Congress passed § 230 of the Communications Decency Act of 19961×1. 47 U.S.C. § 230. to reduce internet intermediaries’ liability for third-party content.2×2. See Eric Goldman, Essay, Why Section 230 Is Better than the First Amendment, 95 Notre Dame L. Rev. Reflection 33, 33, 36 (2019). In the first appellate case to interpret the scope of § 230’s protections, the Fourth Circuit read the statute broadly by invoking the collateral censorship theory.3×3. See Zeran v. Am. Online, Inc., 129 F.3d 327, 331 (4th Cir. 1997); Felix T. Wu, Collateral Censorship and the Limits of Intermediary Immunity, 87 Notre Dame L. Rev. 293, 317 (2011) (noting that Zeran relied on “collateral censorship, although the court did not use that term”). Collateral censorship is a form of indirect censorship that occurs when the imposition of liability on a private intermediary incentivizes it to censor another private party’s speech.4×4. See Wu, supra note 3, at 295–96. Although the theory looms large in § 230 precedent, the Supreme Court has not extended it to First Amendment cases involving internet services.5×5. Jack M. Balkin, Old-School/New-School Speech Regulation, 127 Harv. L. Rev. 2296, 2312–13 (2014). Recently, in Washington Post v. McManus,6×6. 944 F.3d 506 (4th Cir. 2019). the Fourth Circuit relied on the logic of collateral censorship to invalidate a Maryland campaign finance disclosure law aimed at online campaign advertising.7×7. See id. at 516. It did so under an exclusively First Amendment inquiry, despite the fact that plaintiffs had also claimed relief under § 230.8×8. See id. at 510; Wash. Post v. McManus, 355 F. Supp. 3d 272, 284 (D. Md. 2019). By deciding the case on constitutional as opposed to statutory grounds, the Fourth Circuit took a step toward the constitutionalization of collateral censorship as a First Amendment basis for exempting internet intermediaries from otherwise applicable laws.
The 2016 U.S. presidential election revealed an online electioneering landscape that had far outpaced its regulatory counterpart.9×9. See Brian Beyersdorf, Note, Regulating the “Most Accessible Marketplace of Ideas in History”: Disclosure Requirements in Online Political Advertisements After the 2016 Election, 107 Calif. L. Rev. 1061, 1063–64 (2019). This revelation, coupled with partisan gridlock at the federal level,10×10. Id. at 1065. prompted several state efforts to extend campaign disclosure requirements to online advertising.11×11. See McManus, 944 F.3d at 510. Maryland’s Online Electioneering Transparency and Accountability Act12×12. Ch. 834, 2018 Md. Laws 4220 (codified in scattered sections of Md. Code Ann., Elec. Law). was one such effort, and it imposed disclosure obligations on websites that reached a certain circulation and received payment for qualifying ads.13×13. McManus, 944 F.3d at 511–12 (citing Md. Code Ann., Elec. Law §§ 1-101(dd-1), 13-405(b)–(c) (Michie, LEXIS through legislation effective Nov. 6, 2020)). First, platforms had to disclose on their websites “the identity of the purchaser, the individuals exercising control over the purchaser, and the total amount paid for the ad” (the “publication requirement”).14×14. Id. at 511 (citing Md. Code Ann., Elec. Law § 13-405(b) (LEXIS)). The information had to be posted within forty-eight hours of the purchase and kept on the website for at least a year following the relevant election. Id. at 511–12. Second, platforms had to collect and maintain a more extensive record for inspection by the Maryland Board of Elections upon state request (the “inspection requirement”).15×15. See id. at 512, 514 (citing Md. Code Ann., Elec. Law § 13-405(c) (LEXIS)). The inspection record included additional information such as “the candidate or ballot issue” addressed by the ad, the dates of first and last dissemination, a digital copy of the ad’s content, the approximate geographic location of dissemination, a description of the audience targeted or receiving the ad, and “the total number of impressions generated.” Id. at 514. Shortly after the law took effect, a group of news outlets filed for a preliminary injunction, alleging facial and as-applied violations of the First Amendment, among other issues.16×16. See Wash. Post v. McManus, 355 F. Supp. 3d 272, 284 (D. Md. 2019). The publishers also argued that the provision was “unconstitutionally vague, . . . authorize[d] an unconstitutional seizure . . . in violation of the Fourth Amendment,” and was preempted by § 230. Id. The district court found the First Amendment claim sufficient for the plaintiffs to succeed in obtaining a preliminary injunction. Id. at 285.
The district court granted the news outlets’ motion.17×17. Id. at 306. The court limited itself to the outlets’ First Amendment claim, finding that the law was a content-based regulation of political speech and therefore subject to strict scrutiny.18×18. Id.“>https://www.zotero.org/google-docs/?SCA9xf”>Id. at 287, 297. While the court acknowledged that “First Amendment case law is fertile with exceptions to the various fundamental rules,”19×19. Id. at 287. it maintained that this case did not fit within any of those exceptions.20×20. Id. at 297. Specifically, the court rejected the State’s argument that the Act fell under the campaign finance disclosure exception articulated by the Supreme Court in Buckley v. Valeo,21×21. 424 U.S. 1 (1976) (per curiam). and that it should thus be viewed under the less burdensome “exacting scrutiny” standard.22×22. McManus, 355 F. Supp. 3d at 297; see id. at 288–97. Still, the court found that the Act would fail even under the less stringent standard of exacting scrutiny.23×23. Id. at 302. Although the court found it evident that the State had compelling interests in preventing foreign meddling in elections, informing voters, and deterring corruption, the Act was overinclusive and underinclusive because, among other reasons, it was duplicative of advertisers’ disclosure obligations, did not target fake websites and free social media posts, applied to a wide range of platforms, and was triggered only by ad buyers’ disclosure that the law applied to their ads. Id. at 298–305. Maryland appealed to the Fourth Circuit.24×24. See McManus, 944 F.3d at 513.
The Fourth Circuit affirmed on largely the same grounds.25×25. Id. at 510. Writing for a unanimous panel, Judge Wilkinson26×26. Judge Wilkinson was joined by Judges Motz and Floyd. detailed the “compendium of traditional First Amendment infirmities” afflicting the law.27×27. McManus, 944 F.3d at 513. “First, the Act [wa]s a content-based regulation [of] speech” because it applied to “campaign-related speech” only.28×28. Id. Second, the Act targeted political speech.29×29. Id. And last, the court held that the law compelled speech by forcing the outlets to disclose on their websites information that they otherwise would not disclose.30×30. Id. at 514. Moreover, the panel concluded that the Act’s inspection requirement also compelled speech by requiring platforms to “collect and retain” details on campaign ads “to be disclosed to state regulators upon request.”31×31. Id.
The State’s characterization of the law as a campaign finance regulation was unavailing. The court recognized that under Buckley and its progeny, content-based regulations of political speech through campaign finance disclosures are permissible, because although they “may burden the ability to speak,” they “do not prevent anyone from speaking.”32×32. Id. at 516 (quoting Citizens United v. FEC, 558 U.S. 310, 366 (2010)). But that logic is limited to regulations aimed at “direct participants in the political process,” not “neutral third-party platforms” whose primary interest is not political substance, but revenue.33×33. Id. The court reasoned that political actors’ ambition to succeed at the ballot box “generally offsets, at least in part, whatever burdens are posed by disclosure obligations.” Id. It reasoned that the Act’s disclosure obligations and their attendant compliance costs made hosting political ads financially more burdensome, deterring profit-oriented platforms from hosting political ads.34×34. Id. The court found that this deterrence would “necessarily reduce the quantity of [political] expression,” and held that such a law was outside the scope of Buckley.35×35. Id. at 517 (first alteration in original) (emphasis added) (quoting Buckley v. Valeo, 424 U.S. 1, 19 (1976) (per curiam)). The panel added that the Act’s application to the press was especially troublesome. Because media outlets’ advertising choices receive First Amendment protections, the publication requirement was an intrusion into the editorial discretion of the press.36×36. Id. at 518. The inspection requirement was similarly problematic because it created an “unhealthy entanglement” between the state and news outlets.37×37. Id.
While the court declined to declare an exact standard of review for a disclosure law like Maryland’s,38×38. Id. at 520. it held that the law failed even under the less demanding “exacting scrutiny,”39×39. Id. which requires a “substantial relation” between an “important” state interest and the regulation.40×40. Id. (first quoting Buckley, 424 U.S. at 64; and then quoting id. at 66). The court found that the law was underinclusive because it failed to address unpaid social media posts, the primary mechanism of foreign interference.41×41. Id. at 521. And while the law applied to ads for specific candidates or ballot initiatives, it did not reach those focused on general, divisive issues, leaving the “vast majority” of foreign-placed ads untouched.42×42. Id. The court also rejected Maryland’s argument that the law, though underinclusive, was the least restrictive means available to the State because of Maryland’s concern about the First Amendment difficulties attendant in the regulation of unpaid speech. Id. The law was also overinclusive. The court reasoned that Maryland failed to provide evidence of foreign-sourced ads on news sites, and the Act did not properly distinguish among platforms based on size, susceptibility to foreign meddling, or electoral influence.43×43. Id. at 521–22. The Act painted with too “broad [a] brush” because giant social media websites like Facebook and Instagram were the primary playground for foreign interference. Id. at 522.
Sometimes “the argument for a particular rule may be more important than the argument from that rule to the particular case.”44×44. Ronald Dworkin, Taking Rights Seriously 112 (1977). That may be the case with McManus’s collateral censorship argument for exempting online intermediaries from the application of standard First Amendment doctrine. Although the Supreme Court has not extended collateral censorship to its First Amendment jurisprudence involving internet intermediaries, collateral censorship plays a significant role in statutory interpretations of § 230, including the Fourth Circuit’s.45×45. See, e.g., Wu, supra note 3, at 317. Given that the publishers had claimed preemption under § 230,46×46. Wash. Post v. McManus, 355 F. Supp. 3d 272, 284 (D. Md. 2019). the court could have reached the same substantive conclusion using the collateral censorship logic of its § 230 precedent. Instead, the court applied that logic under an exclusively First Amendment analysis. By so deciding, the Fourth Circuit took a step toward the constitutionalization of collateral censorship as a First Amendment basis for exempting internet intermediaries from otherwise applicable laws.
The court’s argument for exempting online platforms from the Buckley framework relied on the logic of the so-called collateral censorship phenomenon47×47. See Michael I. Meyerson, Authors, Editors, and Uncommon Carriers: Identifying the “Speaker” Within the New Media, 71 Notre Dame L. Rev. 79, 116–18 (1995). : when the imposition of liability and compliance costs on private intermediary A incentivizes it to use its power to censor the speech of private speaker B.48×48. J.M. Balkin, Essay, Free Speech and Hostile Environments, 99 Colum. L. Rev. 2295, 2298 & n.14 (1999) (crediting Professor Michael Meyerson for coining the term); see also Seth F. Kreimer, Censorship by Proxy: The First Amendment, Internet Intermediaries, and the Problem of the Weakest Link, 155 U. Pa. L. Rev. 11, 27–33 (2006) (using the term “proxy censorship” instead). The Fourth Circuit reasoned that the Act’s compliance costs would disincentivize platforms from hosting political ads because, unlike primary speakers, intermediaries have a primarily financial rather than expressive interest.49×49. See McManus, 944 F.3d at 516. This concern aligns with collateral censorship theory, which argues that because of their “economic and instrumental” interests, online intermediaries are particularly susceptible to chilling effects50×50. David S. Ardia, Free Speech Savior or Shield for Scoundrels: An Empirical Study of Intermediary Immunity Under Section 230 of the Communications Decency Act, 43 Loy. L.A. L. Rev. 373, 379 (2010); see also Wu, supra note 3, at 304–08. and should therefore be overprotected from otherwise constitutional speech regulations.51×51. See, e.g., Kreimer, supra note 48, at 77 (“[E]fforts to recruit intermediaries as proxy censors should generally be viewed as ‘more intrusive’ for First Amendment purposes than efforts to regulate speakers or listeners directly.”). This theory is often used in arguments for shielding online platforms from liabilities that apply to traditional speakers or listeners.52×52. See, e.g., Wu, supra note 3, at 304 (“[T]he chilling effects on intermediaries are even greater, and the law ought to account for that difference.”). Moreover, collateral censorship can be characterized as a particular application of the speculative “chilling effects” inquiry,53×53. Id. which considers the incidental effect of legitimate regulation on protected speech54×54. Leslie Kendrick, Speech, Intent, and the Chilling Effect, 54 Wm. & Mary L. Rev. 1633, 1649 (2013). and, as in McManus, can result in “a constitutionally mandated exception to a general doctrinal rule.”55×55. Monica Youn, The Chilling Effect and the Problem of Private Action, 66 Vand. L. Rev. 1473, 1483 (2013). But “chilling effects” is not a discrete doctrine,56×56. Id. and given collateral censorship’s doctrinal hurdles, its invocation does not refute the novelty of McManus’s collateral censorship analysis.
One difficulty with McManus stems from the Supreme Court’s increasing aversion to speaker-based distinctions in the context of campaign finance regulations.57×57. See, e.g., Citizens United v. FEC, 558 U.S. 310, 340–41 (2010); Michael Kagan, Speaker Discrimination: The Next Frontier of Free Speech, 42 Fla. St. U. L. Rev. 765, 781–82 (2015); Asaf Wiener, A Speaker-Based Approach to Speech Moderation and First Amendment Analysis, 31 Stan. L. & Pol’y Rev. 187, 209–10, 215 (2020). McManus distinguished between direct political speakers and “neutral third-party platforms”58×58. McManus, 944 F.3d at 516; see id. at 515–16. by arguing that Citizens United limited the exacting scrutiny standard to disclosure requirements that “do not prevent anyone from speaking.”59×59. Id. at 516 (quoting Citizens United, 558 U.S. at 366). But the court’s interpretation is incompatible with Buckley’s analysis of disclosure requirements, which was endorsed by Citizens United.60×60. Citizens United, 558 U.S. at 366–67. The Court in Buckley considered a facial attack seeking to exempt minority parties and independent candidates from such requirements.61×61. Buckley v. Valeo, 424 U.S. 1, 68–69 (1976) (per curiam). The Court foreclosed such facial attacks and rejected blanket speaker-based exemptions.62×62. Id. at 71, 74. It reasoned that although disclosures would “undoubtedly . . . deter some individuals,”63×63. Id. at 68. the state’s substantial interests foreclosed challenges based on “highly speculative”64×64. Id. at 70. or “generally alleged” harms.65×65. Id. at 72. The Court found testimonies similar to ones in McManus too speculative to outweigh the state interest.66×66. Compare id. at 71–72 (“At best [the plaintiffs] offer the testimony of several minor-party officials that one or two persons refused to make contributions because of the possibility of disclosure.”), with McManus, 944 F.3d at 517 (“[A] candidate for Maryland’s House of Delegates has alleged that Google’s dropoff from political advertising harmed his campaign . . . .”). Most importantly, the Court conceded that minor parties are generally more susceptible to chilling effects,67×67. Buckley, 424 U.S. at 71. but it did not carve out an exception to its exacting scrutiny standard as the court did in McManus. Citizens United upheld Buckley’s framework, so its statement that disclosure requirements “do not prevent anyone from speaking” could not reasonably be construed to limit the application of exacting scrutiny based on such generally alleged and speculative harms as those asserted in McManus.68×68. Citizens United v. FEC, 558 U.S. 310, 366 (2010) (quoting McConnell v. FEC, 540 U.S. 93, 201 (2003) (alteration omitted)). Rather, the statement should be understood as drawing a comparison between disclosure requirements and direct restrictions on speech — namely corporate campaign spending — not indirect chilling effects resulting from private censorship.69×69. See id. at 483 (Thomas, J., concurring in part and dissenting in part) (rejecting as a “fallacy” the Court’s conclusion that disclosure requirements do not prevent anyone from speaking).
The doctrinal difficulties of McManus extend beyond the Court’s campaign finance case law. As Professor Jack Balkin has noted, the “closest the Supreme Court has come to recognizing collateral censorship as a First Amendment issue”70×70. Balkin, supra note 5, at 2312. was in Smith v. California,71×71. 361 U.S. 147 (1959). which invalidated a municipal ordinance that imposed strict liability on booksellers for possession of obscene books, citing the risk that the penalty would incentivize booksellers to censor their wares.72×72. Balkin, supra note 5, at 2312; see Smith, 361 U.S. at 153–54. But the Court has not extended Smith-type protections to online platforms,73×73. See Balkin, supra note 5, at 2313; see also Wu, supra note 3, at 312. which may explain Smith’s absence from the McManus opinion. As of now, there is general scholarly consensus that collateral censorship theory has not been widely adopted into the Court’s First Amendment analysis of state regulations,74×74. See, e.g., Balkin, supra note 5, at 2313; Meyerson, supra note 47, at 117 (lamenting the inconsistencies in courts’ grants of intermediary immunity); Rebecca Tushnet, Power Without Responsibility: Intermediaries and the First Amendment, 76 Geo. Wash. L. Rev. 986, 1008 & n.95 (2008) (discussing how before § 230, “the assumption in the law reviews tended to be that the . . . standard [invoked in New York Times Co. v. Sullivan, 376 U.S. 254 (1964),] was the best to be hoped for as a constitutional matter,” Tushnet, supra, at 1008 n.95). including cases involving election-speech intermediaries.75×75. Samuel S. Sadeghi, Comment, Election Speech and Collateral Censorship at the Slightest Whiff of Legal Trouble, 63 UCLA L. Rev. 1472, 1474, 1487–89 (2016). Although the theory appears in cases involving private causes of action,76×76. See Brent Skorup & Jennifer Huddleston, The Erosion of Publisher Liability in American Law, Section 230, and the Future of Online Curation, 72 Okla. L. Rev. 635, 639 (2020). challenges asserting the First Amendment rights of third parties are still typically subject to the traditional vagueness and overbreadth doctrines.77×77. See Erwin Chemerinsky, Constitutional Law § 11.2.2, at 1388, 1391–92 (5th ed. 2015).
Unlike the Court’s First Amendment jurisprudence, § 230 case law and scholarship have taken special notice of collateral censorship.78×78. See, e.g., Wu, supra note 3, at 317–18, 339. The Fourth Circuit is no stranger to this trend. Indeed, Judge Wilkinson wrote the opinion in Zeran v. America Online, Inc.,79×79. 129 F.3d 327 (4th Cir. 1997). “the seminal case that read [§] 230 broadly”80×80. Goldman, supra note 2, at 36. by relying on the logic of collateral censorship.81×81. Wu, supra note 3, at 317; see Zeran, 129 F.3d at 331 (discussing how intermediary tort liability would lead to “an obvious chilling effect”). Section 230 preempts any state law cause of action or liability that is inconsistent with its grant of immunity.82×82. 47 U.S.C. § 230(e)(3). The Fourth Circuit interpreted this immunity to apply to “any cause of action that would make service providers liable for information originating with a third-party user of the service.”83×83. Zeran, 129 F.3d at 330 (emphasis added). Although the language of § 230(c)(1) grants immunity only against claims seeking to treat platforms as publishers,84×84. 47 U.S.C. § 230(c)(1). courts have generally followed the Fourth Circuit in barring any claim arising from third-party content, regardless of the actual cause of action.85×85. Goldman, supra note 2, at 36–37.
Given collateral censorship’s relative force in § 230 precedent, the McManus court would have stood on firmer ground by resolving the case under § 230. Under Fourth Circuit precedent, § 230 bars liability on the basis of a service provider’s exercise of “traditional editorial functions,” which include decisions about “whether to publish, withdraw, postpone or alter content.”86×86. Zeran, 129 F.3d at 330. As the McManus plaintiffs rightly pointed out, Maryland’s law provided “for the removal of political ads on pain of contempt or criminal sanctions.”87×87. Memorandum of Points & Authorities in Support of Plaintiffs’ Motion for Preliminary Injunction at 30, Wash. Post v. McManus, 355 F. Supp. 3d 272 (D. Md. 2019) (No. 18-cv-02527). Therefore, the Act created liabilities for what amounts to publishers’ exercise of traditional editorial functions, which would have allowed the court to remand or affirm the case on statutory grounds.
Read broadly, McManus suggests that established First Amendment rules may be inapplicable to speech intermediaries whose business models refuse to internalize the harms of their practices, irrespective of the availability of alternative business models. But as Professor Mark Tushnet notes, it is not so clear whether the First Amendment should treat speech intermediaries’ business practices as inevitable, which is what the McManus court seems to assume.88×88. Mark Tushnet, Internet Exceptionalism: An Overview from General Constitutional Law, 56 Wm. & Mary L. Rev. 1637, 1660–61 (2015). To the contrary, current copyright and defamation doctrines reject blanket intermediary exemptions despite the laws’ collateral censorship effects.89×89. On the defamation front, scholars often cite to New York Times Co. v. Sullivan as an example of the Court’s recognition of collateral censorship by intermediaries. See, e.g., Kreimer, supra note 48, at 54. But Sullivan rejected underprotection of an intermediary hosting a paid political ad, rather than standing for overprotection of the intermediary. See New York Times v. Sullivan, 376 U.S. 254, 266 (1964) (holding that otherwise-protected statements “do not forfeit that protection because they were published in the form of a paid advertisement”). For an example on the copyright front, see generally Jennifer M. Urban et al., Notice and Takedown in Everyday Practice (2d ed. 2017). A collateral censorship doctrine would need to address not only doctrinal implications for various areas of the law but also the extent to which its logic is limited to internet intermediaries. Insofar as collateral censorship theory demands a medium-specific approach, it may be to the internet what scarcity was to broadcasting: a distinctive feature resulting in “a distinctive place in First Amendment law.”90×90. McManus, 944 F.3d at 519.
Read narrowly, McManus avoids applying established First Amendment rules to a novel context while refusing to articulate a new rule.91×91. Id. at 520. Given the novelty and significance of challenges like online election interference, the court’s hands-off approach demands an active legislative counterpart that can respond to the novelty that McManus recognizes.92×92. Id. But legislatures act against a backdrop of established rules, and uncertainty about applicable doctrines may hamper legislative action. Under such conditions of legal uncertainty, legislatures will find it difficult — if not impossible — to address issues of online campaign advertising while respecting the Fourth Circuit’s undefined conception of the First Amendment’s reach.