrutiny” under the First Amendment. Turner I, 512 U.S. at 642 (quotation omitted). The “principal inquiry in determining content neutrality is whether the government has adopted a regulation of speech because of agreement or 64 Case: 21-51178 Document: 00516474571 Page: 65 Date Filed: 09/16/2022 No. 21-51178 disagreement with the message it conveys.” Ibid. Accordingly, “[a]s a general rule, laws that by their terms distinguish favored speech from disfavored speech on the basis of the ideas or views expressed are content based.” Id. at 643. But “laws that confer benefits or impose burdens on speech without reference to the ideas or views expressed are in most instances content neutral.” Ibid. Or as the Court put it more recently, “the phrase ‘content based’ requires a court to consider whether a regulation of speech ‘on its face’ draws distinctions based on the message a speaker conveys.” Reed v. Town of Gilbert, 576 U.S. 155, 163 (2015); accord City of Austin v. Reagan Nat’l Advert. of Austin, LLC, 142 S. Ct. 1464, 1471–74 (2022). Section 7 is content-neutral. Even assuming viewpoint-based censorship is speech, the burden Section 7 imposes on that speech does not depend on “the ideas or views [it] expresse[s].” Turner I, 512 U.S. at 643. In other words, Section 7’s burden in no way depends on what message a Platform conveys or intends to convey through its censorship. That’s because Section 7 applies equally regardless of the censored user’s viewpoint, and regardless of the motives (stated or unstated) animating the Platform’s viewpoint-based or geography-based censorship. The Platforms have several responses. First, they argue Section 7 is content-based because its definition of “social media platform” excludes news, sports, and entertainment websites. Specifically, Section 7 does not apply to “an online service, application, or website”: (i) that consists primarily of news, sports, entertainment, or other information or content that is not user generated but is preselected by the provider; and (ii) for which any chat, comments, or interactive functionality is incidental to, directly related to, or dependent on the provision of the content described by Subparagraph (i). 65 Case: 21-51178 Document: 00516474571 Page: 66 Date Filed: 09/16/2022 No. 21-51178 Tex. Bus. & Com. Code § 120.001(1)(C). This definition does not render HB 20 content-based because the excluded websites are fundamentally dissimilar mediums. And “the fact that a law singles out a certain medium . . . is insufficient by itself to raise First Amendment concerns.” Turner I, 512 U.S. at 660 (quotation omitted). HB 20 defines “social media platform” to sweep in websites that exist primarily to host and transmit user-generated speech. Section 120.001(1)(C)(i) does not create a content-based exemption from Section 7’s coverage. Rather, it excludes the distinct medium of websites whose primary purpose is not the sharing of user-generated speech but rather the dissemination of information “preselected by the provider.” Under Turner I, targeting a particular medium does not render Section 7 content-based. Second, the Platforms argue Section 7 is content-based because it permits certain narrow kinds of censorship. Section 7 permits Platforms to censor, for example, expression directly inciting criminal activity and specific threats of violence. See Tex. Civ. Prac. & Rem. Code § 143A.006(a). But the Platforms offer no evidence whatsoever that Texas permitted these narrow categories of censorship “because of agreement or disagreement with the message [such censorship] conveys.” Turner I, 512 U.S. at 642 (quotation omitted). Rather, Section 7 permits censorship of expression that’s unprotected by the First Amendment. See, e.g., Brandenburg v. Ohio, 395 U.S. 444, 447–48 (1969) (incitement unprotected). So it’s clear that the narrow permission to censor afforded by § 143A.006 is not “based on hostility—or favoritism—towards the underlying message expressed” by the Platforms’ censorship. R.A.V. v. City of St. Paul, 505 U.S. 377, 386 (1992). Section 143A.006 therefore does not render Section 7 content-based. Third, the Platforms argue that Section 7 triggers strict scrutiny because it targets only the largest social media platforms: those with more 66 Case: 21-51178 Document: 00516474571 Page: 67 Date Filed: 09/16/2022 No. 21-51178 than 50 million users. They contend this alone requires strict scrutiny, relying principally on Minneapolis Star and Tribune Co. v. Minnesota Commissioner of Revenue, 460 U.S. 575 (1983), and Arkansas Writers’ Project, Inc. v. Ragland, 481 U.S. 221 (1987). Minneapolis Star involved a challenge to Minnesota’s “use tax” on paper and ink products used by the press. 460 U.S. at 577. Because the tax exempted the first $100,000 of paper and ink used, only the largest eleven or so publishers incurred any tax liability in a given year. Id. at 578. The Court held that “Minnesota’s ink and paper tax violates the First Amendment not only because it singles out the press, but also because it targets a small group of newspapers.” Id. at 591. Similarly, in Arkansas Writers’ Project, the Court held unconstitutional another tax that “target[ed] a small group within the press,” this time by imposing a sales tax on magazines but exempting religious, trade, professional, and sports magazines. 481 U.S. at 229; see also Grosjean v. Am. Press Co., 297 U.S. 233, 251 (1936) (holding unconstitutional a tax singling out newspapers with weekly circulations above 20,000). These taxation cases are inapposite. As the Court later explained, Minneapolis Star and Arkansas Writers’ Project “demonstrate that differential taxation of First Amendment speakers is constitutionally suspect when it threatens to suppress the expression of particular ideas or viewpoints.” Leathers v. Medlock, 499 U.S. 439, 447 (1991). But “differential taxation of speakers, even members of the press, does not implicate the First Amendment unless the tax is directed at, or presents the danger of suppressing, particular ideas,” as “was the case in Grosjean, Minneapolis Star, and Arkansas Writers’ [Project].” Id. at 453. Section 7’s focus on a particular subset of firms is not directed at suppressing particular ideas or viewpoints, as Minnesota’s and Arkansas’s discriminatory taxes were. Rather, the law aims at protecting a diversity of ideas and viewpoints by focusing on the large firms that constitute “the modern public square.” 67 Case: 21-51178 Document: 00516474571 Page: 68 Date Filed: 09/16/2022 No. 21-51178 Packingham, 137 S. Ct. at 1737. Nor is there any evidence in the record before us that Section 7 could in fact suppress any constitutionally protected speech by anyone. See supra Part III.A. Minneapolis Star and Arkansas Writers’ Project thus provide no basis for subjecting Section 7 to strict scrutiny. Finally, the Platforms argue that Section 7 impermissibly targeted the largest social media platforms because of the Texas legislature’s particular disagreement with those Platforms’ partisan censorship efforts. This argument fails on both the facts and the law. On the facts, the Platforms present no real evidence of the Texas legislature’s alleged improper motives. Instead, they simply ask us to infer an improper motive from various unexplained amendments to the user threshold number and the fact that HB 20 lacks legislative findings regarding the user threshold. But it’s just as plausible to infer that the legislature simply picked a number that would sweep in the largest platforms most salient to public discussion and debate in Texas. And on the law, we may not hold unconstitutional “a statute that is . . . constitutional on its face, on the basis of what fewer than a handful of Congressmen said about it.” United States v. O’Brien, 391 U.S. 367, 384 (1968). We thus hold that even if Section 7 regulated the Platforms’ speech, intermediate scrutiny would apply. 2. Section 7 satisfies intermediate scrutiny. “A content-neutral regulation will be sustained under the First Amendment if it advances important governmental interests unrelated to the suppression of free speech and does not burden substantially more speech than necessary to further those interests.” Turner II, 520 U.S. at 189. We hold that Section 7’s regulation of viewpoint-based censorship meets each of these requirements. First, Section 7 advances an important governmental interest. HB 20’s legislative findings assert that Texas “has a fundamental interest in 68 Case: 21-51178 Document: 00516474571 Page: 69 Date Filed: 09/16/2022 No. 21-51178 protecting the free exchange of ideas and information in this state.” And Supreme Court precedent confirms that this is “a governmental purpose of the highest order.” Turner I, 512 U.S. at 663; see ibid. (“[A]ssuring that the public has access to a multiplicity of information sources is a governmental purpose of the highest order, for it promotes values central to the First Amendment.”); Turner II, 520 U.S. at 189 (“promoting the widespread dissemination of information from a multiplicity of sources” is an important government interest); see also Associated Press v. United States, 326 U.S. 1, 20 (1945) (“[T]he widest possible dissemination of information from diverse and antagonistic sources is essential to the welfare of the public.”). The Platforms argue Miami Herald shows that Section 7 does not further any sufficient government interest to satisfy intermediate scrutiny. That’s because the Miami Herald Court considered Florida’s argument that “the public has lost any ability to respond or to contribute in a meaningful way to the debate on issues” yet found that interest insufficient to justify the right-of-reply law. 418 U.S. at 250. But the Miami Herald Court never discussed or applied intermediate scrutiny, and it didn’t suggest Florida’s interest was unimportant. Rather, Florida’s law was unconstitutional because it imposed an obvious content-based penalty on the newspaper’s speech. Id. at 256. And at any rate, because it only protected the speech of political candidates the newspaper disfavored, it would have done little to advance the State’s broader interest in public debate. Miami Herald thus does not bear on the importance of Texas’s asserted interest in this case. The Platforms also rely on Hurley, but that case also did not apply intermediate scrutiny or weigh the strength of the governmental interest at stake. And Hurley distinguished Turner I by invoking the inherently expressive nature of a parade, as compared to “cable’s long history of serving as a conduit for broadcast signals.” 515 U.S. at 575–77 (quotation omitted). The Platforms do not exercise the same editorial discretion and control that 69 Case: 21-51178 Document: 00516474571 Page: 70 Date Filed: 09/16/2022 No. 21-51178 cable operators do—for example, they do not make ex ante decisions to select a limited repertoire of expression. See supra Part III.C.2.c. So if Hurley distinguished Turner I on that basis, then Hurley a fortiori doesn’t fit this case. In sum, the Platforms’ cases—none of which even applied intermediate scrutiny—do not undercut the Court’s holding that the widespread dissemination of information from a multiplicity of sources is “a governmental purpose of the highest order.” Turner I, 512 U.S. at 663. Second, Section 7 is “unrelated to the suppression of free speech” because it aims to protect individual speakers’ ability to speak. Turner II, 520 U.S. at 189. The Platforms resist this conclusion only by insisting that Section 7 curtails the Platforms’ own speech. That conflates the criteria for triggering intermediate scrutiny with the requirements for satisfying it. Intermediate scrutiny only kicks in when a law curtails speech, so the Platforms’ test would mean that no law triggering intermediate scrutiny could ever satisfy that standard. And that would make little sense. Section 7 is plainly unrelated to the suppression of free speech because at most it curtails the Platforms’ censorship—which they call speech—and only to the extent necessary to allow Texans to speak without suffering viewpoint discrimination. 34 34 In a similar vein, our esteemed colleague in dissent argues that Section 7 does not further the important government interest recognized in the Turner cases because it “strives to promote speech by first targeting the content of others’ speech.” Post, at 15. By contrast, according to the dissent, “[t]he Turner must-carry rules did not directly target cable-operators’ editorial discretion.” Ibid. In our view, Turner is not so easily distinguishable. In Turner, the interference with cable operators’ speech was not the point of the regulations, nor was it gratuitous—it was necessary to further the government’s interest in “the widespread dissemination of information from a multiplicity of sources.” Turner I, 512 U.S. at 662. So too here. Section 7 does not “directly target” the Platforms’ speech any more than the regulations in Turner targeted cable operators’ speech. As in Turner, the law only obstructs the Platforms’ expression to the extent necessary to protect the public’s “access to a multiplicity of information sources.” Id. at 663. The Platforms and the dissent offer no evidence that 70 Case: 21-51178 Document: 00516474571 Page: 71 Date Filed: 09/16/2022 No. 21-51178 Third, Section 7 “does not burden substantially more speech than necessary to further [Texas’s] interests.” Ibid. This is perhaps best illustrated by considering the Platforms’ main argument to the contrary: that “[i]f the State were truly interested in providing a viewpoint-neutral public forum, the State could have created its own government-run social-media platform.” The same network effects that make the Platforms so useful to their users mean that Texas (or even a private competitor) is unlikely to be able to reproduce that network and create a similarly valuable communications medium. See supra at 57–58 & n.29. It’s almost as absurd to tell Texas to just make its own Twitter as it would have been to tell broadcasters to just make their own cable systems. And aside from this bizarre claim, the Platforms offer no less restrictive alternative that would similarly advance Texas’s interest in “promoting the widespread dissemination of information from a multiplicity of sources.” Turner II, 520 U.S. at 189. 35 Section 7 gratuitously targets the Platforms’ speech or imposes a burden on the Platforms’ speech that doesn’t further the goal of protecting Texans’ expression. 35 Our esteemed colleague in dissent argues that “Section 7 burdens substantially more speech than necessary in order to further Texas’s legitimate interests” because it prohibits demonetization, de-boosting, and other forms of discrimination in addition to outright bans or content removal. Post, at 17 (quotation omitted). We disagree for several reasons. First, for some speakers who depend on advertising for their livelihoods, demonetization is tantamount to an outright ban because it dooms the financial viability of their enterprise and hence their speech. See, e.g., Brief for Amici Curiae The Babylon Bee, LLC, et al. at 4 (explaining amici’s reliance on monetization through social media platforms to disseminate speech). Second, demonetization and de-boosting, in addition to outright bans, also thwart “the widest possible dissemination of information from diverse and antagonistic sources,” an interest the Supreme Court has recognized as “essential to the welfare of the public.” Turner I, 512 U.S. at 663 (quotation omitted). They do so by penalizing and disincentivizing the same diversity the Supreme Court has recognized as “essential.” The dissent does not 71 Case: 21-51178 Document: 00516474571 Page: 72 Date Filed: 09/16/2022 No. 21-51178 The Platforms also suggest Section 7 is inadequately tailored because it’s under-inclusive. Specifically, they claim Texas could’ve applied Section 7 to smaller social media platforms too and could’ve excised the carveouts where the Platforms are still permitted to censor (like specific threats of violence). But Texas reasonably determined that the largest social media platforms’ market dominance and network effects make them uniquely in need of regulation to protect the widespread dissemination of information. And regulating smaller platforms would intrude more substantially on private property rights and perhaps create unique constitutional problems of its own. See PruneYard, 447 U.S. at 101 (Powell, J., concurring in part and in the judgment) (implying hosting rules would raise additional First Amendment concerns if applied to small entities). With regard to carveouts, the Platforms do not explain how requiring them to host, say, specific threats of violence or direct incitement of criminal activity would have meaningfully advanced Texas’s interest in protecting a widespread marketplace of ideas—especially when such speech enjoys no constitutional protection. See, e.g., Brandenburg, 395 U.S. at 447–48. Section 7 thus serves Texas’s important interest in protecting the widespread dissemination of information, is unrelated to the suppression of free expression, and does not burden substantially more speech than dispute the importance of Texas’s interest. Yet it’s hard to see how Texas can protect its interest in preserving a “multiplicity of information sources” if the Platforms may make them functionally invisible to users. See ibid. Finally, applying intermediate scrutiny, Texas must show only that its “statutory classification [is] substantially related to an important governmental objective.” Clark v. Jeter, 486 U.S. 456, 461 (1988). It need not be perfect, or even the “least restrictive alternative that can be used to achieve [Texas’s] goal.” Cf. Ashcroft v. ACLU, 542 U.S. 656, 666 (2004). Even if one chooses to nitpick at Texas’s enumeration of prohibited discriminatory acts, they are all at least “substantially related” to the furtherance of its concededly important interest. Clark, 486 U.S. at 461. 72 Case: 21-51178 Document: 00516474571 Page: 73 Date Filed: 09/16/2022 No. 21-51178 necessary to advance Texas’s interest. Section 7 therefore satisfies intermediate scrutiny and would be constitutional on that basis even if its censorship prohibitions implicated the Platforms’ First Amendment rights. IV. The Platforms next contend that they are entitled to pre-enforcement facial relief against Section 2 of HB 20. Again, we disagree. Section 2 requires the Platforms to make certain disclosures that consist of “purely factual and uncontroversial information” about the Platforms’ services. Zauderer v. Off. of Disciplinary Couns., 471 U.S. 626, 651 (1985). Under the relevant Supreme Court precedent, the Platforms are therefore not entitled to facial relief against Section 2. Section 2’s requirements fall into three categories. First, there are what we will call the “one-and-done” disclosures: requirements to publish an acceptable use policy and disclose certain information about the Platforms’ content management and business practices. See Tex. Bus. & Com. Code §§ 120.051–52. Second, there is the biannual transparency- report requirement, which obligates the Platforms to publish a report containing high-level statistics about their content-moderation activities every six months. See id. § 120.053. Third, there is the complaint-and-appeal- process requirement, which obligates the Platforms to explain their content removal decisions, permit affected users to appeal such removals, and generally respond to appeals within 14 business days. See id. §§ 120.101–04. Our review of these disclosure requirements is controlled by the Supreme Court’s decision in Zauderer. That case established that States may require commercial enterprises to disclose “purely factual and uncontroversial information” about their services. 471 U.S. at 651. At the same time, the Court recognized that “unjustified or unduly burdensome disclosure requirements might offend the First Amendment by chilling 73 Case: 21-51178 Document: 00516474571 Page: 74 Date Filed: 09/16/2022 No. 21-51178 protected commercial speech.” Ibid. And disclosure requirements must be reasonably related to a legitimate state interest, like preventing deception of consumers. See ibid. Texas argues—and the Platforms do not dispute—that Section 2 advances the State’s interest in “enabl[ing] users to make an informed choice” regarding whether to use the Platforms. Tex. Bus. & Com. Code § 120.051(b). Therefore, the only question is whether the State has carried its burden to show that the three categories of disclosures required by Section 2 are not unduly burdensome. See Nat’l Inst. of Fam. & Life Advocs., 138 S. Ct. 2361, 2377 (2021) (“NIFLA”). First, the one-and-done disclosures. Texas contends these impose a minimal burden, in part because the Platforms already largely comply with them. The Platforms respond that the one-and-done disclosures are unduly burdensome because Texas might find the disclosures inadequate and file suit. This argument is flawed on several levels. Most fundamentally, the Platforms do not explain how this concern can justify pre-enforcement relief against Section 2. The Platforms all but concede that publishing an acceptable use policy and high-level descriptions of their content and data management practices are not themselves unduly burdensome. Instead, they speculate that Texas will use these disclosure requirements to file unduly burdensome lawsuits seeking an unreasonably intrusive level of detail regarding, for example, the Platforms’ proprietary algorithms. But the Platforms have no authority suggesting the fear of litigation can render disclosure requirements unconstitutional—let alone that the fear of hypothetical litigation can do so in a pre-enforcement posture. Moreover, the Platforms’ argument ignores the fact that under Zauderer, we must evaluate whether disclosure requirements are “unduly burdensome” by reference to whether they threaten to “chill[] protected commercial speech.” 471 U.S. at 651. That is, Zauderer does not countenance a broad inquiry into whether disclosure requirements are “unduly 74 Case: 21-51178 Document: 00516474571 Page: 75 Date Filed: 09/16/2022 No. 21-51178 burdensome” in some abstract sense, but instead instructs us to consider whether they unduly burden (or “chill”) protected speech and thereby intrude on an entity’s First Amendment speech rights. 36 Here, the Platforms do not explain how the one-and-done disclosure requirements—or even the prospect of litigation to enforce those requirements—could or would burden the Platforms’ protected speech, even assuming that their censorship constitutes protected speech. Second, the biannual transparency report. Texas contends this requirement imposes little burden because the Platforms already track many of the statistics required by this report. The Platforms concede this point. They’ve shared and relied on much of that data in this lawsuit, and they do not dispute that reporting many of the required statistics would impose little burden. But they assert, with no explanation, that other required statistics— like how the Platforms were alerted to policy-violating content—would not be feasible to collect. And they again suggest that Texas will try to enforce this disclosure requirement in a particularly intrusive manner, such as by “demand[ing] access to platforms’ raw data.” These objections suffer from the same defects as the Platforms’ arguments against the one-and-done disclosures. At best, they’ve shown that 36 The Supreme Court’s recent decision in NIFLA further illustrates the Zauderer framework. In NIFLA, the Court considered a California law requiring unlicensed clinics serving pregnant women to provide certain notices. 138 S. Ct. at 2376–77. The Court held that the law failed First Amendment scrutiny under Zauderer—not because it was “unduly burdensome” in some administrative or operational sense, but because it would chill the clinics’ protected speech. For example, “a billboard for an unlicensed facility that says ‘Choose Life’ would have to surround that two-word statement with a 29-word statement from the government, in as many as 13 different languages.” Id. at 2378. This would “drown[] out the facility’s own message,” and, as a practical matter, preclude it from speaking that message in the first place. See ibid. NIFLA confirms that we evaluate whether a law is “unduly burdensome” by considering its burden on protected speech. 75 Case: 21-51178 Document: 00516474571 Page: 76 Date Filed: 09/16/2022 No. 21-51178 some of the transparency report’s disclosures, if interpreted in a particularly demanding way by Texas, might prove unduly burdensome due to unexplained limits on the Platforms’ technical capabilities. But none of these contingencies have materialized. And even if they did, a court would need to evaluate them on a case-by-case basis. Additionally, the Platforms have not explained how tracking the other purportedly more difficult statistics would unduly burden their protected speech, as opposed to imposing technical, economic, or operational burdens. So the Platforms are not entitled to facial pre-enforcement relief. See Zauderer, 471 U.S. at 651. Third, the complaint-and-appeal process. Texas again argues that the burden imposed by this requirement is reasonable because the Platforms already do what Section 2 requires for large swaths of content they transmit. And the Platforms again respond that complying with this requirement will prove unduly burdensome and is technically infeasible. But because the Platforms already largely comply with the complaint-and-appeal-process requirement, their only claim of infeasibility is that it’d be difficult to scale up the Platforms’ systems so as to provide a complaint-and-appeal process for all the content they host. And they provide just one example: YouTube comments. They emphasize that YouTube removed over a billion comments in a three-month period in 2020 and that providing an appeal process for comment removals would be substantially more onerous than providing the (existing) system for video removals. The Platforms also argue that the complaint-and-appeal requirement threatens to chill protected speech. That’s because, by requiring an explanation and appeal opportunity every time a Platform censors a user, the complaint-and-appeal requirement disincentivizes censorship in the first place. Even if the Platforms’ censorship was speech, and even assuming Section 2 would chill Google from censoring YouTube comments, that would not entitle the Platforms to facial pre-enforcement relief against Section 2. 76 Case: 21-51178 Document: 00516474571 Page: 77 Date Filed: 09/16/2022 No. 21-51178 The Platforms only argue that the complaint-and-appeal requirement will chill censorship for one subset of one Platform’s content. That falls far short of showing that “a substantial number of [Section 2’s] applications are unconstitutional, judged in relation to the statute’s plainly legitimate sweep.” Bonta, 141 S. Ct. at 2387 (quotation omitted). The Platforms do not allege that any other application of the complaint-and-appeal requirement will chill protected speech. And they couldn’t plausibly do so, because they already provide an appeals process substantially similar to what Section 2 requires for most other categories of content they host. One Platform CEO even testified that “[w]e believe that all companies should be required to provide a straightforward process to appeal decisions made by humans or algorithms.” 37 That’s hardly the stuff of a facial-overbreadth challenge. Perhaps recognizing that Section 2 easily passes muster under Zauderer, the Platforms next contend that case is inapposite. They give two reasons. First, they object that these disclosure requirements are triggered by the same definition of “social media platform” that Section 7 uses—a definition they claim is impermissibly content- and speaker-based. But we’ve already rejected the argument that HB 20’s definition of “social media platform” impermissibly targets particular content or particular speakers. See supra Part III.F.1. Second, the Platforms claim the Zauderer standard does not apply to disclosure laws that implicate the editorial process—that is, laws requiring publishers to disclose their editorial policies or explain how they exercise editorial discretion. They rely on dicta from Herbert v. Lando, 441 U.S. 153 (1979), where the Court suggested a State may not subject a publisher’s “editorial process to private or official examination merely to satisfy curiosity 37 Senate Hearings, supra, at 2 (statement of Jack Dorsey, CEO, Twitter, Inc.). 77 Case: 21-51178 Document: 00516474571 Page: 78 Date Filed: 09/16/2022 No. 21-51178 or to serve some general end such as the public interest.” Id. at 174. But Herbert held that a defamation plaintiff could obtain discovery into the editorial processes that allegedly defamed him. Id. at 175. And in the course of so holding, the Court rejected the editor’s request to create “a constitutional privilege foreclosing direct inquiry into the editorial process.” Id. at 176. The Platforms offer no authority suggesting we may create a constitutional privilege—akin to the one rejected in Herbert—for the disclosures mandated by Section 2. 38 But the more fundamental problem with the Platforms’ reliance on Herbert is that they do not have an “editorial process” that looks anything like a traditional publisher’s. See supra Part III.C.2.c. Herbert involved discovery into how an editor selected, composed, and edited a particular story. See 441 U.S. at 156–57. But the Platforms, of course, neither select, compose, nor edit (except in rare instances after dissemination) the speech they host. So even if there was a different rule for disclosure requirements implicating a newspaper-like editorial process, that rule would not apply here because the Platforms have no such process. Put differently, the question in Herbert was whether the Court should craft a rule protecting activities the Platforms do not even engage in—and even then, the Court answered “no.” We need not decide whether the Platforms might have meritorious as- applied challenges to particular applications of Section 2. We reiterate, however, that the First Amendment protects the Platforms from 38 The Platforms also rely on Washington Post v. McManus, 944 F.3d 506 (4th Cir. 2019), where the Fourth Circuit affirmed a preliminary injunction against a Maryland disclosure law targeting political campaign advertisements on online platforms. McManus is irrelevant for numerous reasons. Among them, Maryland’s law burdened a particular topic of speech (and was therefore content-based), see id. at 513; singled out political speech, see ibid.; compelled speech by actual newspapers, see id. at 517–18; and violated doctrines related to campaign-finance law, see id. at 515–17. 78 Case: 21-51178 Document: 00516474571 Page: 79 Date Filed: 09/16/2022 No. 21-51178 unconstitutional burdens on speech—not disclosure requirements that are burdensome in the abstract. Here, the Platforms have sought pre- enforcement facial relief primarily by objecting to the technical and operational burdens Section 2 will impose, and by highlighting a small number of applications that they contend will prove particularly burdensome. We hold that this does not entitle the Platforms to pre-enforcement facial relief against Section 2. V. Texas was not the first State to enact a law regulating censorship by large social media platforms. In May 2021, Florida enacted SB 7072, which sought to protect political candidates and journalistic organizations from censorship by large social media platforms. See Fla. Stat. §§ 106.072, 501.2041. 39 The Eleventh Circuit recently held that platforms challenging SB 7072 were entitled to a preliminary injunction against most of its provisions. See NetChoice, LLC v. Att’y Gen. of Fla., 34 F.4th 1196 (11th Cir. 2022). The Platforms urge us to follow the Eleventh Circuit’s NetChoice opinion. We will not. Most fundamentally, (A) SB 7072 and HB 20 are dissimilar laws in many legally relevant ways. Much of the Eleventh Circuit’s reasoning is thus consistent with or irrelevant to our resolution of the Platforms’ claims in this case. It’s also true, however, that (B) we disagree with the Eleventh Circuit’s reasoning at three critical junctures. A. Florida’s and Texas’s laws are very different. Three differences bear particular emphasis here. 39 The full text of SB 7072 can be accessed here: https://perma.cc/6WPF-4WC6. 79 Case: 21-51178 Document: 00516474571 Page: 80 Date Filed: 09/16/2022 No. 21-51178 First, SB 7072 only targets censorship of speech by political candidates and journalistic enterprises, as well as censorship of speech “about” political candidates. See Fla. Stat. §§ 106.072(2), 501.2041(2)(h), (2)(j). Under SB 7072, Platforms may not censor speech by or about a political candidate, full stop—no matter whether the speech is obscene or threatening. See id. § 501.2041(2)(h). And Platforms may only censor a journalistic enterprise’s expression if it is obscene. Id. § 501.2041(2)(j). But when it comes to non-journalists’ speech that doesn’t relate to a political campaign, the Platforms may continue to censor for any reason or no reason. 40 Thus, to generalize just a bit, SB 7072 prohibits all censorship of some speakers, while HB 20 prohibits some censorship of all speakers. Texas’s law permits non-viewpoint-based censorship and censorship of certain constitutionally unprotected expression regardless of who the speaker is. And HB 20 applies to all speakers equally, instead of singling out political candidates and journalists for favored treatment. These are of course highly relevant distinctions when deciding whether SB 7072 and HB 20 are impermissibly content- or speaker-based laws and whether they sufficiently tailored to satisfy heightened First Amendment scrutiny. See NetChoice, 34 F.4th at 1229 (relying on the absence of exceptions to hold that Florida’s absolute ban on censoring political candidates’ speech is insufficiently tailored to satisfy intermediate scrutiny). Second, several of SB 7072’s provisions arguably interfere with covered platforms’ own speech, instead of merely regulating how they 40 The only provision of SB 7072 arguably limiting censorship outside the realms of political candidates and journalists is § 501.2041(2)(b), which requires covered platforms to apply censorship standards “in a consistent manner.” SB 7072 does not define “consistent.” 80 Case: 21-51178 Document: 00516474571 Page: 81 Date Filed: 09/16/2022 No. 21-51178 transmit the speech of others. For example, Florida defines censorship to include “post[ing] an addendum to any content or material posted by a user.” Fla. Stat. § 501.2041(1)(b). Additionally, the Platforms may not modify their “rules, terms, and agreements” more than once every 30 days. Id. § 501.2041(2)(c). These provisions restrict the Platforms’ own speech— they can’t append a warning to a candidate’s or journalist’s post, and they can’t explain changes to their terms of service if they’ve already done so in the past month. HB 20, by contrast, does not interfere with the Platforms’ own speech in any way; they remain free to say whatever and whenever they want about their terms of service, about any user’s post, or about anything else. Third, SB 7072’s remedial scheme markedly differs from HB 20’s. Florida may collect fines of up to $250,000 per day for certain violations. Id. § 106.072(3). For others, platform users may win up to $100,000 in statutory damages per claim—along with actual and punitive damages. Id. § 501.2041(6). On the other hand, HB 20 does not permit the recovery of any damages; it only provides for prospective declaratory and injunctive relief. This distinction is significant when considering whether a pre-enforcement facial challenge is appropriate, especially given overbreadth doctrine’s concern with the chilling effect of challenged laws. Cf. NetChoice, 34 F.4th at 1230–31 (noting Florida’s law “provides for up to $100,000 in statutory damages per claim and pegs liability to vague terms like ‘thorough’ and ‘precise’” and holding this threatens to chill protected speech). Because of these and other distinctions between Florida’s and Texas’s laws, the Eleventh Circuit’s reasoning is either inapposite to or consistent with several of our holdings. In particular, our application of heightened First Amendment scrutiny and our evaluation of HB 20’s disclosure requirements are reconcilable with the Eleventh Circuit’s opinion. 81 Case: 21-51178 Document: 00516474571 Page: 82 Date Filed: 09/16/2022 No. 21-51178 B. We part ways with the Eleventh Circuit, however, on three key issues. Unlike the Eleventh Circuit, we (1) do not think the Supreme Court has recognized “editorial discretion” as an independent category of First- Amendment-protected expression. And even if it had, we (2) disagree with the Eleventh Circuit’s conclusion that the Platforms’ censorship is akin to the “editorial judgment” that’s been mentioned in Supreme Court doctrine. Finally, we (3) disagree with the Eleventh Circuit’s conclusion that the common carrier doctrine does not support the constitutionality of imposing nondiscrimination obligations on the Platforms. 41 41 The Eleventh Circuit also held, relying on its own precedent, that the Platforms’ censorship constitutes protected expressive conduct. See NetChoice, 34 F.4th at 1212–13. As noted earlier, the Platforms have not made an expressive-conduct argument in this case. See supra at 31 n.14. Even so, we are perplexed by the Eleventh Circuit’s holding that “social-media platforms engage in content moderation that is inherently expressive notwithstanding [Rumsfeld].” NetChoice, 34 F.4th at 1218. The Eleventh Circuit suggested that the Platforms’ “targeted removal of users’ speech” is different from law schools’ targeted denial of access to military recruiters because “a reasonable observer witnessing a platform remove a user or item of content would infer, at a minimum, a message of disapproval.” Id. at 1217; see also id. at 1217 n.15. But of course, a reasonable observer watching a law school eject a military recruiter would also infer a message of disapproval. The Supreme Court held that doesn’t matter because an observer who merely sees the military recruiting off campus could not know why the recruiter was off campus. See Rumsfeld, 547 U.S. at 66. Maybe it’s more convenient; maybe it’s because the law school ejected the military; maybe it’s some other reason. Likewise with the Platforms. An observer who merely sees a post on “The Democratic Hub,” NetChoice, 34 F.4th at 1214, could not know why the post appeared there. Maybe it’s more convenient; maybe it’s because Twitter banned the user; maybe it’s some other reason. Without more information, the observer has no basis for inferring a “particularized message” that Twitter disapproved the post. Johnson, 491 U.S. at 404. The Eleventh Circuit attempted to thread an eyeless needle. 82 Case: 21-51178 Document: 00516474571 Page: 83 Date Filed: 09/16/2022 No. 21-51178 1. The Eleventh Circuit reasoned that the Supreme Court’s decisions in Miami Herald, PG&E, Turner I, and Hurley establish an “editorial-judgment principle” under which a private entity has a First Amendment right to control “whether, to what extent, and in what manner to disseminate third- party-created content to the public.” NetChoice, 34 F.4th at 1212. But this purported rule is never mentioned by the cases the Eleventh Circuit relied on. And it’s flatly contradicted by other Supreme Court cases that the Eleventh Circuit addressed only as an afterthought. First, none of the cases the Eleventh Circuit relied on recognize an “editorial-judgment principle” or a distinct category of First Amendment protection for “editorial judgment.” Instead, each case explains how the challenged regulation either compelled or restricted speech. In Miami Herald, for example, Florida’s right-of-reply law both forced the Miami Herald to implicitly convey an editorial endorsement of speech it opposed and limited its opportunity to engage in other speech it would have preferred. See 418 U.S. at 256–58. Likewise in Turner I, the Court explained that “must-carry rules regulate cable speech” because they obstruct cable operators’ ability to express or convey the particular messages or programs they’ve chosen. 512 U.S. at 636–37; see also PG&E, 475 U.S. at 9–16; Hurley, 515 U.S. at 572–77. If the Eleventh Circuit’s rule was the Supreme Court’s rule, then all of those cases would have been easy analytical softballs. The Court would have merely needed to explain that the cases involved a private entity that wanted to control—that is, exercise “editorial judgment” over—speech it hosted. And that would have been the end of each case. But that’s not the analytical route the Supreme Court took. Instead, it focused on whether the challenged regulation either compelled or restricted the private entity’s own 83 Case: 21-51178 Document: 00516474571 Page: 84 Date Filed: 09/16/2022 No. 21-51178 speech—and explained at length why the regulations in Miami Herald, PG&E, Turner I, and Hurley did so. Second and more importantly, the Eleventh Circuit’s “editorial- judgment principle” conflicts with PruneYard and Rumsfeld. The Eleventh Circuit tries to square its rule with PruneYard by noting that there, the forum owner didn’t make an editorial-judgment argument. NetChoice, 34 F.4th at 1215. Perhaps, although that writes PruneYard out of the U.S. Reports by making the precedent irrelevant as long as a speech host chants the magical incantation “editorial judgment!” But then we get to Rumsfeld, where the forum owner did make the editorial-judgment argument: The law schools claimed a “First Amendment right to decide whether to disseminate or accommodate a military recruiter’s message” in their forum. 547 U.S. at 53. Yet the Supreme Court unanimously rejected the claimed right to choose who speaks in the law schools’ forum because “[t]he Solomon Amendment neither limits what law schools may say nor requires them to say anything.” Id. at 60. The Eleventh Circuit tried to square its “editorial-judgment principle” with Rumsfeld by asserting that “[s]ocial-media platforms, unlike law-school recruiting services, are in the business of disseminating curated collections of speech.” NetChoice, 34 F.4th at 1216. The Eleventh Circuit thus relied on the fact that social media platforms’ business is disseminating users’ speech, whereas law schools’ core business is not disseminating job recruiters’ speech. On the Eleventh Circuit’s reasoning, the business of disseminating speech is protected editorial judgment even if casual or sporadic dissemination is not. This distinction turns law, logic, and history on their heads. First, law: The Supreme Court’s cases have never stated or implied that this distinction is dispositive. If they had, phone companies and shipping services would be 84 Case: 21-51178 Document: 00516474571 Page: 85 Date Filed: 09/16/2022 No. 21-51178 free to discriminate, while PG&E (whose primary business is providing electricity, not disseminating speech) would have no First Amendment right to decline to share its billing envelope space with a third party. Next, logic: If a firm’s core business is disseminating others’ speech, then that should weaken, not strengthen, the firm’s argument that it has a First Amendment right to censor that speech. In PruneYard, for example, the shopping mall was open to the public—but for the purpose of shopping, not sharing expression. So it was perhaps tenuous for the State to use the public nature of the mall to justify a speech-hosting requirement. Cf. PruneYard, 447 U.S. at 95 (White, J., concurring in part) (noting that California’s hosting requirement involved communication “about subjects having no connection with the shopping centers’ business”). But here, the Platforms are open to the public for the specific purpose of disseminating the public’s speech. It’s rather odd to say that a business has more rights to discriminate when it’s in the speech business than when it’s in some altogether non-speech business (like shopping or legal education). Last, history: Communications firms have historically been the principal targets of laws prohibiting viewpoint-discriminatory transmission of speech. See supra Part III.E. By contrast, if an entity carried speech, people, or other goods only “as a casual occupation,” see Story, Commentaries on the Law of Bailments, supra, § 495, common carrier obligations could not be imposed. So there’s no basis in history, logic, or law for distinguishing Rumsfeld on the ground that law schools’ core business is not disseminating speech. The Eleventh Circuit also distinguished Rumsfeld on the ground that social media platforms, unlike law schools, disseminate “curated collections of speech.” NetChoice, 34 F.4th at 1216. This curation means that social media platforms are engaged in “editorial judgment” while law schools are 85 Case: 21-51178 Document: 00516474571 Page: 86 Date Filed: 09/16/2022 No. 21-51178 not. But that’s backwards. The law schools in Rumsfeld deliberately reviewed the content and viewpoint of bulletin board notices and emails before disseminating them to students on behalf of employers. But social media platforms, after algorithmic screening to filter obscenity and spam, arrange and transmit expression to users while remaining agnostic as to far more than 99% of that expression’s content and viewpoint. See Moody, 546 F. Supp. 3d at 1091–92. If either entity is “curating” expression in the ordinary sense— that is, engaging in substantive, discretionary review to decide what merits inclusion in a collection—it’s the law schools. A person’s social media feed is “curated” in the same sense that his mail is curated because the postal service has used automated screening to filter out hazardous materials and overweight packages, and then organized and affixed a logo to the mail before delivery. And it has never been true that content-agnostic processing, organizing, and arranging of expression generate some First Amendment license to censor. Were it otherwise, not only would Rumsfeld have come out the other way, but all sorts of nondiscrimination obligations currently imposed on communications firms and mail carriers would be unconstitutional. 2. The foregoing explains why the Eleventh Circuit’s articulation of its “editorial-judgment principle” conflicts with Supreme Court precedent. But even if editorial judgment was a freestanding category of First-Amendment- protected expression, the Eleventh Circuit’s explanation of why the Platforms’ censorship falls into that category is unpersuasive. The Eleventh Circuit did not discuss the glaring distinctions between the Platforms’ censorship and the editorial judgment described in Miami Herald and Turner I. For example, cable operators “exercise substantial editorial discretion in the selection and presentation of their 86 Case: 21-51178 Document: 00516474571 Page: 87 Date Filed: 09/16/2022 No. 21-51178 programming”—that is, they select (with great care) beforehand a limited repertoire of channels to transmit. Ark. Educ., 523 U.S. at 673. Newspapers similarly publish a narrow “choice of material” that’s been reviewed and edited beforehand, and they are subject to legal and reputational responsibility for that material. See Miami Herald, 418 U.S. at 258; see also id. at 261–62 (White, J., concurring). The Eleventh Circuit did not suggest the Platforms operate similarly. Instead, the Eleventh Circuit tried to equate the Platforms’ censorship with the editorial processes of newspapers and cable operators by reasoning that “Platforms employ editorial judgment to convey some messages but not others and thereby cultivate different types of communities.” NetChoice, 34 F.4th at 1213. For example, YouTube censors some content to create a “welcoming community”; Facebook censors to “foster authenticity, safety, privacy, and dignity”; and Twitter censors “to ensure all people can participate in the public conversation freely and safely.” Ibid. (quotation omitted). Because the Platforms censor speech to further these amorphous goals, the Eleventh Circuit held, the censorship is protected by the First Amendment. See ibid. Recall that under the Eleventh Circuit’s framework, the presence of editorial judgment generates a First Amendment right to censor. But now, censorship itself—as long as it’s explained by a generalized appeal to some attractive value—constitutes editorial judgment. This is circular: The Platforms have a right to censor because they exercise editorial judgment, and they exercise editorial judgment because they censor. The only arguably non- circular part of this framework is the apparent requirement that the censorship be justified by appealing to something like a “welcoming community” (as opposed to, say, an “unwelcoming one”). But the Eleventh Circuit gives this requirement no meaningful content: The Platforms may establish a First Amendment right to censor by invoking any generalized 87 Case: 21-51178 Document: 00516474571 Page: 88 Date Filed: 09/16/2022 No. 21-51178 interest, like “fostering authenticity,” without even explaining how viewpoint-based censorship furthers that interest. The practical upshot is that telephone companies, email providers, shipping services, or any other entity engaged in facilitating speech can acquire a First Amendment license to censor disfavored viewpoints by merely gesturing towards “safety” or “dignity.” That is not the law, as Miami Herald and Turner I illustrate and PruneYard and Rumsfeld confirm. 3. The Eleventh Circuit quickly dismissed the common carrier doctrine without addressing its history or propounding a test for how it should apply. See id. at 1219–22. This part of the Eleventh Circuit’s opinion is also unpersuasive. The Eleventh Circuit “confess[ed] some uncertainty” as to whether the State’s position was “(a) that platforms are already common carriers” or “(b) that the State can, by dint of ordinary legislation, make them common carriers.” Id. at 1220. It then rejected each position in turn. First, it reasoned that the Platforms are not already common carriers because pre-existing law did not already regulate them as such. See ibid. Moreover, the Platforms don’t currently follow common carrier obligations. 42 And pre-SB 7072 and HB 20 42 In this vein, the Eleventh Circuit found it significant that “social-media platforms have never acted like common carriers” and that users must “accept their terms of service and abide by their community standards.” NetChoice, 34 F.4th at 1220. Of course, violating common carrier obligations has never been sufficient to exempt a firm from common carrier obligations. The dominant telegraph companies, for example, offered discriminatory services before States regulated them as common carriers. See supra Part III.E. Similarly, most or all common carriers have terms of service—for example, one must accept FedEx’s terms to ship a package—and common carriers retain the right to remove unruly passengers or obscene transmissions. The Eleventh Circuit presents no authority suggesting this somehow forecloses common carrier regulation. 88 Case: 21-51178 Document: 00516474571 Page: 89 Date Filed: 09/16/2022 No. 21-51178 judicial decisions note the lack of government regulation of internet forums. 43 Second, it reasoned that the State can’t regulate them as common carriers because they are not already common carriers: That would give the “government authority to strip an entity of its First Amendment rights merely by labeling it a common carrier.” Id. at 1221. So in the Eleventh Circuit’s view, a firm can’t become a common carrier unless the law already recognizes it as such, and the law may only recognize it as such if it’s already a common carrier. Again, that’s circular. And it’s inconsistent with the common-law history and tradition discussed earlier, where common carrier nondiscrimination obligations were extended from ferries, to railroads, to telegraphy, to telephony, and so on. See supra Part III.E. The Eleventh Circuit didn’t purport to reconcile its approach with this history. The implication is that history doesn’t matter because SB 7072 is unconstitutional under the Eleventh Circuit’s “editorial-judgment principle.” But the Eleventh Circuit offers no persuasive justification for reading that principle into the Constitution, especially when it would contravene a deeply rooted common law nondiscrimination doctrine that’s centuries older than the Constitution itself. See supra Part III.E.1. 43 The Eleventh Circuit primarily focused on Turner I, analogizing social media platforms to cable broadcasters. But nothing in Turner I suggests that regulating social media platforms as common carriers would be unconstitutional. The opposite is true: Even the Turner I dissenters—the Justices who were more protective of cable operators’ speech rights—strongly suggested the First Amendment would not prevent regulating cable operators as common carriers. See 512 U.S. at 684 (O’Connor, J., concurring in part and dissenting in part) (“Congress might also conceivably obligate cable operators to act as common carriers for some of their channels . . . . [I]t stands to reason that if Congress may demand that telephone companies operate as common carriers, it can ask the same of cable companies.”). 89 Case: 21-51178 Document: 00516474571 Page: 90 Date Filed: 09/16/2022 No. 21-51178 * * * The First Amendment protects speech: It generally prevents the government from interfering with people’s speech or forcing them to speak. The Platforms argue that because they host and transmit speech, the First Amendment also gives them an unqualified license to invalidate laws that hinder them from censoring speech they don’t like. And they say that license entitles them to pre-enforcement facial relief against HB 20. We reject the Platforms’ attempt to extract a freewheeling censorship right from the Constitution’s free speech guarantee. The Platforms are not newspapers. Their censorship is not speech. They’re not entitled to pre- enforcement facial relief. And HB 20 is constitutional because it neither compels nor obstructs the Platforms’ own speech in any way. The district court erred in concluding otherwise and abused its discretion by issuing a preliminary injunction. The preliminary injunction is VACATED, and this case is REMANDED for further proceedings consistent with this opinion. 90 Case: 21-51178 Document: 00516474571 Page: 91 Date Filed: 09/16/2022 No. 21-51178, NetChoice v. Paxton Edith H. Jones, Circuit Judge, concurring: I concur in Judge Oldham’s conclusion and reasoning that the business of the regulated large social media platforms is hosting the speech of others. Functioning as conduits for both makers and recipients of speech, the platforms’ businesses are closer analytically to the holdings of the Supreme Court in PruneYard and FAIR than to Miami Herald, Pacific Gas & Electric, and Hurley. It follows from the first two cases that in arbitrarily excluding from their platforms the makers of speech and preventing disfavored speech from reaching potential audiences (“censoring,” in the comprehensive statutory term), they are not themselves “speaking” for First Amendment purposes. In particular, it is ludicrous to assert, as NetChoice does, that in forbidding the covered platforms from exercising viewpoint-based “censorship,” the platforms’ “own speech” is curtailed. But for their advertising such “censorship”—or for the censored parties’ voicing their suspicions about such actions—no one would know about the goals of their algorithmic magic. It is hard to construe as “speech” what the speaker never says, or when it acts so vaguely as to be incomprehensible. Further, the platforms bestride a nearly unlimited digital world in which they have more than enough opportunity to express their views in many ways other than “censorship.” The Texas statute regulates none of their verbal “speech.” What the statute does, as Judge Oldham carefully explains, is ensure that a multiplicity of voices will contend for audience attention on these platforms. That is a pro-speech, not anti-free speech result. Another way to look at this case, however, is through the Turner I decision, in which the Supreme Court held that cable TV companies are to some extent engaged in First Amendment-covered “speech” when, as they 91 Case: 21-51178 Document: 00516474571 Page: 92 Date Filed: 09/16/2022 No. 21-51178 “operate” their systems, they determine which cable channels to host. 1 Using intermediate scrutiny, the Court did not reject federal must-carry regulations requiring hosting of certain preferred channels. Instead, the Court distinguished both Pacific Gas & Electric and Miami Herald for three reasons. First, the must- carry regulations were content neutral. Second, they did not force cable operators to modify their own speech, nor were viewers likely to associate the mandatory hosted speech with that of the operators. And third, a cable operator’s selection of channels controlled the flow of information into subscribers’ households, and could “thus silence the voice of competing speakers with the mere flick of a switch.” Turner Broadcasting Sys., Inc. v. FCC, 512 U.S. at 656, 114 S. Ct. at 2466. I find all of these points compellingly applicable to analyzing the regulations imposed on large social media platforms by the Texas statute before us. 2 Finally, even if there is a legitimate basis to argue that the Texas statute may chill the platforms’ “speech,” it is not sufficient to sustain a facial attack, as Judge Oldham explains. Case by case adjudication is a small burden on the Goliaths of internet communications if they contend with Davids who use their platforms. 1 I do not believe it necessary to determine whether the Texas statute survives this facial attack on the theory of common carrier regulation and therefore do not subscribe to that portion of Judge Oldham’s opinion. Turner I, in my view, is applicable irrespective of overarching common carrier theory. 2 And as Judge Oldham notes, the dissenters in Turner I did not disavow the possibility of some regulation in the monopolistic context in which most cable companies then operated. 92 Case: 21-51178 Document: 00516474571 Page: 93 Date Filed: 09/16/2022 No. 21-51178 Leslie H. Southwick, Circuit Judge, concurring in part and dissenting in part: The central question in this case is whether social media platforms engage in First Amendment-protected expression when they moderate their users’ content. The erudite opinion of my colleagues in the majority says no. Although there are parts of the opinion I join, I write separately because, fundamentally, I conclude the answer to the question is yes. First, some points of agreement. As to the discussion of the First Amendment, the majority is certainly correct that a successful facial challenge to a state law is difficult. Consequently, I agree that a facial challenge to the Disclosure and Operations provisions in Section 2 of HB 20 is unlikely to succeed on the merits. These portions of the law ought not to be enjoined at the preliminary injunction stage. I also agree with my colleagues that the social media Platforms represented by NetChoice are “firms of tremendous public importance.” The part they have chosen to play in modern public discourse is at times detrimental to the healthy exchange of competing ideas. The argument here is that the Platforms blatantly censor the views of those with whom they disagree, leaving no equivalent platform available to the speakers they scorn. The Platforms certainly have taken aggressive, inconsistent positions before legislative, regulatory, and now judicial bodies about the relevance of the First Amendment to their actions. They pursue maximum freedom to shape discourse while accepting no liability for the content they host. The legal issues before us, though, must be separated from any disquiet irrelevant to the application of the First Amendment. My disagreement with my colleagues lies in the application of First Amendment principles to the anti-discrimination provisions of Section 7. The majority frames the case as one dealing with conduct and unfair censorship. The 93 Case: 21-51178 Document: 00516474571 Page: 94 Date Filed: 09/16/2022 No. 21-51178 majority’s rejection of First Amendment protections for conduct follows unremarkably. I conclude, though, that the majority is forcing the picture of what the Platforms do into a frame that is too small. The frame must be large enough to fit the wide-ranging, free-wheeling, unlimited variety of expression — ranging from the perfectly fair and reasonable to the impossibly biased and outrageous — that is the picture of the First Amendment as envisioned by those who designed the initial amendments to the Constitution. I do not celebrate the excesses, but the Constitution wisely allows for them. The majority no doubt could create an image for the First Amendment better than what I just verbalized, but the description would have to be similar. We simply disagree about whether speech is involved in this case. Yes, almost none of what others place on the Platforms is subject to any action by the companies that own them. The First Amendment, though, is what protects the curating, moderating, or whatever else we call the Platforms’ interaction with what others are trying to say. We are in a new arena, a very extensive one, for speakers and for those who would moderate their speech. None of the precedents fit seamlessly. The majority appears assured of their approach; I am hesitant. The closest match I see is caselaw establishing the right of newspapers to control what they do and do not print, and that is the law that guides me until the Supreme Court gives us more. What follows is my effort to work with the same material the majority analyzed. My desire is to explain why the Platforms’ moderating third-party- content is speech, where that speech fits into the broader body of First Amendment jurisprudence, and how I analyze the effect of Section 7 of HB 20 on that speech. I. Content moderation and the First Amendment The critical question is whether the anti-discrimination provisions in Section 7 of HB 20 regulate non-expressive conduct or whether they regulate 94 Case: 21-51178 Document: 00516474571 Page: 95 Date Filed: 09/16/2022 No. 21-51178 First Amendment-protected activity. The majority concludes that “Section 7 does not regulate the Platforms’ speech at all; it protects other people’s speech and regulates the Platforms’ conduct.” Maj. Op. at 7. The majority’s perceived censorship is my perceived editing. The Platforms can act with obvious bias. The lack of First Amendment protection for their biases is not so obvious. The majority has discussed the careful work of another circuit on the same essential questions. In assessing a similar law, the Eleventh Circuit held “a private entity’s decisions about whether, to what extent, and in what manner to disseminate third-party-created content to the public are editorial judgments protected by the First Amendment” and that “social-media platforms’ content-moderation decisions constitute the same sort of editorial judgments and thus trigger First Amendment scrutiny.” NetChoice, LLC v. Att’y Gen., Fla., 34 F.4th 1196, 1212 (11th Cir. 2022). I agree. The question we must answer is similar. In explaining my answer, I begin with an overview of what the Platforms currently do with content and a reminder of the obligations imposed by Section 7. The Platforms admit they take an active role in determining which pieces of content reach individual users: “Platforms compile, curate, and disseminate a combination of user-submitted expression, platform-authored expression, and advertisements.” To varying degrees, the Platforms all “control[] who can access their platforms, what kinds of content [are] available, and how that content is presented to users.” Section 7 limits the ability of Platforms to engage in these activities by imposing anti-discrimination policies. Platforms “may not censor a user, a user’s expression, or a user’s ability to receive the expression of another person based on: (1) the viewpoint of the user or another person; (2) the viewpoint represented . . . ; or (3) a user’s geographic location.” Tex. Civ. 95 Case: 21-51178 Document: 00516474571 Page: 96 Date Filed: 09/16/2022 No. 21-51178 Prac. & Rem. Code § 143A.002(a). “Censor” is a defined term that reaches many of the Platforms’ core activities. See id. § 143A.001. The Platforms may engage in the activities in varying frequency, but when the Platforms engage in any content moderation based on the views represented in the content, they “deny equal access or visibility to, or otherwise discriminate against expression” and violate the statute. Id. These activities native to the digital age have no clear ancestral home within our First Amendment precedent. Their closest relative may be what the Supreme Court held newspapers were permitted to do in Miami Herald Publishing Co. v. Tornillo, 418 U.S. 241 (1974). I see the Platforms’ curating or moderating as the current equivalent of a newspaper’s exercise of editorial discretion. This view requires me to consider many of the same authorities reviewed by the majority and explain where my conclusions diverge from those of my colleagues. 1 I start with Miami Herald, which considered a Florida statute that “grant[ed] a political candidate a right to equal space to reply to criticism . . . by a newspaper.” 418 U.S. at 243. The Miami newspaper sought declaratory 1 The majority analyzes several authorities when distinguishing between regulations on “hosting” speech and either requiring the “host” to speak or interfering with the host’s own message. I add one more. See Turner Broad. Sys., Inc. v. FCC, 512 U.S. 622 (1994) (“Turner I”); see Maj Op. Part III.C.1. Although I think Miami Herald is the case closest to the matter at hand, I discuss Turner I here because it interpreted Miami Herald and served as a basis for the decision in the Hurley case. See Turner I, 512 U.S. at 636–41, 653–57; Hurley v. Irish-American Gay, Lesbian, and Bisexual Group of Boston, 515 U.S. 557, 570 (1995). Additionally, as discussed below, Turner I emphasizes that, in the modern communications context, an entity may “host” the speech of others while simultaneously engaging in First Amendment activity of its own. See 512 U.S. at 636–37. Further, I will not engage with the majority’s analysis of the history of prior restraint. It is certainly a detailed review, with debatable points along the way. I limit my analysis to the extent needed to explain why I believe the Platforms are engaged in First Amendment-protected activity. 96 Case: 21-51178 Document: 00516474571 Page: 97 Date Filed: 09/16/2022 No. 21-51178 relief that the law was unconstitutional. 2 The majority recounts the basic facts of the case, then quotes the following passage: A newspaper is more than a passive receptacle or conduit for news, comment, and advertising. 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. Id. at 258. I wish to add, though, what the Court stated in the next sentence: “It has yet to be demonstrated how governmental regulation of this crucial process can be exercised consistent with First Amendment guarantees of a free press as they have evolved to this time.” Id. The majority sees the Court as having held that “[b]ecause a newspaper prints a curated set of material selected by its editors, everything it publishes is, in a sense, the newspaper’s own speech,” and that newspapers “cannot be compelled to ‘publish that which reason tells them should not be published.’” See Maj. Op. at 22 (quoting Miami Herald, at 256). The majority does not, though, understand the Court to have recognized the selection process itself as First Amendment expression. See Maj. Op. at 83. I do. There were at least two levels of publisher speech involved. Certainly, a traditional publisher cannot be forced to adopt speech with which they disagree. That was the first premise that underlay the Miami Herald holding. 418 U.S. at 256. The Court went further, though. It recognized that “[t]he 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 2 418 U.S. at 245; contra Maj. Op. at 13 (including Miami Herald in the contention that all of NetChoice’s cases “involved challenges to concrete applications of an allegedly unconstitutional law, raised by a defendant in state court proceedings”). 97 Case: 21-51178 Document: 00516474571 Page: 98 Date Filed: 09/16/2022 No. 21-51178 of editorial control and judgment” and that the Court did not see “how governmental regulation of this crucial process” was consistent with the First Amendment. Id. at 258. I read this as establishing the selection process itself as First Amendment-protected activity. Six years later, the Court considered the right of high school students to engage in their own First Amendment activity at a local shopping mall. PruneYard Shopping Ctr. v. Robins, 447 U.S. 74, 77 (1980). After mall security told the students to stop distributing political literature, the students sued the shopping mall owner in state court for infringing on their speech rights under the California state constitution. Id. The students succeeded in state courts. Id. at 78. At the Supreme Court, the owner of the shopping mall argued that being forced to host the students’ speech by the State of California violated both the owner’s property rights under the Fifth and Fourteenth Amendments and free speech rights under the First and Fourteenth Amendments. Id. at 76–77. In considering the shopping center owner’s assertion that Miami Herald controlled the case, the Court stated that the precedent “rests on the principle that the State cannot tell a newspaper what it must print,” and that the concerns present in Miami Herald — forced speech, chilling of debate — were not present because the plaintiffs sought “to exercise state-protected rights of expression and petition.” PruneYard, 447 U.S. at 88. The Court, though, made clear in a previous section of its opinion that the rights needed to be exercised in a situation where those “activities [did] not interfere with normal business operations.” Id. at 78. The Supreme Court qualified PruneYard just six years later in Pacific Gas & Electric Company v. Public Utilities Commission of California, 475 U.S. 1 (1986) (plurality op.) (“PG&E”). As the majority in our present case discusses, a plurality of the PG&E Court held that the California Public 98 Case: 21-51178 Document: 00516474571 Page: 99 Date Filed: 09/16/2022 No. 21-51178 Utilities Commission’s order to allocate space in PG&E’s newsletter to third party groups that opposed PG&E’s own messages at certain times throughout the year violated PG&E’s First Amendment rights. Id. at 20–21. In doing so, the plurality explained the limits of PruneYard: “notably absent from PruneYard was any concern that access . . . might affect the shopping center owner’s exercise of his own right to speech.” Id. at 12. Justice Marshall, who contributed the fifth vote and concurred in the judgment, agreed, observing that the PruneYard mall’s owner did not want speech by the students, but “he nowhere alleged that his own expression was hindered in the slightest.” Id. at 24 (Marshall, J., concurring in the judgment). The regulations infringed on PG&E’s speech, though; the order, affording rebuttal space to opposing parties could chill speech if PG&E found that “the safe course [was] to avoid controversy,” and the regulations would “abridge [PG&E’s] own rights in order to enhance the relative voice of its opponents.” Id. at 14 (plurality op.). The Court subsequently applied the principles outlined in those precedents in the context of cable television. Turner Broad. Sys., Inc. v. FCC, 512 U.S. 622 (1994) (“Turner I”). The Court considered federal regulations requiring cable operators to set aside certain channels for commercial broadcast stations. See id. at 630; id. at 674 (Stevens, J., concurring in the judgment). Most obviously, these rules burdened cable programmers “by reducing the number of channels for which they [could] compete.” Id. at 645. Writing for a majority of the Court, though, Justice Kennedy further explained that cable operators also “engage in and transmit speech” when, “[t]hrough ‘original programming or by exercising editorial discretion over which stations or programs to include in its repertoire,’ cable programmers and operators ‘see[k] to communicate messages on a wide variety of topics and in a wide variety of formats.’” Id. at 636 (quoting Los Angeles v. Preferred 99 Case: 21-51178 Document: 00516474571 Page: 100 Date Filed: 09/16/2022 No. 21-51178 Comms., Inc., 476 U.S. 488, 494 (1986) (establishing the same)). 3 In other words, the must-carry provisions “interfere[d] with cable operators’ editorial discretion by compelling them to offer carriage to a certain minimum number of broadcast stations,” even though they did so in a way that did “not depend upon the content of the cable operators’ programming.” Id. at 643– 44. As is relevant to Part II of this opinion, the Turner I majority then considered the level of scrutiny appropriate for the must-carry rules and whether the laws met that level of scrutiny. Id. at 641–61. The majority rejected the Government’s argument that rational basis scrutiny should apply, but also decided against the cable operators’ contention that Miami Herald and PG&E dictated strict scrutiny. Id. at 640–41, 661–62. As part of its scrutiny analysis, the majority found three considerations present in Turner I that were not present in Miami Herald and PG&E : (1) the rules were “content neutral” because they were “not activated by any particular message spoken by cable operators”; (2) the rules would not “force cable operators to alter their own messages to respond to the broadcast programming they are required to carry”; and (3) there was “physical control” by the cable operators over a piece of communications infrastructure. Id. at 655–57. These factors, together, suggested a lower tier of scrutiny should be applied. Id. at 661–62. In sum, First Amendment rights were exercised in two ways in Turner I: (1) the speech of cable programmers when they transmitted their own message, and (2) the exercise of “editorial discretion.” Id. at 636. The 3 Although the four dissenting Justices did not join this part of the opinion, they agreed that the must-carry rules implicated the First Amendment rights of cable operators. Id. at 675 (O’Connor, J., dissenting). They would have labeled the rules as unconstitutional content-based restrictions on the cable operators’ speech. Id. at 685. 100 Case: 21-51178 Document: 00516474571 Page: 101 Date Filed: 09/16/2022 No. 21-51178 regulations were held to be content neutral regulations — though unquestionably regulations on First Amendment-protected expression — and the case was remanded for further factfinding to determine whether summary judgment for the Government was appropriate. Id. at 662–63; id. at 669 (Stevens, J., concurring in the judgment). 4 The must-carry rules were then upheld under the intermediate scrutiny standard for content neutral regulations on speech when the case returned to the Supreme Court. Turner Broad. Sys., Inc. v. FCC, 520 U.S. 180, 224–25 (1997) (“Turner II”). The very next term, the Supreme Court decided Hurley v. Irish- American Gay, Lesbian, and Bisexual Group of Boston, 515 U.S. 557 (1995). In considering whether a parade, recognized by Massachusetts’s highest court as a public accommodation but organized by a private party, could be forced under state law to include participation by an organization of gay, lesbian, and bisexual individuals, the Court held that a parade was “a form of expression.” Id. at 568. In identifying protected expression, the Hurley Court did not stop there: “The protected expression that inheres in a parade is not limited to its banners and songs, however, for the Constitution looks beyond written or spoken words as mediums of expression.” Id. at 569. The Court analyzed Turner I and Miami Herald, reiterating that “[c]able operators . . . are engaged in protected speech activities even when they only select programming originally produced by others,” and that “the presentation of an edited compilation of speech generated by other persons . . . fall[s] squarely within the core of First Amendment security . . . 4 The majority gleans a separate insight from Turner I: “Most significant for our purposes, even the four dissenting Justices believed Congress could have permissibly imposed more modest common carrier regulations.” See Maj. Op. at 60. I discuss common carrier treatment below. Most significant for me, though, is that all Justices — in the majority and dissent — understood that some degree of First Amendment scrutiny attended the must-carry rules. See 512 U.S at 675 (O’Connor, J., dissenting). 101 Case: 21-51178 Document: 00516474571 Page: 102 Date Filed: 09/16/2022 No. 21-51178 as does even the simple selection of a paid noncommercial advertisement for inclusion in a daily paper.” Id. at 570 (citing Turner I, 512 U.S. at 636; Miami Herald, 418 U.S. at 258; New York Times v. Sullivan, 376 U.S. 254, 265–66 (1964)). This selection needed not be based on any particular theme, as the Court pointed out that one “does not forfeit constitutional protection simply by combining multifarious voices, or by failing to edit their themes to isolate an exact message as the exclusive subject matter of the speech.” Id. at 569– 70. This constituted the Court’s clear statement that protected expression lies not merely in the message or messages transmitted but in the process of collecting and presenting speech. Finally, there is Rumsfeld v. Forum for Academic and Institutional Rights, Inc., 547 U.S. 47 (2006) (“FAIR”). In FAIR, the Court analyzed Miami Herald, PG&E, and Hurley in the context of a group of law schools seeking a declaratory judgment against the enforcement of the Solomon Amendment, a law that denied federal funding to schools that did not give military recruiters “access to students that is at least equal in quality and scope to the access provided other potential employers.” Id. at 54, 63 (quotation marks and citation omitted). In upholding the constitutionality of the Solomon Amendment, the Court held that “[t]he compelled speech violation in each of our prior cases . . . resulted from the fact that the complaining speaker’s own message was affected by the speech it was forced to accommodate.” Id. at 63. “[B]ecause the schools are not speaking when they host interviews and recruiting receptions,” the regulation did “not affect the law schools’ speech.” Id. at 64. This was because “[u]nlike a parade organizer’s choice of parade contingents, a law school’s decision to allow recruiters on campus is not inherently expressive” and “recruiting services lack the expressive quality of a parade, a newsletter, or the editorial page of a newspaper.” Id. Further, as in PruneYard, there was “little likelihood that the views of those engaging in expressive activity would be 102 Case: 21-51178 Document: 00516474571 Page: 103 Date Filed: 09/16/2022 No. 21-51178 identified with the [property] owner” because “[n]othing about recruiting suggests that law schools agree with any speech by recruiters, and nothing in the Solomon Amendment restricts what the law schools may say about the military’s policies.” Id. at 65. This review of the authorities provides the material for my conclusion that the Miami Herald opinion is the most comparable to what is before us in this appeal. When the Platforms curate their users’ feeds, which are the behaviors prohibited in Section 7 of HB 20, they are exercising their editorial discretion. That is a type of First Amendment-protected activity recognized in Miami Herald, PG&E, Turner, and Hurley. The majority disagrees that editorial discretion is a category, instead asserting that the Supreme Court has merely “treated editorial discretion as a factual consideration supporting their legal conclusions about the presence or absence of protected speech.” Further, the majority concludes that “[n]either [Miami Herald nor Turner I] implied that editorial discretion is itself a freestanding category of constitutionally protected expression.” Maj. Op. at 35 (emphasis omitted). Respectfully, such an interpretation disregards the Supreme Court’s recognition that there may be more than one type of First Amendment activity occurring by the same speaker when, for instance, an article is selected and printed in a newspaper — or, in our context, a tweet posted or video listed. If anything, the majority’s research and reasoning supports the Platforms’ contention that First Amendment protections attend the publishing process as well as the actual published content. I do not read PruneYard and FAIR to suggest anything to the contrary. The hosting mandate upheld by the PruneYard Court did not interfere with speech published and adopted by the shopping mall, nor did it interfere with a selection process for determining which speech was permitted. As the Eleventh Circuit recently remarked, “the only First Amendment interest that the mall owner asserted was the right ‘not to be forced by the state to use 103 Case: 21-51178 Document: 00516474571 Page: 104 Date Filed: 09/16/2022 No. 21-51178 [its] property as a forum for the speech of others.’” NetChoice, 34 F.4th at 1215 (quoting PruneYard, 447 U.S. at 85)). PG&E and Hurley both suggest that this lack of alleged speech activity by the PruneYard proprietor was operative in the analysis. See PG&E, 475 U.S. at 11–12; Hurley, 515 U.S. at 580. FAIR also demonstrates this distinction. In that case, the law schools attempted to draw an analogy to Hurley, arguing that hosting military recruiters unconstitutionally compelled the schools to accommodate the military’s message. FAIR, 547 U.S. at 63. The FAIR Court distinguished Hurley by making clear that “the expressive nature of a parade was central to [the] holding,” and that because “‘every participating unit affects the message conveyed by the . . . organizers,’” a law dictating inclusion of a particular group “alters the expressive content of the parade.” Id. (quoting Hurley, 515 U.S. at 572–73). There was no “inherently expressive” nature to a law school’s decision to allow recruiters on campus, though. Id. at 64. The Court explained that “a law school’s recruiting services lack the expressive quality of a parade, a newsletter, or the editorial page of a newspaper.” Id. The same simply cannot be said for the Platforms. Expression is the very core of their identity and existence. In short, although PruneYard and FAIR establish situations in which the Supreme Court has “upheld government regulations that effectively compelled private actors to ‘host’ others’ speech,” in neither case did the Supreme Court uphold regulations that interfered with the private actors’ own speech. See NetChoice, 34 F.4th at 1215–16. I see no importance to the fact that the Platforms’ moderation will usually follow actual publication. Contra Maj. Op. at 37–38. In the Platforms’ world, it is usually the only practical means to moderate content. Certainly, in those instances in which a particular speaker is barred entirely, 104 Case: 21-51178 Document: 00516474571 Page: 105 Date Filed: 09/16/2022 No. 21-51178 the discretion is exercised in advance. Platforms may also use technology to screen out content they believe does not match their terms of service. Unlike traditional publications, though, where editorial discretion will precede publishing, the majority of decisions on moderating what has been posted can only be made, as a practical matter, after the appearance of the content on the Platform. As discussed later, Congress recognized this reality through the passage of Section 230. I am aware of no authority that denies First Amendment rights to otherwise-protected speech based on similar questions about timing. Editorial discretion is exercised when it is sensible and, in many situations, even possible to do so. The First Amendment fits new contexts and new technologies as they arise. II. Implications of content moderation as speech With the understanding that the Platforms are in fact engaging in First Amendment expression, I turn to the task of determining whether it is likely that HB 20 impermissibly infringes on that expression. As the Supreme Court discussed in a compelled-speech case last term, plaintiffs bringing facial challenges usually “must establish that no set of circumstances exists under which the [law] would be valid or show that the law lacks a plainly legitimate sweep.” Americans for Prosperity Found. v. Bonta, 141 S. Ct. 2373, 2387 (2021) (quotation marks and citation omitted). “[T]he First Amendment context,” though, implicates “a second type of facial challenge, whereby a law may be invalidated as overbroad if a substantial number of its applications are unconstitutional, judged in relation to the statute’s plainly legitimate sweep.” Id. (quoting United States v. Stevens, 559 U.S. 460, 473 (2010)). Overbreadth analysis is proper in challenges to compelled speech as in Bonta and in challenges to statutory limitations on speech as was the case in Stevens, where the Court considered 105 Case: 21-51178 Document: 00516474571 Page: 106 Date Filed: 09/16/2022 No. 21-51178 a law criminalizing the creation, sale, or possession of depictions of animal cruelty, widely defined. See Stevens, 559 U.S. at 464, 474. I join the majority in concluding that the overbreadth doctrine is the proper mode of analysis for this case. See Maj. Op. Part III.A. The question is whether “a substantial number of [HB 20’s] applications are unconstitutional, judged in relation to the statute’s plainly legitimate sweep.” Bonta, 141 S. Ct. at 2387. Because the First Amendment applies, we must decide the applicable level of scrutiny. I do not have confidence about the level of scrutiny that the Supreme Court will one day apply to activities such as those engaged in by these platforms. It is sufficient now to accept the majority’s conclusion that intermediate scrutiny applies to Section 7. 5 I can agree because Section 7’s restrictions on the Platforms’ speech do not survive such scrutiny. Intermediate scrutiny analysis in the First Amendment context allows content neutral regulations upon the finding of three elements: A content-neutral regulation will be sustained . . . [1] if it furthers an important or substantial governmental interest; [2] if the governmental interest is unrelated to the suppression of free expression; and [3] if the incidental restriction on alleged First Amendment freedoms is no greater than is essential to the furtherance of that interest. Turner I, 512 U.S. at 662 (quoting United States v. O’Brien, 391 U.S. 367, 377 (1968)). 5 See also Netchoice LLC, 34 F.4th at 1223–27 (acknowledging that strict scrutiny may apply to several provisions of a similar law but analyzing those provisions under intermediate scrutiny since the provisions were unlikely to withstand even the lower tier of scrutiny). I also question whether at least some of Section 7’s provisions are content neutral. 106 Case: 21-51178 Document: 00516474571 Page: 107 Date Filed: 09/16/2022 No. 21-51178 Texas can satisfy the first two elements if it establishes that Section 7 of HB 20 serves an important or substantial government interest unrelated to the suppression of free expression. Certainly, this does not mean Texas must adopt the “least restrictive alternative,” a test for a different level of scrutiny, but these elements are significant demands. In Turner II, a plurality of the Court referred to three substantial governmental interests: “1) preserving the benefits of free, over-the-air local broadcast television, 2) promoting the widespread dissemination of information from a multiplicity of sources, and 3) promoting fair competition in the market for television programming.” 520 U.S. at 189 (quotation marks and citation omitted); see id. at 226 (Breyer, J., concurring in part) (accepting rationales 1 and 2 but recognizing that must- carry regulation “extracts a serious First Amendment price. It interferes with the protected interests of the cable operators to choose their own programming.”). My able colleagues in the majority argue the second interest applies here — “promoting the widespread dissemination of information from a multiplicity of sources.” Unlike in Turner, though, the Texas statute strives to promote speech by first targeting the content of others’ speech, i.e., it prohibits Platform “censorship” on the basis of viewpoint. (I acknowledge that, yet again, the fundamental division between my view and that of the majority is whether the Platforms are “speaking” when they exercise their editorial discretion.) Texas argues this satisfies the interest recognized in the Turner opinions because it will increase the multiplicity of views on the Platforms — arguably a good result. That justification, though, alters the interest that Turner actually recognized. The Turner must-carry rules did not directly target cable-operators’ editorial discretion. Instead, the must-carry rules supported the interest of the non-cable subscribing public in accessing information without needing to use the cable operators’ platforms. The regulations sought to improve the 107 Case: 21-51178 Document: 00516474571 Page: 108 Date Filed: 09/16/2022 No. 21-51178 viability of traditional commercial broadcast media in order “to prevent too precipitous a decline in the quality and quantity of programming choice for an ever-shrinking non-cable-subscribing segment of the public.” Turner II, 520 U.S. at 226 (Breyer, J., concurring) (adding the fifth vote to affirm the Government’s interest in “promoting widespread dissemination of information from a multiplicity of sources”). Indeed, any interference with the cable operators’ speech to promote the traditional broadcaster’s ability to speak was the “price” and not the purpose of the regulation. Id. Here, of course, interference with expressing views is both the purpose and the price. HB 20 directly interferes with the editorial choices the Platforms make — which I consider First Amendment expression — as both a means and end. In Turner, the cable operators could displace any programming they wanted in order to make room for local commercial broadcast media, thereby helping local broadcast stations survive that new technology. See Turner I, 512 U.S. at 636–37 (acknowledging the set-aside for local broadcasters and that it would be “more difficult for cable programmers to compete for carriage on the limited channels remaining”). Had the justification for the must-carry rules been only a governmental interest of having cable operators express additional views, the rules should have been struck down because of Miami Herald. The Court has recognized that the state “may not burden the speech of others in order to tilt public debate in a preferred direction.” Sorrell v. IMS Health Inc., 564 U.S. 552, 578–79 (2011); see also Buckley v. Valeo, 424 U.S. 1, 48–49 (recognizing that there is no interest in “restrict[ing] the speech of some elements of our society in order to enhance the relative voice of others” as the First Amendment “was designed to secure the widest possible dissemination of information from diverse and antagonistic sources” (quotation marks and citation omitted)). 108 Case: 21-51178 Document: 00516474571 Page: 109 Date Filed: 09/16/2022 No. 21-51178 I agree with the Eleventh Circuit when it reiterated the message from Miami Herald: “preventing unfairness to certain users or points of view isn’t a substantial government interest; rather private actors have a First Amendment right to be ‘unfair’ — which is to say, a right to have and express their own points of view.” NetChoice, 34 F.4th at 1228 (quotation marks and citation omitted). That is the case here. Further regarding the relevance of unfairness, the majority considers it extraordinary that counsel for one of the Platforms at oral argument answered a question from the court by agreeing a Platform could, as the majority opinion states, “ban all pro-LGBT speech for no other reason than its employees want to pick on members of that community.” Maj. Op. at 2. Extreme hypotheticals necessarily lead to extreme answers when a First Amendment right is involved. The First Amendment does not moderate its protections based on the content of the speech, with irrelevant exceptions. In no manner am I denying the reasonableness of the governmental interest. When these Platforms, that for the moment have gained such dominance, impose their policy choices, the effects are far more powerful and widespread than most other speakers’ choices. The First Amendment, though, is not withdrawn from speech just because speakers are using their available platforms unfairly or when the speech is offensive. The asserted governmental interest supporting this statute is undeniably related to the suppression of free expression. The First Amendment bars the restraints. Setting aside that the purpose of Texas’s law is related to suppressing First Amendment activity, I also believe there is a strong likelihood that Section 7 burdens “substantially more speech than necessary in order to further [Texas’s] legitimate interests.” See Turner I, 512 U.S. at 662. The scope of conduct prohibited by Section 7 is broad: 109 Case: 21-51178 Document: 00516474571 Page: 110 Date Filed: 09/16/2022 No. 21-51178 A social media platform may not [block, ban, remove, deplatform, demonetize, de-boost, restrict, deny equal access or visibility to, or otherwise discriminate against] a user, a user’s expression, or a user’s ability to receive the expression of another person based on [the user’s viewpoint, the viewpoint represented, or geographic location]. Tex. Civ. Prac. & Rem. Code, §§ 143A.001–002. If Texas’s interest is in “protecting the free exchange of ideas and information in this state,” prohibitions (for example) on demonetization, de- boosting, “denying equal access or visibility to” or “otherwise discriminat[ing] against,” likely go too far. If the goal is only to make more speech available, there is no reason that the Platforms should have to publish — as an extreme example — pro-Nazi expression, while monetizing, recommending, and giving equal treatment to such content as might be given to anti-Nazi expression. When Platforms elevate certain third-party content above other third-party content, they engage in their own First Amendment expression, and the broad-based prohibition against engaging in this editorial discretion whenever “viewpoint” is at issue is hardly narrow tailoring that “does not burden substantially more speech than necessary” to further a legitimate interest. See Turner I, 512 U.S. at 602. III. Common carrier designation, Section 230, and other rationales for abrogating First Amendment rights One of my colleagues concludes that common carrier classification of the Platforms and Section 230 provide further support for the constitutionality of HB 20. I address both arguments. A common carrier designation, which I doubt is appropriate, would not likely change any of my preceding analysis. Few of the cases cited in the discussion on common carrier law concern the intersection of common carrier obligations and First Amendment speech rights. The only precedents 110 Case: 21-51178 Document: 00516474571 Page: 111 Date Filed: 09/16/2022 No. 21-51178 that do discuss this intersection reinforce the idea common carriers retain their First Amendment protections for their own speech. See id. at 636. Section 230 also does not affect the First Amendment right of the Platforms to exercise their own editorial discretion through content moderation. My colleague suggests that “Congress’s judgment” as expressed in 47 U.S.C. § 230 “reinforces our conclusion that the Platforms’ censorship is not speech under the First Amendment.” Maj. Op. at 39. That opinion refers to this language: “No provider or user of an interactive computer service” — interactive computer service being a defined term encompassing a wide variety of information services, systems, and access software providers — “shall be treated as the publisher or speaker of any information provided by another content provider.” 47 U.S.C. § 230(c)(1). Though I agree that Congressional fact-findings underlying enactments may be considered by courts, the question here is whether the Platforms’ barred activity is an exercise of their First Amendment rights. If it is, Section 230’s characterizations do not transform it into unprotected speech. The Platforms also are criticized for what my colleague sees as an inconsistent argument: the Platforms analogize their conduct to the exercise of editorial discretion by traditional media outlets, though Section 230 by its terms exempts them from traditional publisher liability. This may be exactly how Section 230 is supposed to work, though. Contrary to the contention about inconsistency, Congress in adopting Section 230 never factually determined that “the Platforms are not ‘publishers.’” Maj. Op. at 41. As one of Section 230’s co-sponsors — former California Congressman Christopher Cox, one of the amici here — stated, Section 230 merely established that the platforms are not to be treated as the publishers of pieces of content when they take up the mantle of content moderation, which was precisely the problem that Section 230 set out to solve: “content moderation . . . is not only consistent with Section 230; its protection is the 111 Case: 21-51178 Document: 00516474571 Page: 112 Date Filed: 09/16/2022 No. 21-51178 very raison d’etre of Section 230.” In short, we should not force a false dichotomy on the Platforms. There is no reason “that a platform must be classified for all purposes as either a publisher or a mere conduit.” In any case, as Congressman Cox put it, “because content moderation is a form of editorial speech, the First Amendment more fully protects it beyond the specific safeguards enumerated in § 230(c)(2).” I agree. IV. Other preliminary injunction factors In reviewing the grant of a preliminary injunction, this court considers three other factors in addition to the likelihood of success on the merits: the substantial threat of irreparable harm should the injunction not be granted, the balance of harms, and the public interest. See Atchafalaya Basinkeeper v. United States Army Corps of Eng’rs, 894 F.3d 692, 696 (5th Cir. 2018). “Loss of First Amendment freedoms, even for minimal periods of time, constitutes irreparable injury.” Ingebretsen ex rel. Ingebretsen v. Jackson Pub. Sch. Dist., 88 F.3d 274, 280 (5th Cir. 1996). Accordingly, to prevent the Platforms from establishing that the balance of such harms weighs in their favor, Texas “would need to present powerful evidence of harm to its interests.” See Opulent Life Church v. City of Holly Springs, 697 F.3d 279, 298 (5th Cir. 2012). Further, “[i]njunctions protecting First Amendment freedoms are always in the public interest.” Id. (quotation marks and citation omitted). Because I see the enforcement of the anti-discrimination provisions of Section 7 of HB 20 as likely unconstitutional infringements on First Amendment freedoms, these factors would also favor preliminary relief against those provisions. V. Conclusion This is a difficult case. We are seeking the closest analogies among the precedents. The Supreme Court will, as always, have the final word. For now, I conclude Section 7’s anti-discrimination provisions are an 112 Case: 21-51178 Document: 00516474571 Page: 113 Date Filed: 09/16/2022 No. 21-51178 unconstitutional infringement on the Plaintiffs’ rights to edit or remove, after the fact, speech that appears on their private Platforms. My understanding of their rights does not mean that “email providers, mobile phone companies, and banks could cancel the accounts of anyone who sends an email, makes a phone call, or spends money in support of a disfavored political party, candidate or business,” as suggested by the majority. It does mean that when the social media Platforms who are in the business of speech make decisions about which speech is permitted, featured, promoted, boosted, monetized, and more, they are engaging in activity to which First Amendment protection attaches. Balance and fairness certainly would be preferable, but the First Amendment does not require it. I concur with the judgment in Part IV of the majority’s opinion. I respectfully dissent from the remainder. 113