The Law and Economics of Online Republication

I.     Introduction

Jerry publishes unlawful content about Newman on Facebook, Elaine shares Jerry’s post, the share automatically turns into a tweet because her Facebook and Twitter accounts are linked, and George immediately retweets it. Should Elaine and George be liable for these republications? This question is neither theoretical nor idiosyncratic. On occasion, it reaches the headlines, such as when Jennifer Lawrence’s representatives announced she would sue every person involved in the dissemination, through various online platforms, of her illegally obtained nude pictures.1 Yet this is only the tip of the iceberg. Numerous potentially offensive items are reposted daily, their exposure expands in widening circles, and they sometimes “go viral.” Figure 1 illustrates this phenomenon: The black dot denotes the person who generated the content (hereinafter “the originator”), and each circle represents the marginal expansion of exposure attributed to the publication of that content by members of the preceding circle.

Figure 1. Exposure Expansion through Republication


This Article is the first to provide a legal and economic analysis of the question of liability for online republication. Its main thesis is that liability for republication generates a specter of multiple defendants which might dilute the originator’s liability and undermine the deterrent effect of the latter. The Article concludes that, subject to several exceptions and methodological caveats, only the originator should be liable. This seems to be the American rule, as enunciated in Batzel v. Smith2 and Barrett v. Rosenthal.3 It stands in stark contrast to the prevalent rules in other Western jurisdictions4 and has been challenged by scholars on various grounds since its inception.5

Part II of this Article provides a systematic analysis of applicable law.6 Section II.A starts with the development of liability for republication of one’s own publications. Historically, each delivery of unlawful content to a third party was considered a separate publication which gave rise to a new cause of action.7 This is known as the multiple publication rule.8 Later on, American courts recognized an exception, applicable to mass communications, whereby numerous deliveries of a single publication can only underlie a single cause of action.9 The single publication rule was extended to online publication at the turn of the millennium.10

Section II.B turns to liability for republication of third-party content. At common law, if a person republishes unlawful content already published by another, a new publication occurs, and a new cause of action against that person arises.11 This rule is subject to several limited exceptions, such as the “fair report privilege.”12 In the absence of legislative intervention, the common law rule—along with its exceptions—could extend to online publications. However, the enactment of § 230 of the Communications Decency Act (“CDA”)13 has afforded individual republishers of third-party content an almost absolute immunity.14 Section II.B shows that this regime is extreme and unique from a global perspective, by comparing it to more generous liability schemes prevalent in other Western jurisdictions.15 Lastly, Section II.B explains that irrespective of the republisher’s status, the content originator is liable for all foreseeable consequences of the publication, including those of foreseeable republication by others.16

Part III of this Article offers an economic justification for the American model. Section III.A presents the notion of liability dilution and its limits. According to economic theory, efficient deterrence entails internalization by the wrongdoer of the social harm caused by the wrongful conduct.17 Law and economics literature has already observed that attributing liability for a constant indivisible harm to multiple injurers, where each could have single-handedly prevented that harm (“alternative care” settings), might lead to “dilution of liability.” As the overall harm is constant, increasing the number of liable injurers reduces the burden incurred by each one, and impairs the incentives to take cost-effective precautions. Several tort doctrines have been defended in terms of preventing liability dilution.18 Online republication differs from previously analyzed settings, primarily because it does not involve alternative care situations, and because every republication increases exposure and consequent harm.19 Section III.A concludes, however, that neither alternative care nor constant harm is an indispensable precondition for the applicability of the dilution argument.

Section III.B discusses the possible dilution of liability and its impact on deterrence in the context of online republication. It starts with a simple and unrealistic model involving non-overlapping contact lists of average size, explaining how liability is diluted and when this might lead to under-deterrence.20 Subsequently, the Section introduces overlaps and variance in size among user contact lists.21 Section III.C examines another potentially unwarranted outcome of liability for republication: a considerable upsurge in administrative costs. Liability for republication is not only unnecessary when the originator is properly deterred; it might also undermine the deterrent effect of the originator’s liability due to dilution.22 Consequently, the administrative costs cannot be justified in terms of deterrence.

Part IV of this Article considers the possible limits of the American model. Section IV.A maintains that the exclusive originator liability rule must be relaxed, and first-order republisher liability should be recognized, when the originator is unidentifiable (anonymous or pseudonymous) or judgment-proof, and when the republisher’s identity or the republication’s audience is unforeseeable. Section IV.B discusses quasi-exceptions, namely situations in which imposition of liability on a republisher is justified but cannot be regarded as a true exception to the exclusive originator liability rule. The first is a case of republication accompanied by original, substantial, and unlawful content. The republisher’s liability in such a case is not for the act of republication, but for the act of publishing new unlawful content.23 The second is a case of republication of illegally obtained content. Republication with knowledge that the content was illegally obtained is an independent intentional wrong that needs to be addressed either through tort law, using punitive damages to overcome the dilution of liability, or through criminal law.24 Section IV.C answers other possible criticisms of the Article’s thesis—the arguments that dilution does not necessarily result from recognition of liability for republication, and is not always undesirable. Finally, Part V concludes.

II.     The Legal Framework

A.     Republication of Self-Created Content

1.     Offline Publications

The earliest cases in which the question of liability for republication of unlawful content arose were cases of republication by the content’s originator. In the paradigmatic incident, a newspaper or a book publisher was sued for a relatively late delivery of previously published material. To overcome limitation periods which barred civil actions for the original, earlier publications, plaintiffs argued that the late deliveries constituted new publications which gave rise to new unbarred causes of action.

The seminal case was Duke of Brunswick v. Harmer,25 decided by the English Court of Queen’s Bench in 1849. The defendant, a newspaper with minuscule circulation, published an article defaming the plaintiff in 1830.26 More than 17 years later, the plaintiff’s agent purchased a copy of the same issue from the defendant’s office, and the plaintiff brought an action for libel.27 The limitation period for libel was six years, so the defendant argued that the cause of action was time-barred.28 The court established the “multiple publication rule,” whereby each delivery of a defamatory statement to a third party constitutes a separate publication which gives rise to a new cause of action.29 Interestingly, the defendant was found liable even though the only person exposed to the late publication was the plaintiff’s agent.30 American courts have not followed the Duke of Brunswick in considering delivery to the plaintiff’s agent a new publication.31 Yet, they generally endorsed the multiple publication rule in the late-nineteenth century.32 This rule not only enables plaintiffs to overcome limitation periods, but also allows numerous lawsuits to arise from a single defamatory statement published multiple times.33

The technological advances of the twentieth century, particularly the emergence of mass communication, have raised several concerns. First, the multiple publication rule could open the floodgates of litigation, burdening the courts.34 Second, it might unduly harass the defendant by repeated suits.35 Third, numerous suits might result in excessive damages.36 Fourth, allowing claims following belated deliveries might undermine statutes of limitation.37 Fifth, a multiple publication rule, particularly in a federal system, might generate significant administrative costs if the publication is delivered in several places, due to the scattering of claims among different jurisdictions,38 and the potential diversity in applicable substantive law.39

Consequently, American courts started to deviate from the English case law, and developed an exception known as the “single publication rule.”40 Relying on accumulated case law, section 577A(3) of the Second Restatement of Torts provides that “[a]ny one edition of a book or a newspaper, or any one radio or television broadcast, exhibition of a motion picture or similar aggregate communication is a single publication.”41 Put differently, a single defamatory communication is considered a single publication, even when published to many persons.42 A defamation victim has only one cause of action for the mass or aggregate publication of a single defamatory statement.43 Most states follow the single publication rule, although a few remain undecided.44 In 2013, the British Parliament adopted a somewhat constrained version of the rule, providing that when a defamatory statement is published to the public, any cause of action for subsequent publication of the same statement is regarded “as having accrued on the date of the first publication.”45

The Uniform Single Publication Act set forth a more general version of the rule, applicable to all publication-based torts: “No person shall have more than one cause of action for damages for libel or slander or invasion of privacy or any other tort founded upon any single publication or exhibition or utterance . . . .”46 This rule covers one edition of a newspaper, a book or a magazine, one presentation to an audience, a single broadcast over radio or television, and one exhibition of a motion picture.47 It was endorsed by several state legislatures, including those of Arizona, California, Illinois, and Pennsylvania,48 and by the Texas Court of Appeals.49

The single publication rule has two important qualifications. First, it only applies to aggregate or mass communications, such as a newspaper issue or a television broadcast.50 If the publication is not made through mass communication, each subsequent delivery is considered a separate and distinct publication.51 Second, the rule only applies to the numerous deliveries of a single communication. If an additional delivery is not part of a single broadcast, edition or exhibition, a new publication occurs and gives rise to a new cause of action.52 In other words, each new edition or broadcast is a new publication. The justification is that “the second publication is intended to and does reach a new group.”53 Thus, the single publication rule would not apply when the subsequent publication is intended for a different audience.54 In contrast, it would apply where additional deliveries of a publication, such as printing of extra copies of the first edition of a book, are necessary to meet demand by the same audience after the original supply is exhausted.55

2.     Online Publications

Online publication has changed the way information is published and disseminated. To begin with, online publications are not attached to physical objects, such as books, and may be easily accessible by everyone in a vast geographical area for a long, potentially infinite, period of time. Moreover, Web 2.056 enables people to connect to the internet and contribute content with a potentially wide reach. Users can (1) ask questions or provide answers at online forums; (2) write blogs or make comments on others’ blog posts; (3) publish customer reviews of travel-related services at or of books at; (4) take part in multi-user discussions on social networking services, such as Facebook, Twitter or LinkedIn; (5) share photos and videos or make comments on others’ visual content on Instagram or YouTube; and (6) participate in collaborative writing projects, such as Wikipedia.57 Finally, and most importantly, technology enables users to readily and effortlessly republish existing content through actions such as “sharing,” “retweeting,” providing URLs, and forwarding e-mails. Therefore, while expanding the audience of traditional mass communication usually entails a new edition or broadcast, the audience of online communication can organically expand without any intervention of the content’s originator. These differences appear to strengthen the justifications for a single publication rule.

While application of the rule to online communications was proposed in the late 1990s,58 the legal authority crystallized only at the turn of the millennium. In the seminal case of Firth v. State, the defendant provided an executive summary of a report which was highly critical of the plaintiff’s conduct, with links to the full text, on the government’s website.59 The plaintiff brought an action more than a year following the publication, and the defendant argued “that the claim was time-barred [by] the one-year statute of limitations for defamation.”60 The plaintiff argued, first, “that the ongoing availability of the report via the Internet constituted a continuing wrong or new publication” (with every viewing) and, second, that the defendant subsequently modified the website, and that modification constitutes new publication.61 The New York Court of Appeals held that the justifications for the single publication rule apply a fortiori to online publications.62 Such publications have a much broader reach in terms of accessibility, time, and space, so “a multiple publication rule would implicate an even greater potential for endless retriggering of the statute of limitations, multiplicity of suits and harassment of defendants.”63 Furthermore, allowing a claim for each viewing might inhibit “the open, pervasive dissemination of information and ideas over the [i]nternet.”64 As regards the second argument, the Court of Appeals held that the “addition of unrelated information to a Web site” is not equivalent to the repetition of a defamatory statement in a new edition of a book or a newspaper.65

Following Firth, other courts have applied the single publication rule to online publications.66 However, courts have recognized two qualifications. First, republication occurs and a new cause of action arises when the content has been modified by the publisher after initial publication—something that could not normally happen in traditional mass media. This was explicitly stated by the Court of Claims and the Appellate Division in Firth,67 implicitly endorsed by the Court of Appeals in the same case,68 and followed in other jurisdictions.69 Second, as in the case of offline publication, if the same content is knowingly disseminated to a new audience, the additional dissemination is considered a new publication.70 Arguably, when a publisher solicits republication by others, the consequent republication may be deemed a new publication by the same publisher,71 because he or she knowingly made the content available to a new audience.72 Similarly, if the website starts a new advertising campaign, this may be deemed a conscious decision to attract a new audience.73

B.     Republication of Third-Party Content

1.     Offline Publications

Let us now turn to republication of third-party content. At common law, if a person republishes a defamatory statement already published by another, a new publication occurs, and a new cause of action against that person arises, in addition to any cause of action against the statement’s originator.74 Ascribing the statement to the originator does not exempt the republisher from liability.75 For example, when a news agency provides a defamatory article to newspapers, each paper that prints the article is liable.76 This rule is often defended on the theory that “[t]alebearers are as bad as talemakers.”77 Republication “threatens the [victim’s] reputation as much as . . . the original publication.”78

However, courts have recognized several limited exceptions to this rule. The first exception is the “fair report privilege.”79 Repetition of a defamatory statement may be privileged if it is part of an accurate and complete or fairly abridged report of official actions or proceedings.80 The fair report privilege is not absolute. It does not apply when the publisher, “with actual malice, . . . adopts the defamatory statement as its own,” instead of merely reporting that it was made.81 The rationale for this privilege is that the public has an interest in having access to information detailing what occurs in official proceedings and public meetings, in order to oversee such events.82 Moreover, the publisher here acts as an agent of the public, reporting things that people could hear for themselves if they attended the proceedings.83

A second exception is the “neutral reportage privilege.”84 In some jurisdictions, repetition of a defamatory statement is privileged if it is an (1) accurate (in the publisher’s reasonable good faith judgment), (2) disinterested (neutral), (3) report of serious charges, (4) against a public figure, (5) by “a responsible, prominent organization.”85 The exception applies regardless of the publisher’s views about the validity of the allegations.86 The rationale for this exception is that charges made by reputable organizations against public figures are “newsworthy”: The press must have the freedom to report such charges due to the public interest in being informed about controversies around sensitive issues.87 Note, however, that this exception has not yet been accepted by the Supreme Court,88 and was rejected in many jurisdictions.89

A third exception is the “wire service defense.”90 Republishing news does not give rise to liability if the news was previously “published by a recognizable reliable source of daily news,” such as a news agency (Associated Press, Reuters); appeared to be accurate (because nothing on the face of the release led to suspicion of inaccuracy); and was republished without substantial changes or actual knowledge of their falsity.91 The rationale is that modern publishers cannot afford to verify the authenticity of each and every news item they receive from a reliable source of news, and promptly disseminate newsworthy material.92 A verify-or-litigate choice would overburden news publishers, particularly small and local businesses, and curtail the flow of information.93

2.     Online Publications: The American No-Liability Model

In the absence of legislative intervention, the common law rule of liability for republication, along with its recognized exceptions, could extend to online publications. However, problems arising from the interaction between old law and new technology in a closely related context resulted in a fundamental reform, which turned the law of online republication 180 degrees.94

i.     The Failure of the Common Law and the Introduction of § 230

Traditionally, non-speaker liability for defamation has been subject to a distinction between publishers, distributors, and common carriers.95 This distinction is based on the level of the defendant’s control over the published content. “Publishers” of content created by others, such as newspapers or broadcast organizations, exercise significant control over published material and are subject to strict liability because they adopt the published material as their own.96 In contrast, “distributors” of content created by others, such as newsstands, bookstores, and libraries, distribute content without control over it and are liable only if they knew or had reason to know that the relevant material consisted of defamatory statements.97 Finally, “common carriers” or “conduits,” such as telephone companies, only transmit information and are not liable for defamation in transmitted material.98 However, this framework resulted in problematic rulings when courts began to apply it to online publications in the 1990s.99

In Cubby, Inc. v. CompuServe, Inc.,100 the court found that CompuServe, which provided users with online access to a daily newsletter without reviewing its content, was a mere distributor and therefore not liable for false and defamatory statements made in that newsletter.101 Yet in Stratton Oakmont, Inc. v. Prodigy Services Co.,102 the court held that Prodigy, a bulletin board operator that exercised editorial control over user-generated content, was a publisher, and thus could be held liable for defamatory statements made by an anonymous user concerning a securities brokerage firm.103 The joint reading of Cubby and Stratton Oakmont incentivized online platforms to avoid moderating online discourse, because moderating content exposed them to potential liability.104

Pressures from the internet industry105 quickly led to the enactment of § 230 of the CDA,106 whereby providers and users of interactive computer services shall not be considered publishers “of any information provided by another information content provider.”107 In Zeran v. America Online, Inc., the court held that under § 230 a message board operator could not be found liable for defamatory postings by an anonymous user, even though the operator had relevant knowledge after a certain point and would have been considered a publisher under traditional defamation law.108 Following Zeran, § 230 has afforded online service providers effective immunity from liability for the publication of third-party content.109

The immunity offered by § 230 has been broadly interpreted.110 First, the common view, following Zeran itself, is that the statutory immunity forecloses not only publishers’ liability, in the traditional sense, but also distributors’ fault-based liability.111 In other words, § 230 protects service providers with complete editorial control prior to publication, as well as service providers with no pre-publication control,112 and it applies even if the service provider knew the published statements were false or defamatory.113 Second, the immunity was granted to a wide array of online service providers, including online bulletin board services, chat room hosts, electronic mailing list operators, interactive dating websites, and even the feedback forum of an internet auction website.114 Third, the immunity applies to a very broad range of publication-based causes of action, such as defamation, infringement of privacy, and intentional infliction of emotional distress.115 Lastly, several courts extended the immunity even further, holding that it prohibits not only civil liability but also declaratory and injunctive relief, such as removal orders against service providers with respect to content created by third parties.116

ii.     Section 230 and User Liability

Less debated is the fact that § 230 also applies to users of interactive computer services, even though it was primarily intended to address the skewed incentive structure of online service providers under traditional law. The court in Batzel v. Smith117 was the first to use this feature. The administrator of a website and an electronic newsletter dedicated to stolen art posted on his website, and included in the newsletter, an e-mail sent by a third party that attributed possession of stolen art to the plaintiff.118 The classification of the website and the newsletter as “interactive computer services” was contested.119 The court held that it was unnecessary to decide whether the defendant provided an interactive computer service, because the defendant definitely used interactive computer services to operate the website and the newsletter, and “the language of [§ 230] confers immunity not just on ‘providers’ of [interactive computer] services, but also on ‘users.’”120 In this case, however, the person invoking § 230 was actually a service provider, that is, an administrator of a website and a newsletter, and a “user” only in the sense that he relied on other internet services to provide his own­ services. In other words, he was not a “pure user” of internet services.121

The applicability of § 230 to “pure users,” those who merely use internet services and do not provide them, was first discussed in the landmark case of Barrett v. Rosenthal.122 The defendant posted an article e-mailed to her by a third party on the websites of two newsgroups operated by others.123 The article contained allegedly defamatory statements about the plaintiffs.124 The Supreme Court of California first rejected the plaintiffs’ argument that § 230 does not foreclose distributor liability.125 Next, it held that although the legislative record does not indicate why § 230 applies to users,126 the statutory language clearly demonstrates that Congress did not intend to treat users and service providers differently.127 Lastly, the court rejected the plaintiffs’ argument that § 230 affords immunity to “passive users,” who merely receive offensive information or screen and remove it, but not to “active users,” who actively republish offensive information.128 The court concluded that “Congress has comprehensively immunized republication by individual internet users,” and that any “expansion of liability must await [c]ongressional action.”129

Based on Barrett, it appears individual users and service providers are equally immunized against liability for publishing content generated by third parties.130 Specifically, actions like “sharing” another person’s post on Facebook, “liking” a post to the extent that it brings the content to others’ attention, “retweeting” on Twitter, and “reblogging” on Tumblr, are mere dissemination of third-party content, and cannot give rise to liability.131 Forwarding an email with offensive content similarly enjoys § 230 immunity,132 as does hyperlinking to unlawful content, even if accompanied by explicit and enthusiastic endorsement of that content.133

The question is whether the Barrett ruling is consistent not only with the language of § 230, as explained above, but also with the legislative intent. Section 230 had two main purposes—to prevent the chilling effect of liability on the freedom of speech and the dissemination of information and ideas on the internet, and to encourage self-regulation of user-generated content by online platforms.134 Immunity for service providers is likely consistent with the legislative intent of § 230. To promote self-regulation, § 230 not only overcomes, through immunity, the Cubby–Stratton Oakmont incentive to avoid editorial control.135 It also immunizes providers (and users) from liability for good-faith removal of materials they consider offensive.136 The two purposes also underlie the judicial inclination to apply § 230 even if the defendant knew or had reason to know of the offensive nature of the content published. A knowledge-based standard would require service providers to investigate every notification of potentially offensive content, and then make a legal judgment and an editorial decision.137 Due to the complexities of applicable law and the extent of published content, service providers would have to choose between bearing an impossible economic burden and simply removing any material upon notification of its potentially offensive nature.138 This would have a chilling effect on online speech. Those displeased with online content would have a very simple way of censoring it (that is, by notifying the service provider that the content is “offensive”).139 Moreover, knowledge-based liability would deter service providers from self-regulation, because discovering suspected material would increase the chances of liability.140

Does the legislative intent also support absolute immunity to users? The goal of fostering free speech on the internet clearly supports such an extension. Active users “provide much of the ‘diversity of political discourse,’ the pursuit of ‘opportunities for cultural development,’ and the exploration of ‘myriad avenues for intellectual activity’ that the statute was meant to protect.”141 Users may not be threatened with liability as often as service providers, because they do not publish other users’ content as frequently, but “their lack of comparable financial and legal resources makes that threat no less intimidating.”142 In theory, users can escape liability by avoiding republication of unlawful material. Unfortunately, individual users have neither the expertise nor the resources to make the necessary judgments on a rolling basis. They cannot evaluate every item generated “by a third party . . . for veracity and for tendency to cause reputational harm” before republishing it,143 and might cope with the uncertainties by avoiding republication altogether. Therefore, a chilling effect is a real risk.144 The goal of encouraging self-regulation also seems to support user immunity. Indeed, individual users do not need to handle “the massive volume of third-party postings that [service] providers encounter,” so self-regulation appears “far less challenging” for them.145 But, again, users are more sensitive to the risk of liability, and less equipped than service providers to evaluate content and make accurate judgments about it. To the extent that exercising editorial discretion exposes them to liability, they will simply avoid it or, even worse, refrain from any republication.

Still, individual users differ from service providers in a very important respect. They are more likely to actively engage in the development and propagation of offensive material.146 Several qualifications to the user-immunity have been adopted or proposed to address this concern. First, “active involvement in the creation of a[n] [unlawful] [i]nternet posting,” even if based on another’s publication, should at a certain point turn the party involved into an original content provider, rather than a mere intermediary, and expose that party to liability.147 Many courts deny the immunity when the defendant “materially contributed” to the unlawful content.148 In Barrett, the line was not crossed by the defendant “[b]ecause [she] made no changes in the article she republished on the newsgroups.”149 Second, in a concurring opinion in Barrett, Justice Moreno opined that the immunity should not apply if the originator and the republisher conspired to publish offensive material, as otherwise, conspirators might get off scot-free.150 Consider a conspiracy between an “anonymous” or a judgment-proof provider of content, and another user who republishes that content (and currently enjoys immunity under § 230).151 Third, Danielle Citron contended that users should not be immune from liability if they republish offensive material knowing that it was illegally obtained152 (as in the Jennifer Lawrence incident mentioned in Part I).153

Critics of user immunity argue that it is unfair to impose liability for offline republication of third-party content,154 while affording absolute immunity for online republication.155 Supporters of the immunity respond that online republication is so different from offline republication that it justifies a different regime. They suggest three main distinguishing factors. First, online user republication is more easily deterred by expected litigation and liability than offline republication. Arguably, offline republication is dominated by “a relatively small number of powerful” entities—newspapers, magazines, broadcasters.156 These publishers may have in-house counsel, a deep pocket, and legal expenses and liability insurance.157 Online republication is carried out by millions of average citizens.158 The average citizen “is more likely to be dissuaded from [speech]” by the prospect of liability159 and by the costs of potential litigation.160 This is a relatively weak justification for the offline-online divide because the laws applicable to offline and online republication do not distinguish between republication by powerful entities and average citizens.

Second, supporters of user immunity argue that offline republishers are treated like the originators because they consciously adopt the statements as their own, usually following factual examination and legal analysis.161 Online users are not as equipped as offline republishers to determine whether the content they wish to republish is unlawful. Traditional offline republishers have time to review content between its creation and actual publication, whereas online republishers respond immediately.162 Thus, requiring online republishers to exercise close scrutiny would curtail online dialogue. Furthermore, offline republishers often have experienced staff members who carry out the review, whereas the online republisher is usually a single decisionmaker.163 Lastly, the average online republisher has neither the resources and expertise to investigate, nor the legal training to make a decision based on the facts revealed.164 If the law expected online republishers to investigate and make professional decisions prior to republication, the entire medium would be devastated. However, online users’ inability to make informed judgments is contested;165 and at any rate, pre-CDA common law also applied to individual non-professional republishers.166

Third, proponents of an extensive immunity emphasize the unique value of online republication. “Once a single actor introduces a[n] [interesting] piece of information [or argument] into the online realm, [others immediately] discuss and disseminate it.”167 Online republishing enables a more profound, thorough, inclusive, and open discussion of factual statements, as well as competing views and ideas, and helps “ascertain the truth” and crystalize opinions.168

3.     Online Publications: Alternative Liability Models

The American legal regime is exceptional from a global perspective, not only with respect to service providers, but also with respect to individual republishers. Other Western jurisdictions employ a variety of more generous liability regimes. Three conceptual models seem to dominate online republishers’ liability: the endorsement model, the knowing expansion of exposure model, and the repetition model.169 A few introductory analytical comments are due. First, the three models usually represent different perceptions of what constitutes “publication” for the purpose of publication-based torts. Second, given the variance among jurisdictions employing each model, and frequently among courts within each jurisdiction, the conceptual division does not neatly correlate with case outcomes. Courts applying the same model may reach different outcomes in similar cases, and courts applying different models may reach the same outcomes in similar cases. Third, the conceptual distinction generally crosses the common law–civil law divide. None of the three is a uniquely common law model or a uniquely civil law model. At this moment, however, the endorsement model is relatively widespread,170 whereas the other two have been adopted primarily in common law jurisdictions.171

i.     The Endorsement Model

Under the endorsement model, republication may give rise to liability only if the republisher expressed some form of endorsement of the offensive content, making the publication his or her own.172 Different jurisdictions applying this model have reached patently different results. In Germany, for example, liability for republication is considered under the general principles of civil liability, primarily those embodied in § 823 of the Civil Code (BGB).173 Two Higher State Courts held that “sharing” another person’s post on Facebook is not in itself a new publication that can give rise to a cause of action.174 The sharing user can be liable only if he or she adopted the shared content by adding a positive comment or expressing identification with it.175 For example, when a user shared a post and added that it was “too worthy of consideration to be withheld” (“zu erwägenswert, um ihn zu unterschlagen”), the court found that the comment constituted identification with the content and imposed liability.176 Swiss law may more readily impose liability under the same model. In 2017, the Zurich District Court convicted a Facebook user for “liking” comments made by others, which accused an animal rights activist of racism and antisemitism.177 The court held that “[b]y clicking the like button, ‘the defendant clearly endorsed the unseemly content and made it his own’”; and “by liking the comments, the [user] . . . disseminated them to his . . . Facebook contacts, and ‘ . . . made them accessible to a large number of people.’”178 Thus, while German courts do not consider “sharing” per se as an endorsement, at least one Swiss criminal court considered mere “liking” sufficient ground for liability. This demonstrates that using the same conceptual model in similar cases can yield dissimilar outcomes in different jurisdictions.

The endorsement model is not alien to common law jurisprudence. In a concurring opinion in Crookes v. Newton, which will be discussed in more detail below, Chief Justice McLachlin and Justice Fish of the Supreme Court of Canada agreed with the majority that a mere reference, including a hyperlink, is not enough to find publication.179 However, as opposed to the majority, which adopted the repetition model, the two opined that a hyperlink should constitute publication if, read contextually, the text that includes the hyperlink “indicates adoption or endorsement of the content of the hyperlinked text.”180 Presumably, this standard would mean that “sharing” an offensive posting is not in itself actionable, but it can become actionable with an expression of endorsement, even if modest. In Australia, the Supreme Court of New South Wales adopted a similar approach in Visscher v Maritime Union of Australia.181 The court, following Justices McLachlin and Fish, held that liability for hyperlinking can be imposed if “there was an approval, adoption, promotion or some form of ratification of the content of the hyperlinked material.”182 Hyperlinking to a defamatory newspaper article following a short introduction of the subject and an invitation to read the “full story” in the referenced piece constituted publication of the article by the hyperlinker.183 Further, in Bolton v Stoltenberg, the second defendant “liked” 64 Facebook posts written by the first defendant.184 The Supreme Court of New South Wales held “that clicking the ‘like’ button on a Facebook page [does not] constitute[] a level of endorsement of the publication to render the person liable as a publisher.”185 Thus, what constitutes “endorsement” under the endorsement model may be disputed.

ii.     The Expansion of Exposure Model

Under the expansion of exposure model, republication may give rise to liability only if the republisher knowingly expanded the content’s audience.186 This model was recently adopted by the Israeli Supreme Court in Shaul v. Nidaily Communication Ltd.187 The plaintiff was the publisher of a weekly local newspaper.188 The defendants “shared” a third party’s Facebook post, which insulted the newspaper and called for boycotting it, and “liked” another post containing similar statements.189 The question was whether “sharing” and “liking” Facebook posts constitute new publications and may underlie direct liability for defamation.190 The court of first instance (“Tel Aviv Magistrate Court”) answered in the negative and rejected the claim.191 The plaintiff appealed, and given the importance of this matter, the Court of Appeals (“Tel Aviv District Court”) asked the Attorney General to provide an impartial opinion.192

The Attorney General proposed a distinction between “liking” and “sharing” on social media.193 “Liking” may result in exposure of the content to additional users—those connected to the user who “liked” the content but not connected to the originator.194 However, additional exposure is uncertain.195 The user can neither predict nor control such exposure, which depends on social media algorithms and different variables unknown to average users.196 Moreover, from the user’s perspective, liking (especially given the forgone opportunity to share the same content) is not an act intended to disseminate content but an expression of emotion that should not be legally constrained.197 Dissemination, if it even occurs, is an unintended and uncontrolled by-product.198 Liking, therefore, is not a new publication for the purpose of publication-based torts.199 In contrast, “sharing” on Facebook, “retweeting” on Twitter, “reblogging” on Tumblr, forwarding an e-mail to a group of people, or linking to content on a different platform, are intentional acts of dissemination, which necessarily expose the content to additional users.200 Such acts may constitute new publications, but the fact that the offensive statement has already been published by another is taken into account in the assessment of damages.201

 The Court of Appeals wholly endorsed this distinction, along with its explanation, finding for the plaintiff with respect to the “sharing.”202 The defendants appealed to the Supreme Court, which followed the Court of Appeals in adopting the Attorney General’s opinion.203 The court emphasized that “sharing,” as opposed to “liking,” actively expands exposure.204 It held, however, that lawsuits for “sharing” offensive content are constrained by the prohibition against abuse of process,205 and by a liberal construction of the good faith publication defense.206 It added that “commenting” on an offensive post is akin to “liking”: It might expand the post’s exposure, but it cannot be regarded as an independent publication of the same content because the expansion is unintended and unpredicted.207 Although the Supreme Court discussed only a limited question, namely whether “sharing” and “liking” are publications for the purpose of defamation law,208 the ruling lays the foundations for liability for most forms of online republication of offensive content. It leaves some doubts about less active forms of expansion, such as approving or failing to remove a “tag,” which is added to the published content by the originator but results in exposure of the content to the tagged person’s connections.209

The expansion of exposure model was also employed in at least one Australian case. In Bolton v. Stoltenberg, after holding that “liking” does not give rise to liability under the endorsement model, the Supreme Court of New South Wales added that “liking” cannot be actionable under an expansion of exposure model either.210 “Liking” a Facebook post is not tantamount to participation in publication by “drawing the attention of another to” the offensive content, as the annotated hyperlink in Visscher did.211 If “liking” placed the content in the Facebook feed of other people, it would “draw[] the attention of another to” the publication and constitute participation therein.212 But “[t]here was . . . no sufficiently compelling evidence . . . that . . . ‘liking’ . . . ha[s] that effect.”213 Therefore, what counts as republication under the expansion of exposure model seems relatively straightforward.

iii.     The Repetition Model

Finally, under the repetition model, republication may give rise to liability only if the republisher repeated the offensive content.214 This is not necessarily the most generous liability regime, because endorsement and expansion of exposure can occur without repetition. Annotated hyperlinks to external sources provide a good example: They endorse external sources and expand their exposure without repeating them. However, in the context of online reposting, such as Facebook “sharing” and Twitter “retweeting,” the repetition model seems to offer the most lenient test for liability. The Supreme Court of Canada adopted this model in Crookes v. Newton.215 The majority held that references, particularly hyperlinks, “communicate that something exists [(and point at its location)], but do not . . . [convey] its content.”216 A reference “require[s] some act on the part of . . . third part[ies] before [they] gain[] access to the content.”217 “Making reference to the existence []or location of content by hyperlink or otherwise . . . is not publication of th[e] content.”218 Only if the person providing the reference “presents content from the [referenced] material in a way” which constitutes repetition, can this person be regarded as a publisher.219 In other words, the test for liability is actual repetition of the content. Although Crookes applies to new forms of references, it seems reasonable to deduce that while hyperlinks (references without repetition) are not considered publications, references with repetition, such as Facebook “shares” and Twitter “retweets,” may be deemed actionable publications. The technical feature of referencing and including the original content makes a difference in law.

4.     Originator’s Liability

Most jurisdictions seem to agree that, irrespective of the republisher’s status, the content originator is liable for all foreseeable consequence of the publication, including those of foreseeable republication by others.220 Recall that in the United States, individual internet users are immune from liability for republication of unlawful content.221 The Supreme Court of California thus held in Barrett v. Rosenthal that the victim of republished online defamation “may only seek recovery from the original source of the statement.”222 Legal literature accedes that because the victim can sue neither the platform enabling the publication (such as Facebook or Twitter) nor those who used it to republish the content, he or she “[is] left with the publisher of the original [content] as the sole possible defendant.”223 Refusing to impose liability on the originator for the consequences of foreseeable online republication would leave the victim with very limited redress. Unsurprisingly, and regardless of the single publication rule, the victim of a wrongful publication can seek recovery from the original publisher for republication by third parties, as long as it “is a natural and probable consequence . . . or . . . reasonably foreseeable.”224

The originator’s liability for foreseeable republication is preserved even in jurisdictions allowing online republisher liability. For example, in the Canadian case of Pritchard v. Van Nes,225 the defendant published defamatory postings about the plaintiff on Facebook and, due to republication, they reached a much wider audience than the defendant’s contacts.226 The British Columbia Supreme Court held that while a person is generally only responsible for his or her own defamatory publications, not for their repetition by others, there are several exceptions. Most notably, the Court articulated that the “defendant may be liable if the repetition was the natural and probable result of his or her publication.”227 Given the nature of social media and the structure of the platform, the Court reasoned that anyone who posts content “must appreciate that some degree of dissemination at least, and possibly widespread dissemination, may follow.”228 Republication is a “natural and probable result” of posting unlawful content on social media.229 The Court concluded, therefore, that the defendant in the case at bar “is liable for all of the republication through Facebook”230 and other means (such as e-mail).231 In Shaul v. Nidaily Communication, Ltd., the Israeli Supreme Court endorsed this component of Pritchard.232

III.     The Economic Case Against Liability

Part II presented the competing legal regimes. This Part evaluates the alternatives from an economic perspective. It first explains that attributing liability for a constant indivisible harm to multiple injurers in alternative care settings might lead to dilution of liability that would impair the incentives to take reasonable precautions. It argues that neither alternative care nor constant harm is an indispensable precondition for the applicability of the dilution argument. This Part then discusses the possible dilution of liability and its impact on deterrence in the context of online republication. It starts with a simple model involving non-overlapping contact lists of average size and, subsequently, introduces overlaps and variance in size among user contact lists. Finally, this Part examines the considerable upsurge in administrative costs associated with liability for online republication, arguing that these costs cannot be justified in terms of deterrence.

A.     Dilution of Liability

1.     The Basic Argument

“Efficient deterrence [requires] internalization by the wrongdoer of the social harm caused by . . . [the] wrongful conduct.”233 Potential injurers internalize the externalized costs of their conduct ex ante, and are induced to take cost-effective precautions, “[o]nly if the expected liability is equivalent to (or greater than234) the expected externalized costs.”235 For example, if Jack can prevent a $100 loss to Jill by taking precautions that cost $50, imposing liability for the loss caused by failing to take these precautions will induce Jack to take them, because he will save $100 at a cost of $50. Liability “can secure optimal deterrence only if: (1) wrongdoers are always liable for harms caused by their wrongful conduct; and (2)” the extent of liability is aligned “with the social cost of the wrongful conduct.”236 In the example above, if Jack has an 80% chance of evading liability, he will not take the necessary precautions because 20%×$100<$50. Similarly, if Jack will be liable for only 25% of Jill’s loss, he will not take the necessary precautions because 25%×$100<$50.

A multiplicity of defendants might, in some cases, undermine the deterrent effect of tort liability. Law-and-economics literature demonstrates that attributing liability for a constant indivisible harm to multiple injurers, where each could have single-handedly prevented that harm (“alternative care” settings), leads to dilution of liability.237 As the overall harm is constant, increasing the number of liable injurers reduces the burden incurred by each one, and impairs the incentives to take cost effective precautions. Consider the following example. D1, D2, and D3 can each prevent an expected harm of $1000 to P at a cost of $500. Taking precautions is socially desirable ($500<$1,000). But if all are liable for failing to take the necessary precautions, none will take them. Assuming the total harm will be equally allocated among the three wrongdoers, each expects liability of $333, so taking precautions at $500 is not worthwhile ($333<$500).238 Accordingly, no one would take precautions, even though the efficient course of action is that only one actor would. Assume now that D1 and D2 can prevent P’s harm at a cost of $500, and that D3 can prevent it at a cost of $400. In this case, the efficient course of action is that only D3 will take precautions. However, if liability is shared among all those who could have cost-effectively prevented the loss, no one will take precautions ($333<$500 and $333<$400). Thus, imposing liability on multiple injurers in cases of alternative care with indivisible harm might dilute liability and undermine its deterrent effect.

Arguably, the risk of dilution will be overcome if “one or a few salient” members of the larger injurer-group can be singled out and bear liability, while all others are exempted.239 If only one person is liable, he or she will have the incentive to invest in the necessary precautions. This person must be salient so that all potential injurers will know ex ante that he or she is the only one expected to take the necessary precautions. The salient injurer may be the cheapest cost avoider240 or the person “who has a ‘special relationship’ with the [potential] victim.”241 However, “sometimes what justifies the attribution of liability to one [person] [over] another is” merely the arbitrary fact that the former “can be clearly and unambiguously singled out from [all] others.”242

Scholars have supported several tort doctrines as preventing dilution of liability. First, the doctrine of comparative negligence243 might dilute liability, resulting in under-deterrence, where either party could prevent the harm at a cost lower than the expected harm but higher than that party’s relative share of the burden.244 For example, if D can prevent a $100 harm to P for $60, and P can prevent his own harm at the same cost, a comparative negligence regime imposing 50% of the resulting harm on each negligent party would induce both to avoid taking precaution (because $60>$50). A contributory negligence rule245 avoids this dilution because it imposes all of the burden on the negligent victim ex post, thereby inducing potential victims to invest in precautions ex ante.

Second, the doctrine of indemnity246 “shifts the . . . burden [of liability] from . . . joint tortfeasors to [the one] who is better situated to avoid the accident, rather than dividing it between [all] tortfeasors.”247 If D1 can prevent a $100 harm to P for $55, and D2 can prevent it for $60, apportioning liability equally between them will cause neither to take precautions (because $60>$50 and $55>$50).248 Ideally, the doctrine of indemnity will transfer the entire burden to D1 and generate an efficient incentive structure.249 However, this idea is not always simple to implement because there may be more than one cheapest cost avoider and, even when there is only one, potential injurers might not have sufficient information to identify that person ex ante and let only him or her take precautions.250

Third, the doctrine of no liability for mere omissions (nonfeasance)251 has also been defended in terms of dilution avoidance. Where many bystanders fail to take reasonable action to rescue a person in need, imposing liability on all ex post would dilute the burden and lead all to avoid the necessary action ex ante.252 For example, if the law imposed liability on 100 bystanders where each could single-handedly avert a $900 harm at a cost of $10, none would take care ex ante ($10>$900/100=$9); but if instead the law imposed liability only on one person, singled out based on a simple salient feature, that person would take care and the harm would be avoided ($900>$10).253 The law thus imposes liability for omission only on the salient wrongdoer, where one can be readily identified.254 The salient party can be the one who (1) voluntarily decided to act (and thereby singled out oneself),255 (2) has a special relationship with the victim (an easy-to-verify sorting device),256 (3) accidentally created a risk that now needs to be removed,257 or (4) is the cheapest cost avoider (the economist’s choice); but it can also be an arbitrary (though practical) criterion.258

Note that while singling-out the cheapest cost avoider is the ideal solution from an economic perspective, it is not always practical. First, where there is more than one cheapest cost avoider, the economic criterion cannot single out one person who will bear liability ex post and take the necessary precautions ex ante.259 Second, even when a single cheapest cost avoider really exists, his or her identity is not always obvious to all potential tortfeasors. Uncertainty might skew the incentives of all parties. If no one stands out ex ante, many might be (mistakenly) perceived as equally cheap cost avoiders, and expected liability will be diluted.260 Third, there may be a cheapest cost avoider that is difficult to identify and pursue ex post, so the victim might sue the entire group. Given the ex post uncertainty, all may be found liable and liability will once again be diluted. The real cheapest cost avoider has no incentive to clarify the situation as this will increase his or her liability.261

2.     Extensions of the Basic Argument

As mentioned above, the dilution of liability argument has been generally applied to settings satisfying three conditions: (1) a large number of potential tortfeasors; (2) the entire harm can be single-handedly prevented by each of the potential tortfeasors (“alternative care” settings); and (3) the harm does not increase with the number of tortfeasors (constant and indivisible harm).262 If each could prevent the harm, and is therefore liable, the constant harm is distributed among numerous injurers, liability is diluted, and its deterrent effect diminishes. Online republication scenarios often meet the first condition. Facebook posts, Twitter tweets, and e-mails, will usually attract attention when they are widely circulated by different users through actions like “sharing,” “retweeting,” copying URLs, and forwarding e-mails. The dynamic of “Web 2.0”263 in general, and social media in particular, is that of growing exposure through republication. However, online republication differs from previously analyzed phenomena. First, these are not alternative care situations, because no one but the content originator could do anything to prevent the entire harm. Most of the harm would still occur but for each republisher’s actions. Second, every republisher increases exposure and consequent harm, so the harm is neither constant nor indivisible. Do these differences avert dilution and the related concerns?

To begin with, the dilution argument should not be limited to alternative care situations.264 Consider, for example, several arsonists, each holding a torch, who simultaneously set the victim’s car on fire. This is clearly not an alternative care situation, because none of the injurers could have single-handedly prevented the harm.265 Nonetheless, “[t]he expected liability of each arsonist decreases as the number of [arsonists] increases.”266 Assuming the value of the car is $100, if there are two arsonists, “the expected liability of each . . . would be $50”; if there are three, the expected liability of each would decline to $33.33.267 Liability is diluted. Now assume that each of the arsonists was willing and able to pay $60 to inflict the harm. This is an illicit gain that should not be taken into account in the cost-benefit analysis.268 Otherwise, group arson will be deemed “efficient,” because $60×2>$100. Each would be deterred if acting alone ($60<$100), but not as one of two ($60>$100/2), three ($60>$100/3) or more. Dilution of liability undermines deterrence as it does in the alternative care case.269 More generally, then, the dilution argument may be applicable not only in alternative care settings, where each tortfeasor’s conduct is a necessary cause, but also “where each tortfeasor’s conduct is an independent and sufficient [but non-necessary] cause of the [same] injury.”270

In theory, the dilution argument may apply even where each tortfeasor’s conduct is neither necessary nor independently sufficient. This can happen, for example, where there are X concurrent tortfeasors, and harm would occur if any Y of them were negligent, assuming X>Y>1.271 In such a case, the negligence of each injurer is not a necessary cause, because the harm would occur even if that person was not negligent, as long as Y others were. Here, as opposed to the arsonists’ case, the negligence of any potential injurer is not a sufficient cause either, because the harm could not occur unless additional parties were negligent. Imposing joint-and-several liability on all negligent parties, as many courts would,272 might dilute liability and undermine its deterrent effect.

Admittedly, the case of independent and sufficient causes of the same injury and the case of combined non-necessary and insufficient causes are relatively rare. More importantly, neither reflects the online republication dynamic. But recognizing the possibility of extending the dilution argument beyond alternative care cases is an important step. To the extent that alternative care, a feature emphasized in prior liability-dilution literature, is not a theoretical necessity, the fact that online republication is not an alternative care situation should not discourage further analysis of the possibility and effect of liability dilution in this context.

Next, alternative care cases and all extensions discussed above share a common feature: The harm allocated among the injurers is constant and indivisible. The fact that more injurers were at fault and might be liable does not change the scope of the resulting harm. So far, therefore, dilution arguments were based on the very simple arithmetic idea that assuming c is a constant, c/x is smaller if x is larger; and if c/x is sufficiently small, no potential injurer will take the necessary precautions.273 The online republication dynamic lacks this characteristic, because republication increases exposure, and therefore increases aggregate harm. Applying the dilution argument to online republication requires further extension to cases of increasing harm.

Theoretically, liability might be diluted even if harm increases with the number of tortfeasors, as long as each additional tortfeasor causes a smaller increase in the aggregate indivisible harm. Assume, for example, that D1’s conduct exposes P to a 10% probability of a $10,000 harm, and that D1 can eliminate the risk by taking precautions for $800. Failing to take precautions is negligent, and if D1 can anticipate liability for the full harm ex post, the expected liability will incentivize him to take cost-effective precautions ex ante ($1,000>$800). Now assume D2’s negligence increases the expected harm to $1,300, D3’s to $1,500, and D4’s to $1,600. If the harm is divisible so that portions of $1,000, $300, $200, and $100 can be specifically attributed to the four injurers, liability will not be inefficiently diluted. Each will be liable for his or her exact contribution, and will be incentivized to take cost-effective precautions. However, if the harm is indivisible, so that the wrongdoing of each tortfeasor combines with the wrongdoing of others to generate greater harm, liability for the aggregate harm will be joint-and-several, and D1’s liability will be diluted. In this example, D1 will pay $1,000 if he alone is negligent, but his liability will decrease with the addition of D2’s, D3’s, and D4’s wrongdoing to $650 ($1,300/2), $500 ($1,500/3), and $400 ($1,600/4), respectively.274 Any additional wrongdoer dilutes D1’s liability to a point of inefficient deterrence because the cost of taking precautions exceeds his expected liability for failing to take them ($800>$650, $500, $400).

B.     Adaptation to Online Republication

1.     Average-Reach No-Overlap Republications

The dynamic of online republication differs from any of the previously discussed scenarios. It is certainly not an alternative care setting because no potential defendant, except for the content originator, can single-handedly prevent the entire harm. Furthermore, no individual conduct is a sufficient cause of the entire harm. In fact, none—except for the originator’s conduct—is even a sufficient cause of any part of the harm, because no harm caused by a single republisher could have occurred without the original publication and any prior republication that brought the content to his or her attention. But the main hurdle for the dilution argument is the fact that the scope of the harm increases with the number of tortfeasors, and is at least partly divisible. This entails close examination of the impact of the multiplicity of injurers on an individual injurer’s liability and its effect on deterrence.

The simplest possible model involves expansion of exposure by average users with no overlap between users’ contact lists. In the first “tier,” the originator publishes the content to c contacts. In the second tier, r×c of the recently exposed users (0<r<1) republish the content, each to additional c contacts, increasing exposure by rc2; in the third tier, r2c2 republish the content, each to additional c contacts, increasing exposure by r2c3; and in tier n, the content reaches additional rn−1cn contacts. Aggregate exposure in n tiers is ∑rx−1cx (x runs from 1 to n). Assume, for example, that c=100, and r=0.02 (2%). Two of the originator’s contacts (2%×100) republish the post, reaching 200 additional users. Four of those (2%×200) republish the republication, reaching other 400 users, and so on. Marginal and total exposures are depicted in Table 1 below. Now assume the direct harm caused by exposing the offensive content to a single user (excluding any additional harm caused if that individual republishes the content) is $1. The dollar value of the marginal harm for each publication tier is therefore equal to the marginal increase in the number of users exposed to the content, as shown in Table 1.

Table 1. Exposure and Harm

Finally, assume each publisher shares liability for each portion of the harm caused (in a “but-for” sense) by his or her publication with any republisher whose conduct was a necessary cause of the same harm. Thus, the originator will share liability for the harm caused by a first-order republication (second-tier publication) with the first-order republisher, for the harm caused by a second-order republication with both the second-order republisher and the first-order republisher who communicated it to the second, and so forth. The harm caused by an nth-order republication will be shared by the originator and the n republishers in the chain, each bearing 1/(n+1) thereof. The originator’s marginal liability will be the product of the number of nth-order republishers (2, 4, 8, 16, etc.) and the originator’s share of the liability for each nth-order republication (100/2, 100/3, 100/4, 100/5, etc.). Table 2 demonstrates the increase in the originator’s liability, which can be compared to the actual harm shown in Table 1.

Table 2. Originator’s Liability

The gap between total harm and total liability increases with the number of republications. This dilution of the originator’s liability might result in under-deterrence for one of two reasons. First, under-deterrence might occur where the originator can prevent the harm at a certain cost which is lower, potentially much lower, than the expected harm. Assume that the originator in the numerical example above can prevent the offensive publication at a cost of $1,400, which is the cost of the research required to reveal that the content is defamatory and false. The originator will be incentivized to take these precautions if he or she is liable for the harm caused by at least four publication tiers ($1,400<$1,500). However, if each publisher ends up sharing liability for each portion of the harm caused by his or her publication with every republisher whose conduct was a necessary cause of the same harm, the originator will internalize only a fraction of the harm caused. The originator will not be incentivized to take the necessary precautions unless he or she can predict at least seven publication tiers, because only then does the expected liability ($2,300) exceed the cost of avoidance ($1,400).

Second, under-deterrence might occur if the originator’s gain from the publication is dependent on aggregate exposure. The gain can be legitimate, such as where newspapers’ advertising revenues or journalists’ reputational gains are based on or related to exposure. It can also be illicit, such as where a person wishes to harm an adversary with false allegations and derives more satisfaction from greater impact. Assume that the originator in the numerical example above gains $0.5 from the viewing of the content by every single user. If each publisher ends up sharing liability for each portion of the harm caused by his or her publication with every republisher whose conduct was a necessary cause of the same portion, the originator will not be incentivized to avoid the publication if he or she can predict at least three tiers of publication ($0.5×700>$333).

Republishers’ deterrence is even more susceptible to dilution. To begin with, the farther republishers are removed from the originator in the publication chain, the lower their contribution to the overall exposure and harm. More importantly, the farther republishers are removed from the originator, the more diluted their liability is. A republisher R shares liability not only with those who republished R’s republication but also with all the previous links in the publication chain, which brought the content to R’s attention. The numerical example used above can illustrate these two observations. In regard to contribution to the aggregate harm, a first-order republisher is causally responsible for 50% of the aggregate harm excluding the $100 harm directly caused by the original publication. In regard to dilution, Table 3 shows that the ratio between the republisher’s liability and the harm he or she caused not only decreases with the number of subsequent republications (the dilution effect), but does so faster than the originator’s liability-to-harm ratio. The Table includes publication tier 1, for which the republisher is not responsible, to facilitate comparison with the originator.

Table 3. Republisher’s Liability

In the second publication tier, each of the first-order republications reaches 100 new users and causes harm of $100. Although this is the immediate and direct impact of the republication, liability is already diluted. A first-order republisher shares liability for this harm with the originator, $50 each. In the third publication tier, two of the republisher’s followers repost his or her posting, and each causes additional $100 harm. Liability for this harm is shared among the originator, the first-order republisher, and the second-order republisher, $33.3 each. Because there are two republishers of the republication, this increases the first-order republisher’s liability by $67, to $117, out of the $300 harm resulting from his or her republication. In the fourth publication tier, the harm caused by each third-order republication ($100) is shared among the originator and the three subsequent republication links, $25 each. The first-order republisher is responsible for four third-order republications, so his or her liability increases by $100, to $217, out of the $700 harm he or she caused. The fact that the first-order republisher’s liability is diluted more than the originator’s makes under-deterrence an even greater concern. It can be easily demonstrated that the nth-order republisher’s liability is diluted more than the (n−1)th-order republisher’s liability, exacerbating the under-deterrence problem.

In summary, republisher’s liability not only dilutes the originator’s liability and impairs its deterrent effect, but also fails to efficiently deter republishers. The exclusive originator liability rule secures efficient deterrence of originators by preventing dilution of their liability, ideally making republishers’ deterrence redundant. As in other contexts, errors in application might undermine the incentive structure. But the impact of such errors, when leading to republication of unlawful content, may be mitigated by the fact that republication under an exclusive originator liability regime, as a risk-free potentially impulsive activity, may be perceived as less reliable and generate less harm than under alternative regimes.

2.     Overlapping Audiences

In reality, when D1 publishes content to her contacts, and one of the contacts, D2, republishes the content to his contacts, there may be an overlap between the two contact lists. If there is no overlap, the analysis is the same as in Section III.B.1. If there is full overlap, republication does not increase exposure and aggregate harm. Anything in-between will inhibit, but not prevent, the expansion of exposure through republication. The greater the overlap, the smaller the marginal harm caused by republication. Presumably, more publication tiers entail increasing overlap between republishers’ and prior publishers’ contact lists.

Consider the following change in the numerical example. In the second publication tier, 10% of the users reached by the republishers are also the originator’s contacts, and have already been exposed to the content. Only 90% make a new audience, and 2% of them republish. In the third publication tier, only 80% of the republishers’ contacts are a new audience, of which 2% republish, in the fourth tier 70% are new, and so on, until the 11th tier, where republication can no longer reach new users. How does this affect the dilution of liability and its impact on deterrence? To begin with, overlaps reduce the marginal exposure and consequent harm per republication, as Table 4 demonstrates.

Table 4. Exposure/Harm with Overlaps

Table 5 shows the originator’s liability where the law imposes liability on republishers. Again, all those whose conduct was a necessary cause of a specific portion of the harm share liability for it. To the extent that expected liability is lower than the externalized social cost, under-deterrence might still occur. For instance, if the cost of precautions that could prevent the harm was $700, taking precautions would be desirable if more than three tiers of publication are foreseeable ($700<$971), but the diluted liability for harm caused by even seven tiers of publication ($619) would not induce the originator to take them ($700>$619).

Table 5. Originator’s Liability

Note, however, that when two or more republishers have the same person on their contact lists, social media algorithms may increase the salience and visibility of the content for that person. For example, it may increase the probability that the content will appear in that person’s “news feed.” In such a case, the fact that republishers’ contact lists overlap may reduce the marginal reach of each republication, but at the same time increase the likelihood that followers will actually notice the content and even republish it. This may undercut some of the “inhibitory” effect of overlaps on the expansion of exposure, and push it even closer to the no-overlap scenario.

3.     Substantial Reach

The originator and the republishers have audiences of different sizes, and these differences can sometimes be significant. For example, a newspaper like the New York Times may have millions of followers on Facebook275 and Twitter,276 whereas an average online reader may have hundreds. If the originator is a high-circulation newspaper, the initial impact of the publication may be so powerful that dilution of liability for republication is unlikely to have a significant effect on the originator’s inclination to take the necessary precautions. If the initial publication reaches a million people, directly causing considerable harm to a person’s reputation or privacy, liability for this harm will usually be sufficient to incentivize the newspaper to take precautions. Imposing liability for republication will dilute liability for some of the repercussions, but this alone will not justify preclusion of such liability. Theoretically, an exception to the rule of no liability for republication may be recognized where the originator is a highly influential content provider. However, in such cases, the rule of no liability may be justified for other reasons.277

If the originator is an average user, and one of the first republishers is a large-scale content provider, imposing liability for republication will dilute the originator’s liability but, again, not to an extent that might undermine the deterrent effect of liability. Consider, for example, a case in which the New York Times republishes content created by a freelance journalist with 100 followers. The newspaper’s republication is then reposted by readers, by their followers, etc. The New York Times will once again reach a million readers. Imposing liability for republication will leave the originator with a heavy burden (half of the harm caused by the New York Times’s republication)—a sufficient incentive to take the necessary precautions. Here too, however, an exception to the rule of no liability for republication may be justified. But this will be a somewhat different exception that will be discussed in Section IV.A.2 below.

C.     Administrative Costs

Allowing liability for republication has another potentially undesirable outcome: a considerable upsurge in administrative costs. If the average cost of litigating or settling a tort case against the publisher of offensive content is l, then the administrative cost of an exclusive-originator-liability rule will be l, whereas the administrative cost of liability-for-republication rule will be l+nl, where n is the number of republishers.278 The difference nl represents the price, in terms of administrative costs, of shifting from the former regime to the latter, and it may be considerable. To justify paying this price, the benefit must be larger.

As shown above, imposing liability for republication has very little, if any, benefit in terms of deterrence. First, an exclusive originator liability rule, which covers all foreseeable republication, secures full internalization by the originator and provides the proper incentive to take cost-effective precautions.279 Imposing liability on any other party is redundant, even in the absence of liability dilution. The administrative cost of such liability outweighs the negligible or nonexistent deterrent benefit. Second, a liability-for-republication rule might undermine originators’ incentive to take the necessary precautions due to the dilution of their liability for harms caused by republications. Additional claims, with their administrative costs, result in less efficient deterrence. Third, a liability-for-republication rule might not induce any efficient changes in the conduct of republishers, because of the more rapid dilution of their own liability.

Notably, the administrative costs argument supports an exclusive-originator-liability rule even where the liability dilution line of argument is inapplicable. As explained above, dilution of liability does not justify exclusion of liability for republication where the originator is a highly influential content provider. In such a case, the immediate impact of the publication is so significant that it provides a sufficient incentive to take the necessary precautions even if liability for subsequent republications is intensely diluted. However, because efficient deterrence is achieved through the originator’s liability, republisher liability is redundant in terms of deterrence, and wasteful in terms of administrative costs. No liability for republication is still economically justified.

Lastly, the administrative costs argument may be consistent with an exception to the no-liability for republication rule where the originator is an average user, and one of the first republishers is a large-scale content provider. Imposing liability for republication will dilute the originator’s liability, but not to an extent that might undermine its deterrent effect. However, the originator’s deterrence might be affected by the fact that he or she is an average user who is also judgment proof. Under these circumstances, imposing liability on the large-scale republisher may be necessary to ensure that the party who is responsible for most of the harm, and can actually bear the burden of liability, internalizes the costs, and takes the necessary precautions.280

IV.     The Limits of the Theoretical Argument

Part III presented the theoretical justification for the exclusive originator liability rule. This Part examines its limits. First, this Part maintains that economic theory requires recognition of three exceptions to the exclusive originator liability rule when (1) the originator of the content in unknown, (2) the originator is judgment-proof, or (3) the republication is somehow unforeseeable. Second, this Part examines two additional cases that may be considered quasi-exceptions to the exclusive originator liability rule—supplemented republication and republication of illegally obtained content. Finally, this Part challenges the underlying assumptions of the economic argument by examining whether dilution is indeed likely and undesirable.

A.     Necessary Exceptions

1.     Unidentifiable Originator

Exclusive originator liability raises a serious problem where the originator is anonymous or pseudonymous. The victim cannot sue an unidentifiable wrongdoer, and no action means no internalization and no deterrence. To identify an anonymous originator, “[t]he victim needs to obtain [his or her] Internet Protocol (IP) address from the [online platform used to publish the] content . . . and then obtain the [originator’s] identity from the [internet service provider (ISP)], as identified by the IP address.”281 Such a process puts the speaker’s freedom of speech and right to privacy at risk, so even if the law enables it,282 it is inevitably cautious, potentially complicated, and rather costly.283 In addition, users can conceal their IP addresses through various technological tools, such as Tor.284 Furthermore, identifying the wrongdoer’s IP address will not enable attribution of the wrongdoing to a specific person if the same address was used by others. This may occur when the wrongdoer uses a public hotspot285 or illegally connects to another person’s private network.286 The plaintiff will face an additional practical obstacle if the information necessary for the identification of the originator is not retained by the online platform used for the wrongdoing or by the originator’s ISP.287 Finally, the use of legal de-anonymization tools may have territorial boundaries, so anonymous users who publish offensive content on websites or through ISPs outside the court’s jurisdiction may evade identification and liability.288 Publication in a different country will probably pose an even greater problem for plaintiffs than publication in another state.289

Accordingly, “identifying an online anonymous [originator] might be very difficult.”290 When originators “[are] not identified, [they] evade[] liability, the costs of anonymous [wrongdoing] are not fully internalized, and potential wrongdoers are not efficiently deterred.”291 When originators

[are] identified through a costly process, [they] . . . internalize the costs of [the] wrongdoing, but the administrative costs may outweigh the benefits in terms of cost-reducing deterrence. Alternatively, the high administrative costs associated with identifying the primary wrongdoer might render another party (for example, [the online platform operator]) a more cost-effective target for enforcement efforts.292

In an online republication setting, the problem may have an easier solution. If the originator is unidentifiable, liability can be imposed on first-order republishers, each for his or her contribution to the aggregate harm. Put differently, a limited exception to the general rule of no liability for republication may be recognized where the originator is unidentifiable. There is some risk of under-deterrence, because the cost of universal precautions that can prevent the harm (such as the costs of verifying the authenticity of the content) may fall short of the aggregate harm caused by the publication but exceed the harm attributable to each of the first-order republishers. In the numerical example presented in Section III.B.2, with seven foreseeable publication tiers, exclusive originator liability can induce the originator to take precautions for up to $12,700 (Table 1), but first-order republishers’ liability will induce them to take precautions for up to $6,300 (Table 3). If precautions cost $8,000, the solution will not work. But, presumably, this will be rare, and first-order republisher liability will usually suffice to overcome the impracticability of imposing the full burden on the unidentified originator.

The proposed exception also addresses a real concern raised in a concurring opinion in Barrett.293 Justice Moreno envisioned a conspiracy whereby one person anonymously publishes offensive content, and another republishes it. The former cannot be traced, and the latter is immune under § 230.294 The proposed exception prevents this double impunity and frustrates the conspiracy by holding the republisher liable when the originator is unidentifiable. If the republisher argues that the exception does not apply, namely that the originator is identifiable, the former needs to disclose the latter’s identity, and the conspiracy once again fails. The proposed exception is also preferable to malice-based republisher liability as a solution to the potential conspiracy problem, because the administrative costs of proving that the originator is unidentifiable are much lower than those of establishing the republisher’s malice.

Of course, the exception can provide the necessary incentives only if the first-order republisher knows or has reason to know that the originator is unidentifiable. Otherwise, the republisher believes that the originator will exclusively bear the burden and is not deterred by the prospect of liability for republication. To avoid wasteful litigation that yields no deterrence, first-order republisher liability can hinge on the republisher’s actual or constructive knowledge of the originator’s unidentifiability. This limitation may nonetheless be unwarranted from an economic perspective given the presumably low probability of republishers’ misperception of unidentifiable originators as identifiable on the one hand, and the cost of litigating the question of the republisher’s knowledge on the other hand. In addition, the knowledge requirement might leave the victim with no legal redress where first-order republishers mistakenly believe that an anonymous originator is identifiable.

2.     Judgment-Proof Originator

An additional problem with exclusive originator liability is the possibility that the originator will be judgment-proof. If the only person liable for the harm caused by the publication is unable to fully compensate the victim, she will not internalize the social cost of that conduct. From her perspective, the expected expense may be considerably lower than the expected (social) harm. The incentive for choosing the optimal level of care is impaired.295 For example, assume there is a probability of 0.2 that A’s conduct will cause a $1,000 loss to B, and that A can reduce the probability of harm to 0.1 by adopting a certain precaution for $80 (for example, verifying the accuracy of a factual statement). The cost of care ($80) is lower than the ensuing reduction in expected harm ($100), so from an economic perspective, taking this precaution is desirable. Now assume that the expected value of A’s assets during the subsequent litigation is $300 and that A is risk-neutral. Even if liability is certain, it will not provide an adequate incentive for choosing the optimal level of care. The expected sanction that would be imposed on A for failing to take the optimal level of care would be only 0.2×$300=$60, whereas the cost of the precaution is $80 (and under a strict liability rule, this adds to expected liability of 0.1×$300=$30). Simply put, the wrongdoer’s ability to pay is a de facto cap on liability and internalization.296

Although the judgment-proof defendant is a general problem in tort law, it is particularly common in cases of online speech torts. Almost everyone in the developed world uses the internet.297 The ease of access and the veil of anonymity encourage everyone to participate. The typical user is essentially the average citizen with average assets and average income. Consequently, online speakers are often judgment-proof individuals.298 Typical internet users may not have sufficient assets to pay for the harms caused by their offensive statements.299 Thus, exclusive originator liability for online publication might result in under-deterrence. Obviously, a wrongdoer’s inability to fully pay for the harm caused results not only in under-deterrence, but also in under-compensation. Exclusive originator liability might thus leave the victim with partly compensated harm.300 Although under-compensation is distinct from under-deterrence, and seems most relevant from a corrective justice perspective, it may have economic repercussions to the extent that satisfactory compensation is required to prevent deterioration of the victim’s condition.

“The proponents of the economic approach to tort law traditionally propose several solutions to this problem.”301 The most plausible is imposition of liability on another.

[I]f there is a third party with sufficient financial resources who has some control over the . . . wrongdoer’s conduct, that party may be held vicariously liable for [that] conduct . . . or liable for negligently failing to prevent it. In many cases there is no such person or organization, and even where such a person exists, the degree of control is rarely sufficient to ensure the efficient conduct of judgment-proof actors.302

One alternative in the context of online publication is imposing liability on the platform allowing the publication (such as Facebook). Platforms have some control over user-generated content, and frequently have the resources to compensate defamation victims. This solution and its problems were discussed elsewhere.303 Another option in the context of online republication is allowing liability of first-order republishers, as an exception to the exclusive originator liability rule. This is an imperfect solution because sharing liability for the entire harm among the originator and first-order republishers still dilutes liability. However, the dilution of liability is limited, as subsequent republishers do not share in the burden. Imposing liability on this small group may provide sufficient incentives in many—though not all—cases.

3.     Unforeseeable Rerouting

At times, the initial publication of unlawful content is unlikely to cause real harm, but republication of the same content by a different person to a different group might be devastating. If the publisher can reasonably foresee the particular kind of republication, the additional exposure may be attributed to him or her, and the exclusive originator liability rule should apply. Yet the republisher’s identity and the republication’s audience, both affecting the extent of ultimate harm, might be unforeseeable. To begin with, the republisher might be an unpredictably reliable person. For example, D1 sent a hoax concerning P to a small group of friends. One of the addressees shared the text with D2, a prominent journalist, through a private message, and the latter republished the story on Twitter. The original publication was made by D1 and perceived by his friends as a hoax, and D1 did not foresee that it would be reported as genuine news. The initial publication caused little harm, if any, to P, but given D2’s professional prestige and reliability, the unforeseeable republication gave rise to considerable harm. Similarly, republication might reach a fundamentally different—unforeseeable—audience. For instance, D1 sent an e-mail complaining about a lawyer P to a few friends, one of whom forwarded the e-mail to D2, who reposted it in a large Facebook group of lawyers. Again, while foreseeable harm was limited, the republication generated highly harmful repercussions that the originator did not and could not reasonably expect.

From an economic perspective, the republisher should be liable for the unforeseeable rerouting. Such liability does not dilute the originator’s liability because the originator is not, and should not be, liable for unforeseeable ramifications of the original publication anyway. The republisher’s liability is necessary because in its absence, no one is liable for that harm ex post and incentivized to prevent it ex ante. Whether imposing liability for republication constitutes a real exception to the exclusive originator liability rule depends on the rule’s structure. Section 230 of the CDA, as interpreted by the courts, does not allow liability for republication at all.304 Recognizing republishers’ liability is a proposed exception to such a rule. A different version of the exclusive originator liability rule may hold that the originator’s liability excludes republishers’ liability only if the former is actually liable, namely, where the republication is foreseeable. If this version is adopted, liability for unforeseeable rerouting is not a real exception.

B.     Quasi-Exceptions

1.     Supplemented Republication

In the simple case of republication, the original content is republished as is by “sharing” the original post on Facebook, “retweeting” a tweet on Twitter, or forwarding a received e-mail. Yet in other cases, the republisher adds new and original content. The supplement can relate to the original content positively or negatively, with different levels of detail and passion. If the supplement is negative, in the sense that it rejects or criticizes the original content, it cannot be considered part of the harmful dissemination. Thus, it does not call for any exception to the exclusive originator liability rule. If the supplement is positive but very thin, it is an ordinary link in the publication chain.

In contrast, positive endorsement of published content with a substantial addition may increase the harm beyond any increase caused by sharing the original content. In such a case, the republication and the original publication should be analyzed separately. The republication is subject to the no-liability rule. The original addition, interpreted in light of the republished content, must be regarded as a new publication, giving rise to new originator liability, and if it is republished—as a new source of a republication chain.305 Although this distinction entails republishers’ liability, it does not constitute a real exception to the exclusive originator liability rule. The republisher’s liability is not for the act of republication, but for the act of publishing new unlawful content.

2.     Illegally Obtained Content

In the Jennifer Lawrence incident, the unlawful content circulated among millions was illegally obtained by hacking her phone.306 While the prospect of ex post dilution of liability might affect the originator’s ex ante decisions, republication with knowledge that the content was illegally obtained is an independent intentional wrong that needs to be addressed. Tort law can hold all republishers who knew about the illegality liable and counteract the dilution of liability through punitive damages. However, it may well be that criminal law, with its built-in constitutional constraints and prosecutorial discretion, is a more appropriate solution to this specific problem. Knowingly publishing illegally obtained content may be an offense under federal legislation, such as the Electronic Communications Privacy Act.307 It cannot constitute an offense if the content is of public concern and the publisher did not take part in the criminal activity,308 but it might be deemed criminal in the absence of public interest in the publication. Prosecutors can take into account factors like the nature of the content, the extent of each offender’s contribution to its dissemination, and the offender’s state of mind, in deciding whether to indict and offer a plea bargain.

C.     Challenging the Underlying Assumptions

1.     The Likelihood of Dilution

One of the main assumptions of the economic analysis is that without the exclusive originator liability rule the originator would share liability for the harm caused by republication with the republishers. Liability will not be diluted if the originator does not actually share it with others. In theory, even if the victim can sue republishers, he or she may choose not to do so, and focus solely on the originator. In such a case, the originator will internalize the full harm, and recognition of republisher liability will not have an adverse effect on the originator’s incentives. The response to this argument is twofold. First, while the victim may indeed choose whom to sue, it may often be in the plaintiff’s best interest to bring an action against as many defendants as possible to secure full compensation. Additionally, by suing several injurers the victim can gain a strategic advantage, because several defendants liable for the same harm will try to attribute greater parts of the harm to the others, thereby helping the victim establish fault, causation, etc. Second, even if the victim sues only the originator (or a small subgroup of the republishers), the defendant may spread the cost by issuing third party notices to other liable parties. True, a newspaper is less likely to implicate readers who chose to republish, as this might alienate them and affect their inclination to disseminate future content. But the paradigmatic republication chain involves average users, who may be less reluctant to share the burden. In fact, the possibility of sharing the burden may induce users to seek republication in the first place.

Even if neither the victim nor the originator pursues republishers ex post, the ex ante probability that republishers will be sued affects conduct. A sufficiently high probability of dilution may thwart the legal incentive. For example, if the expected aggregate harm is $1,000 and the cost of precautions that can eliminate the risk is $750, taking these precautions is desirable and should be encouraged. Now assume that 80% of the aggregate harm is attributable to republication, and that there is a 75% probability that this portion will be shared by the originator and at least one republisher. The originator’s expected liability for failing to take the necessary precautions will be $700: 20%×$1000=$200 (harm without republication) plus 80%×$1,000×(75%×1/2+25%×1)=$500 (republication harm borne by the originator). The originator will not have an incentive to take the necessary precautions ($750>$700). Average users do not usually engage in such accurate calculations. The numerical example merely illustrates why the ex ante possibility of sharing the burden has an impact on the originator’s conduct. The higher the probability of sharing the burden, and the larger the number of republishers sharing it, the greater the impact.

2.     The Benefits of Dilution

Critics may argue that dilution of liability in the republication context may have benefits that it does not have in other contexts. Specifically, exclusive originator liability makes first-publication more costly. In theory, this reduces the incentive to bring to light controversial content that may serve the public interest, either by helping people protect or promote important personal interests, or by facilitating public debate. Those possessing newsworthy material might withhold publication, at least until it is published by others, who will bear the risk of liability for the full harm. If all potential publishers are facing the same first-publisher liability threat, the material might not be published at all. Dilution of liability reduces first-publisher risk and may encourage first publication.

However, these benefits are questionable. To start with, dilution of liability may be beneficial only to the extent that the publication in hand needs to be encouraged. The law does not normally aim to encourage unlawful publications, so dilution of liability is inappropriate where the underlying publication is unlawful. If, on the other hand, the publication is lawful, it should not give rise to liability; liability is not a hindrance to publication, and dilution is unnecessary. Dilution may have some benefit if the potential publisher fears liability for non-wrongful publication, due primarily to judicial error, but the likelihood and impact of such error are probably too speculative to justify deviation from the exclusive originator liability rule.

Furthermore, if the content is newsworthy, the first publisher may gain substantial benefit from the publication. For example, publishing an exclusive may increase readership or viewership and advertising income. Additionally, if the content is so valuable that others wish to republish it, the originator (for instance, a freelance journalist) can charge for republication. Put differently, those in possession of newsworthy material will not withhold publication, because a sufficiently strong incentive to preempt others exists. As long as the publication is not wrongful, this incentive only needs to overcome the speculative fear of liability due to judicial error. On the other hand, if the publication is wrongful, there is nothing patently flawed in making the wrongdoer internalize its costs and weigh them against any potential benefit.

V.     Conclusion

This Article proposed an economic justification for the notorious online republication rule embodied in § 230 of the CDA, as interpreted in Batzel v. Smith309 and Barrett v. Rosenthal.310 Part II presented the legal framework. It discussed the rules applicable to republication of self-created content, focusing on the emergence of the single publication rule and its natural extension to online republication. It then turned to republication of third-party content. American law makes a clear-cut distinction between offline republication, which gives rise to a new cause of action against the republisher (subject to a few limited exceptions), and online republication, which enjoys an almost absolute immunity. Other Western jurisdictions employ more generous republisher liability regimes, which usually require endorsement, a knowing expansion of exposure, or repetition.

Part III evaluated the alternatives from an economic perspective. Law-and-economics literature showed that attributing liability for constant indivisible harm to multiple injurers, where each could have single-handedly prevented that harm (“alternative care” settings), leads to dilution of liability. Online republication scenarios often involve multiple tortfeasors. However, they differ from previously analyzed phenomena because they are not alternative care situations, and because the harm—increased by the conduct of each tortfeasor—is not constant and indivisible. Part III argued that neither feature precludes the dilution argument. It explained that the impact of the multiplicity of injurers in the online republication context on liability and deterrence provides a general justification for the American rule. This rule’s relatively low administrative costs afford additional support.

Part IV discussed the limits of the theoretical argument. It demonstrated that exceptions to the exclusive originator liability rule should be recognized when the originator is unidentifiable or judgment-proof, and when either the republisher’s identity or the republication’s audience was unforeseeable. It also explained that the rule does not preclude liability for positive endorsement with a substantial addition, which constitutes a new original publication, or for the dissemination of illegally obtained content, which is an independent wrong. Lastly, Part IV addressed possible challenges to the main argument’s underlying assumptions, namely that liability dilution is a real risk and that it is undesirable.

The Article leaves out interesting questions that do not fit its structure and method. For instance, the analytical part precludes the unique case in which the defendant republished content that the plaintiff created or owned, thereby infringing the latter’s copyright.311 Apart from being factually distinct from all cases discussed in this Article, copyright violations are not covered by the CDA immunity.312 Similarly, the normative part does not consider whether holding an originator exclusively liable for all harm caused by foreseeable republication is morally defensible. The Article aimed to provide an economic justification for the American exclusive originator liability rule and enrich the judicial and scholarly debate on the subject, leaving other relevant perspectives for future research. Hopefully, other Western jurisdictions with more generous republisher-liability regimes will take note of this analysis and reconsider their current law.

  1. [1]. Danielle Citron, Can and Should Perez Hilton be Held Liable for Reposting Celebrities’ Private Nude Photos Without Their Consent, Forbes (Sept. 3, 2014, 4:41 PM),
    sites/daniellecitron/2014/09/03/can-and-should-perez-hilton-be-held-liable-for-reposting-celebrities-private-nude-photos-without-their-consent [].

  2. [2]. Batzel v. Smith, 333 F.3d 1018, 1030–31 (9th Cir. 2003) (interpreting 47 U.S.C. § 230 as precluding liability of internet service users for online republication of content created by another).

  3. [3]. Barrett v. Rosenthal, 146 P.3d 510, 528–29 (Cal. 2006) (adopting Batzel and applying it to “users,” regardless of whether they are active or passive users).

  4. [4]. See infra Section II.B.3.

  5. [5]. See, e.g., John C.P. Goldberg & Benjamin C. Zipursky, Recognizing Wrongs 330–37 (2020) (criticizing the use of § 230(c)(1) to preclude liability for online republication); Daniel J. Solove, The Future of Reputation: Gossip, Rumor, and Privacy on the Internet 157–59 (2007) (criticizing the broad immunity afforded under § 230, using Batzel as an example); Benjamin C. Zipursky, The Monsanto Lecture: Online Defamation, Legal Concepts, and the Good Samaritan, 51 Val. U. L. Rev. 1, 40–52 (2016) (criticizing the preclusion of liability for republication). For a more general criticism of § 230 see, for example, Vanessa S. Browne-Barbour, Losing Their License to Libel: Revisiting § 230 Immunity, 30 Berkeley Tech. L.J. 1505, 1553–56 (2015) (proposing revision of § 230); and Danielle Keats Citron & Benjamin Wittes, The Internet Will Not Break: Denying Bad Samaritans § 230 Immunity, 86 Fordham L. Rev. 401, 415–19 (2017) (proposing a narrower reading or revision of § 230).

  6. [6]. This Article does not discuss online platform liability, but it has been thoroughly discussed elsewhere. See generally Ronen Perry & Tal Z. Zarsky, Who Should Be Liable for Online Anonymous Defamation?, 82 U. Chi. L. Rev. Dialogue 162 (2015) [hereinafter Perry & Zarsky, Who Should Be Liable?] (exploring the interrelation between user and platform liability); Ronen Perry & Tal Z. Zarsky, Liability for Online Anonymous Speech: Comparative and Economic Analyses, 5 J. Eur. Tort L. 205 (2014) [hereinafter Perry & Zarsky, Online Anonymous Speech] (examining comparative and economic models for handling the problem of anonymous online defamation).

  7. [7]. See infra Section II.A.1.

  8. [8]. See infra Section II.A.1.

  9. [9]. See infra notes 40–43 and accompanying text.

  10. [10]. See infra notes 58–66 and accompanying text.

  11. [11]. See infra note 74.

  12. [12]. See infra notes 79–80.

  13. [13]. 47 U.S.C. § 230 (2018).

  14. [14]. See infra notes 105–09, 117–33 and accompanying text.

  15. [15]. Infra Section II.B.3.

  16. [16]. Infra Section II.B.4.

  17. [17]. See infra note 233.

  18. [18]. See infra notes 243–58 and accompanying text.

  19. [19]. Infra Section III.A.2.

  20. [20]. Infra Section III.B.1.

  21. [21]. Infra Section III.B.2–.3.

  22. [22]. Infra Section III.C.

  23. [23]. Infra Section IV.B.1.

  24. [24]. Infra Section IV.B.2.

  25. [25]. Duke of Brunswick & Luneberg v. Harmer (1849) 117 Eng. Rep. 75, 75 (QB).

  26. [26]. Id.

  27. [27]. Id. at 76.

  28. [28]. Id.

  29. [29]. Id.

  30. [30]. Id. at 76–77 (“The defendant, who, on the application of a stranger, delivers to him the writing which libels a third person, publishes the libellous matter to him, though he may have been sent for the purpose of procuring the work by that third person.... [H]e lowers the reputation of the principal in the mind of the agent ....”).

  31. [31]. See Renfro Drug Co. v. Lawson, 160 S.W.2d 246, 251 (Tex. 1942) (denying recovery for defamation following sale of defamatory material to plaintiff’s attorney); Robert F. Somers, Slander? Prove It: Why a Two Hundred-Year-Old Defamation Law Should Be Changed, 19 Sw. J. Int’l L. 133, 151 (2012) (“[N]o American case has followed Brunswick regarding publication to a decoy agent.”).

  32. [32]. See, e.g., O’Reilly v. Curtis Publ’g Co., 31 F. Supp. 364, 365 (D. Mass. 1940) (holding that publication of the same defamatory content in several “states gives rise to separate [and distinct] causes of action”); Staub v. Van Benthuysen, 36 La. Ann. 467, 469 (1884) (“Every sale or delivery of a written or printed copy of a libel is a fresh publication, and every person who sells a written or printed copy of it may be sued therefor ....”).

  33. [33]. See, e.g., Harris v Perkins [2001] NSWSC 258, ¶ 25 (Austl.) (“[T]here is a separate publication (and thus a separate cause of action) in relation to each copy delivered to a reader .... If a newspaper circulates 100,000 copies of the one edition (defamatory of the plaintiff), he has available to him at least 100,000 causes of action.” (citation omitted)).

  34. [34]. See, e.g., Keeton v. Hustler Mag., Inc., 465 U.S. 770, 777 (1984) (“This rule reduces the potential serious drain of libel cases on judicial resources.”); Buckley v. N.Y. Post Corp., 373 F.2d 175, 179–80 (2d Cir. 1967) (holding that the single publication rule “protect[s] ... the courts ... from a multiplicity of suits”); Firth v. State, 775 N.E.2d 463, 465 (N.Y. 2002) (holding that the multiple publication rule leads to “a multiplicity of actions ... draining [] judicial resources”); Restatement (Second) of Torts § 577A cmt. c (Am. L. Inst. 1977) (“The rule is justified by the necessity of protecting defendants and the courts from numerous suits ....”); Sapna Kumar, Comment, Website Libel and the Single Publication Rule, 70 U. Chi. L. Rev. 639, 644 (2003) (same).

  35. [35]. See, e.g., Keeton, 465 U.S. at 777 (“It also serves to protect defendants from harassment resulting from multiple suits.”); Buckley, 373 F.2d at 179–80 (explaining that the single publication rule protects the defendant from a multiplicity of suits); Hickey v. St. Martin’s Press, Inc., 978 F. Supp. 230, 236 (D. Md. 1997) (“[T]he multiple publication rule allows a plaintiff to harass a defendant by bringing numerous causes of action against the same defendant ....”); Firth, 775 N.E.2d at 465 (explaining that the multiple publication rule might “lead[] to potential harassment”); Restatement (Second) of Torts § 577A cmt. d (same).

  36. [36]. See, e.g., Firth, 775 N.E.2d at 465 (referring to “excessive liability”).

  37. [37]. See, e.g., Buckley, 373 F.2d at 180 (explaining that the single publication rule prevents “endless tolling of the statute of limitations”); Hickey, 978 F. Supp. at 236 (same); Firth, 775 N.E.2d at 466 (same); Kumar, supra note 34, at 644 (same).

  38. [38]. See, e.g., Firth, 775 N.E.2d at 465 (“[T]he single publication rule actually reduces the possibility of hardship to plaintiffs by allowing the collection of all damages in one case commenced in a single jurisdiction.” (citation omitted)).

  39. [39]. See, e.g., Buckley, 373 F.2d at 180 (discussing diversity of applicable law).

  40. [40]. See Hartmann v. Time, Inc., 166 F.2d 127, 132 (3d Cir. 1947) (“[T]he one issue of a newspaper or magazine, although it consists of thousands of copies widely distributed, gives rise to one cause of action, there being but one publication ....”); Gregoire v. G.P. Putnam’s Sons, 81 N.E.2d 45, 47 (N.Y. 1948) (“[T]he publication of a defamatory statement in a single issue of a newspaper, or a single issue of a magazine, although such publication consists of thousands of copies widely distributed, is, in legal effect, one publication which gives rise to one cause of action ....”).

  41. [41]. Restatement (Second) of Torts § 577A(3) (Am. L. Inst. 1977).

  42. [42]. 50 Am. Jur. 2d Libel and Slander § 244 (2020) (“This ‘single publication rule’ protects defendants from being sued separately for each copy of a book or newspaper containing the allegedly defamatory statement.”).

  43. [43]. Kumar, supra note 34, at 642–43.

  44. [44]. See id. at 642–43, 643 n.24. Montana still adheres to the multiple publication rule. See Lewis v. Reader’s Digest Ass’n, Inc., 512 P.2d 702, 704 (Mont. 1973) (“Despite the numerical weight of authority following the single publication rule, we consider it unsound.”).

  45. [45]. Defamation Act 2013, c. 26, § 8(3) (U.K.).

  46. [46]. Unif. Single Publ’n Act § 1, 9C U.L.A. 173 (1952); see also generally Comment, Problems Under the Uniform Single Publication Act, 15 Wash. & Lee L. Rev. 321 (1958) (discussing the Act).

  47. [47]. Unif. Single Publ’n Act § 1.

  48. [48]. See Ariz. Rev. Stat. Ann. § 12-651 (2020); Cal. Civ. Code § 3425.3 (West 2020); 740 Ill. Comp. Stat. 165/1 (2020); 42 Pa. Cons. Stat. § 8341 (1978); see also Idaho Code § 6-702 (2020); N.M. Stat. Ann. § 41-7-1 (2020); N.D. Cent. Code § 14-02-10 (2019).

  49. [49]. Holloway v. Butler, 662 S.W.2d 688, 690 (Tex. App. 1983) (“[W]e now take the opportunity to adopt the ... Uniform Single Publication Act[.]”).

  50. [50]. See Kumar, supra note 34, at 643.

  51. [51]. Restatement (Second) of Torts § 577A(1) (Am. L. Inst. 1977) (“Except as stated in Subsections (2) and (3), each of several communications to a third person by the same defamer is a separate publication.”).

  52. [52]. See, e.g., Schneider v. United Airlines, Inc., 256 Cal. Rptr. 71, 74–75 (Ct. App. 1989) (“[T]he single publication rule ... does not include separate aggregate publications on different occasions.... In these cases the publication reaches a new group and the repetition justifies
    a new cause of action.” (alterations in original) (quoting Kanarek v. Bugliosi, 166 Cal. Rptr.
    526, 530 (Ct. App. 1980))); Firth v. State, 775 N.E.2d 463, 466 (N.Y. 2002) (“Republication, retriggering the period of limitations, occurs upon a separate aggregate publication from the original, on a different occasion, which is not merely ‘a delayed circulation of the original edition.’” (quoting Rinaldi v. Viking Penguin, Inc., 420 N.E.2d 377, 382 (N.Y. 1981))); Swafford v. Memphis Individual Prac. Ass’n, No. 02A01-9612-CV-00311, 1998 WL 281935, at *4 (Tenn. Ct. App. June 2, 1998) (“[T]he single publication rule does not apply to ‘aggregate publications on different occasions.’” (quoting Restatement (Second) of Torts § 577A cmt. d)); Restatement (Second) of Torts § 577A cmt. d (“[T]he single publication rule ... does not include separate aggregate publications on different occasions. Thus if the same defamatory statement is published in the morning and evening editions of a newspaper, each edition is a separate single publication and there are two causes of action.”); Kumar, supra note 34, at 645, 652 (same).

  53. [53]. Restatement (Second) of Torts § 577A cmt. d.

  54. [54]. Id.

  55. [55]. Id.

  56. [56]. The term “Web 2.0” refers to websites that enable users to contribute content
    and interact with each other. See William L. Hosch, Web 2.0, Britannica, https:// (last updated Apr. 6, 2017).

  57. [57]. Perry & Zarsky, Who Should Be Liable?, supra note 6, at 162.

  58. [58]. See Byron F. Marchant, On-Line on the Internet: First Amendment and Intellectual Property Uncertainties in the On-Line World, 39 How. L.J. 477, 492 & n.90 (1996) (suggesting the application of the rule to online publications).

  59. [59]. Firth v. State, 775 N.E.2d 463, 464 (N.Y. 2002).

  60. [60]. Id.

  61. [61]. Id. at 464–65.

  62. [62]. Id. at 465–66.

  63. [63]. Id.

  64. [64]. Id. at 466.

  65. [65]. Id.

  66. [66]. See, e.g., Van Buskirk v. N.Y. Times Co., 325 F.3d 87, 89 (2d Cir. 2003) (endorsing Firth); 50 Am. Jur. 2d Libel and Slander § 244 (2020) (“[T]he rule applies to publications on the Internet so that continuous access to an allegedly defamatory article posted via hyperlinks to a website
    is not a republication.”). But cf. Odelia Braun, Comment, Internet Publications and Defamation:
    Why the Single Publication Rule Should Not Apply
    , 32 Golden Gate U. L. Rev. 325, 332 (2002) (suggesting that the rule should not apply to internet publications).

  67. [67]. Firth v. State, 731 N.Y.S.2d 244, 247 (App. Div. 2001), aff’d, 775 N.E.2d 463 (N.Y. 2002) (concluding “the single publication rule applies to” a statement “placed on the Internet ... and never modified thereafter”); Firth v. State, 706 N.Y.S.2d 835, 843 (Ct. Cl. 2000), aff’d, 731 N.Y.S.2d 244 (App. Div. 2001), aff’d, 775 N.E.2d 463 (N.Y. 2002) (“[I]n the absence of some alteration or change in form its continued availability on the Internet does not constitute a republication acting to begin the Statute of Limitations anew each day.”).

  68. [68]. Firth, 775 N.E.2d at 466–67.

  69. [69]. Yeager v. Bowlin, 693 F.3d 1076, 1082 (9th Cir. 2012) (holding that when the republished statement is substantially altered it constitutes republication).

  70. [70]. Id. (holding that when the republished statement is directed to a new audience, it constitutes new publication); Kumar, supra note 34, at 655, 657–58 (“If a website initially attracts a small audience, but the size of the audience later increases, courts should investigate the intent of the publisher and find that republication occurred if the new distribution was intentional.”).

  71. [71]. Adeline A. Allen, Twibel Retweeted: Twitter Libel and the Single Publication Rule, 15 J. High Tech. L. 63, 89–90 (2014).

  72. [72]. Id. at 91–94.

  73. [73]. Kumar, supra note 34, at 658.

  74. [74]. See, e.g., Dixson v. Newsweek, Inc., 562 F.2d 626, 631 (10th Cir. 1977) (“[T]he republication of false defamatory statements is as much a tort as the original publication.”); Khawar v. Globe Int’l, Inc., 965 P.2d 696, 704 (Cal. 1998) (“[O]ne who republishes a defamatory statement is deemed thereby to have adopted it and so may be held liable, together with the person who originated the statement ....”); Frommoethelydo v. Fire Ins. Exch., 721 P.2d 41, 46 (Cal. 1986) (“When one person repeats another’s defamatory statement, he may be held liable for republishing the same libel or slander.”); Morse v. Times-Republican Printing Co., 100 N.W. 867, 871 (Iowa 1904) (“Every repetition or republication of a libel is a new libel, and each publisher is answerable for his act to the same extent as if the calumny originated with him.”); Restatement (Second) of Torts §§ 578, 578 cmt. b (Am. L. Inst. 1977) (“[O]ne who repeats or otherwise republishes defamatory matter is subject to liability as if he had originally published it.”); 50 Am. Jur. 2d Libel and Slander §§ 244–245 (2020) (“Every repetition of a defamation is a publication in itself”; “the publisher of a false statement made by another person, when the publisher knows the statement to be false, is not protected by the fact that someone else
    made the statement.”); Goldberg & Zipursky, supra note 5, at 322 (“[U]nder the venerable republication rule, various actors who repeat another’s defamatory writing are subject to liability as if they were the original authors ....”); Joel Timmer, Fighting Falsity: Fake News, Facebook, and the First Amendment, 35 Cardozo Arts & Ent. L.J. 669, 689–90 (2017); Zipursky, supra note 5, at 4 (discussing the traditional republication rule); Matt C. Sanchez, Note, The Web Difference: A Legal and Normative Rationale Against Liability for Online Reproduction of Third-Party Defamatory Content, 22 Harv. J.L. & Tech. 301, 302 (2008) (same).

  75. [75]. See, e.g., Brennan v. Kadner, 814 N.E.2d 951, 959 (Ill. App. Ct. 2004) (holding that disclosing the originator’s name does not exempt the republisher from liability); Herron v. Trib. Publ’g Co., 736 P.2d 249, 259 (Wash. 1987) (same).

  76. [76]. Restatement (Second) of Torts § 578 cmt. b, illus. 1.

  77. [77]. Flowers v. Carville, 310 F.3d 1118, 1128 (9th Cir. 2002) (alteration in original); see also Barry v. Time, Inc., 584 F. Supp. 1110, 1122 (N.D. Cal. 1984) (using the same language as quoted above); Harris v. Minvielle, 19 So. 925, 928 (La. 1896) (same); Hous. Chron. Publ’g Co. v. Wegner, 182 S.W. 45, 48 (Tex. App. 1915) (same).

  78. [78]. Condit v. Dunne, 317 F. Supp. 2d 344, 363 (S.D.N.Y. 2004); see also Goldberg & Zipursky, supra note 5, at 322 (“The harm done by republication is often equal to or greater than that caused by the original statement.”).

  79. [79]. See, e.g., Agbapuruonwu v. NBC Subsidiary (WRC-TV), LLC, No. 19-1236, 2020 U.S. App. LEXIS 26163, at *10–15 (4th Cir. Aug. 18, 2020) (discussing the “fair report privilege” in Virginia); Nix v. ESPN, Inc., 772 F. App’x 807, 809–10, 812–13 (11th Cir. 2019) (discussing the “fair report privilege” in Florida and New York).

  80. [80]. See, e.g., Hogan v. Herald Co., 446 N.Y.S.2d 836, 841 (App. Div. 1982) (discussing the privilege); Restatement (Second) of Torts § 611 (“The publication of defamatory matter concerning another in a report of an official action or proceeding or of a meeting open to the public that deals with a matter of public concern is privileged if the report is accurate and complete or a fair abridgement of the occurrence reported.”); Amanda Groover Hyland, The Taming of the Internet: A New Approach to Third-Party Internet Defamation, 31 Hastings Commc’ns & Ent. L.J. 79, 95 (2008); Matthew E. Kelley & Steven D. Zansberg, A Little Birdie Told Me, “You’re A Crook”: Libel in the Twittersphere and Beyond, Commc’ns Law., Mar. 2014, at 1, 37 (“[The] fair report privilege shielding from defamation liability reports regarding material in government documents and proceedings.”).

  81. [81]. Chapin v. Knight-Ridder, Inc., 993 F.2d 1087, 1098 (4th Cir. 1993); Hyland, supra note 80, at 95.

  82. [82]. Restatement (Second) of Torts § 611 cmt. a.

  83. [83]. Hogan, 446 N.Y.S.2d at 841.

  84. [84]. See, e.g., Khawar v. Globe Int’l, Inc., 965 P.2d 696, 704–08 (Cal. 1998) (discussing the origins and limits of the “neutral reportage privilege”).

  85. [85]. Edwards v. Nat’l Audubon Soc’y, Inc., 556 F.2d 113, 120 (2d Cir. 1977) (“[W]hen a responsible, prominent organization ... makes serious charges against a public figure, the First Amendment protects the accurate and disinterested reporting of those charges ....”); Lasky v. Am. Broad. Cos., 631 F. Supp. 962, 970 (S.D.N.Y. 1986) (following Edwards); Hyland, supra note 80, at 95; Khawar, 965 P.2d at 698 (“[T]here are two types of public figures: ‘Some occupy positions of such persuasive power and influence that they are deemed public figures for all purposes. More commonly, those classed as public figures have thrust themselves to the forefront of particular public controversies in order to influence the resolution of the issues involved. In either event, they invite attention and comment.’” (citation omitted)).

  86. [86]. Edwards, 556 F.2d at 120.

  87. [87]. Chapin v. Knight-Ridder, Inc., 993 F.2d 1087, 1097 (4th Cir. 1993).

  88. [88]. Harte-Hanks Commc’ns, Inc. v. Connaughton, 491 U.S. 657, 661 n.1 (1989) (“Petitioner did not argue ... that the neutral reportage doctrine immunized its coverage of [another’s] allegations. Accordingly, we do not review this aspect of the District Court’s judgment.”).

  89. [89]. See, e.g., Dickey v. CBS, Inc., 583 F.2d 1221, 1226 (3d Cir. 1978) (“[A] constitutional privilege of neutral reportage is not created, as appellee would have us find, merely because an individual newspaper or television or radio station decides that a particular statement is newsworthy.”); McCall v. Courier-J. & Louisville Times Co., 623 S.W.2d 882, 886–87 (Ky. 1981) (rejecting the doctrine); Postill v. Booth Newspapers, Inc., 325 N.W.2d 511, 517–18 (Mich. Ct. App. 1982) (same); Hogan v. Herald Co., 446 N.Y.S.2d 836, 841–43 (App. Div. 1982) (same).

  90. [90]. See, e.g., Brown v. Courier Herald Publ’g Co., 700 F. Supp. 534, 537 (S.D. Ga. 1988) (discussing “the wire service defense”); Nelson v. Associated Press, Inc., 667 F. Supp. 1468, 1476
    –77 (S.D. Fla. 1987) (same).

  91. [91]. Winn v. Associated Press, 903 F. Supp. 575, 579 (S.D.N.Y. 1995); Brown, 700 F. Supp.
    at 537; Nelson, 667 F. Supp. at 1476–77; Layne v. Trib. Co., 146 So. 234, 238–39 (Fla. 1933); Appleby v. Daily Hampshire Gazette, 478 N.E.2d 721, 725–26 (Mass. 1985); James E. Boasberg, With Malice Toward None: A New Look at Defamatory Republication and Neutral Reportage, 13
    Hastings Commc’ns & Ent. L.J. 455, 458–65 (1991) (discussing the history, application, and interpretation of the wire service defense); Hyland, supra note 80, at 96 (same).

  92. [92]. See, e.g., Winn, 903 F. Supp. at 579; Appleby, 478 N.E.2d at 725–26.

  93. [93]. See, e.g., Brown, 700 F. Supp. at 537; Appleby, 478 N.E.2d at 725–26.

  94. [94]. See Goldberg & Zipursky, supra note 5, at 334 (explaining that the legislative reform, as interpreted by the courts, “is both wildly in derogation of the common law and stands largely to undermine the civil recourse principle as it applies to a basic and long-recognized legal wrong: the wrong of libel”); Zipursky, supra note 5, at 14 (“[A]t least in several prominent and populous jurisdictions—the republication rule is no longer operative for defamatory statements posted on the Internet.”).

  95. [95]. See Goldberg & Zipursky, supra note 5, at 322–23 (explaining the distinction between “publishers” and “distributors”); Hyland, supra note 80, at 96–97 (discussing the trichotomy); Bryan J. Davis, Comment, Untangling the “Publisher” Versus “Information Content Provider” Paradox of 47 U.S.C. § 230: Toward a Rational Application of the Communications Decency Act in Defamation Suits Against Internet Service Providers, 32 N.M. L. Rev. 75, 79–83 (2002) (same).

  96. [96]. See Daniel J. Solove & Paul M. Schwartz, Information Privacy Law 184 (4th ed. 2011) (discussing publishers’ liability); Sanchez, supra note 74, at 303–04 (same).

  97. [97]. Restatement (Second) of Torts § 581(1) (Am. L. Inst. 1977) (allowing liability of distributors following notice); Sewali K. Patel, Immunizing Internet Service Providers from Third-Party Internet Defamation Claims: How Far Should Courts Go?, 55 Vand. L. Rev. 647, 651–52 (2002) (discussing distributors’ liability); Sanchez, supra note 74, at 303–04 (same); Zipursky, supra note 5, at 21 (explaining that section 581(1) of the Restatement (Second) of Torts is the source of distributor liability).

  98. [98]. See, e.g., Anderson v. N.Y. Tel. Co., 320 N.E.2d 647, 649 (N.Y. 1974) (holding that common carriers or conduits, such as a telephone company, are not liable for defamation by customers); Austin v. CrystalTech Web Hosting, 125 P.3d 389, 392 (Ariz. Ct. App. 2005) (same); Patel, supra note 97, at 651 (same); Zipursky, supra note 5, at 21 (explaining that common carriers are not liable as publishers).

  99. [99]. The main problem revolves around the distinction between publishers and distributors because common carriers, including internet access providers, are still not liable (Lunney v. Prodigy Servs. Co., 723 N.E.2d 539, 542–43 (N.Y. 1999)), but online service providers, as opposed to internet access providers, cannot normally be regarded as common carriers.

  100. [100]. Cubby, Inc. v. CompuServe, Inc., 776 F. Supp. 135 (S.D.N.Y. 1991).

  101. [101]. Id. at 139–42.

  102. [102]. Stratton Oakmont, Inc. v. Prodigy Servs. Co., 1995 WL 323710 (N.Y. Sup. Ct. May 24, 1995), superseded by statute, Communications Decency Act of 1996, Pub. L. 104-104, 110 Stat. 133, as recognized in Shiamili v. Real Est. Grp. of N.Y., 952 N.E.2d 1011 (N.Y. 2011).

  103. [103]. Id. at *10­–11; see also id. at *13 (“[Prodigy’s] conscious choice, to gain the benefits of editorial control, has opened it up to a greater liability than CompuServe and other computer networks that make no such choice.”); Anupam Chander, How Law Made Silicon Valley, 63 Emory L.J. 639, 650–51 (2014) (“Prodigy could be liable as a publisher because it had advertisedits editorial control over the site.”); Zipursky, supra note 5, at 24–26 (explaining that Prodigy exercised editorial control as a marketing ploy, offering a family-friendly experience to clients).

  104. [104]. See Zeran v. Am. Online, Inc., 129 F.3d 327, 331–33 (4th Cir. 1997).

  105. [105]. See Zipursky, supra note 5, at 17 (“[T]he fledgling industry went ballistic.”).

  106. [106]. Telecommunications Act of 1996, Pub. L. No. 104-104, § 230, 110 Stat. 56, 137–39, (codified as amended at 47 U.S.C. §230 (2018)).

  107. [107]. 47 U.S.C. § 230(c)(1).

  108. [108]. Zeran, 129 F.3d at 330–33.

  109. [109]. Empirical studies have shown, however, that more than one-third of such claims survive the §230 defense, and, accordingly, websites often have to engage in long and expensive legal battles. See 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, 493(2010); Chander, supra note 103, at 653–55.

  110. [110]. In addition to its broad judicial interpretation, the section’s protection has been extended by Congress. For example, 28 U.S.C. § 4102(c)(1) provides that U.S. courts “shall not recognize or enforce” foreign defamation judgments that are inconsistent with § 230.

  111. [111]. Zeran, 129 F.3d at 332–33; see also Blumenthal v. Drudge, 992 F. Supp. 44, 52 (D.D.C. 1998) (“Congress made no distinction between publishers and distributors in providing immunity from liability.”); Doe v. Am. Online, Inc., 783 So. 2d 1010, 1014–17 (Fla. 2001) (holding that imposing distributor liability on internet service providers would frustrate the objectives of § 230); Hyland, supra note 80, at 107 (explaining that courts draw no distinction between “publishers” and “distributors” for the purpose of the CDA § 230 immunity); Joshua Landau & Kate Willcox, Within the Law: Dealing With Non-Confidential Sensitive Information in the Age of Online Legal Tabloids, 23 Geo. J. Legal Ethics 667, 680 (2010) (same); Timmer, supra note 74, at 691 (same). The California Court of Appeals diverged from this position in Barrett v. Rosenthal, 9 Cal. Rptr. 3d 142, 166 (Ct. App. 2004), holding that distributor (knowledge based) liability survived the congressional grant of immunity. But the California Supreme Court aligned itself with Zeran and its progeny. Barrett v. Rosenthal, 146 P.3d 510, 513 (Cal. 2006) (“[S]ection 230 prohibits ‘distributor’ liability for Internet publications.”); see also id. at 518–26 (rejecting the Court of Appeals’ justifications).

  112. [112]. Hyland, supra note 80, at 107.

  113. [113]. Section 230 applies a fortiori to ISPs in the narrow sense, which “only transmit packets of data,” and can be regarded as common carriers. Goldberg & Zipursky, supra note 5, at 320 n.32.

  114. [114]. Hyland, supra note 80, at 106–07.

  115. [115]. See, e.g., Carafano v., Inc., 339 F.3d 1119, 1125 (9th Cir. 2003) (extending § 230 immunity to a case involving invasion of privacy, defamation, and misappropriation of the right to publicity); Chander, supra note 103, at 651, 653 n.58 (providing an extensive list of cases and concluding that §230 “largely immunized online service providers from secondary liability for most torts committed through their service”); Zipursky, supra note 5, at 9, 13–14 (“[Section 230] has been applied in scores of cases to undercut not only claims for defamation, but other common law tort speech claims—such as invasion of privacy—as well as a variety of state and federal statutory claims ....”).

  116. [116]. See, e.g., Hassell v. Bird, 420 P.3d 776, 789 (Cal. 2018) (“Where, as here, an Internet intermediary’s relevant conduct in a defamation case goes no further than the mere act of publication—including a refusal to depublish upon demand, after a subsequent finding that the published content is libelous—section 230 prohibits this kind of directive.”); Medytox Sols., Inc. v., Inc., 152 So. 3d 727, 731 (Fla. Dist. Ct. App. 2014) (“An action to force a website to remove content on the sole basis that the content is defamatory is necessarily treating the website as a publisher, and is therefore inconsistent with section 230.”).

  117. [117]. Batzel v. Smith, 333 F.3d 1018 (9th Cir. 2003).

  118. [118]. Id. at 1021–22.

  119. [119]. Id. at 1030.

  120. [120]. Id.

  121. [121]. The court in Batzel also rejected the proposed distinction between deciding to publish third-party content and refusing to remove such content, holding that § 230 applies to both. Id. at 1032, 1035. This point is criticized by Goldberg & Zipursky, supra note 5, at 330–37, who agree with the dissent in Batzel that § 230 prevents liability for transmitting unlawful content or failing to remove it after receiving a notice, but not for actively posting content created by others. See also Zipursky, supra note 5, at 42–44 (endorsing the active/passive distinction).

  122. [122]. Barrett v. Rosenthal, 146 P.3d 510, 515 (Cal. 2006) (“This appears to be the first published case in which section 230 immunity has been invoked by an individual who had no supervisory role in the operation of the Internet site where allegedly defamatory material appeared, and who thus was clearly not a provider of an ‘interactive computer service’ under the broad definition provided in the CDA.”).

  123. [123]. Id. at 513–14.

  124. [124]. Id.

  125. [125]. Id. at 514, 518–26.

  126. [126]. Id. at 526.

  127. [127]. Id. at 527.

  128. [128]. Id. at 527–29 (“The Batzel majority [concluded] that no logical distinction can be drawn between a defendant who actively selects information for publication and one who screens submitted material, removing offensive content.”).

  129. [129]. Id. at 529.

  130. [130]. Julio Sharp-Wasserman, Note, Section 230(C)(1) of the Communications Decency Act and the Common Law of Defamation: A Convergence Thesis, 20 Colum. Sci. & Tech. L. Rev. 195, 229 (2018).

  131. [131]. See Allen, supra note 71, at 86 (discussing republication through social-media); Kelley & Zansberg, supra note 80, at 37–39 (same); Timmer, supra note 74, at 690 (same).

  132. [132]. See, e.g., Mitan v. A. Neumann & Assocs., LLC., Civ. No. 08-6154, 2010 WL 4782771, at *4 (D.N.J. Nov. 17, 2010) (“[Immunity is provided to] persons who republish the work of other persons through internet-based methodologies, such as websites, blogs, and email.”); Novins v. Cannon, Civ. No. 09-5354, 2010 WL 1688695, at *2 (D.N.J. Apr. 27, 2010) (“[Those who e-mailed the content] acted as re-publishers of another person’s information, and as such they are protected by the CDA.”); Phan v. Pham, 105 Cal. Rptr. 3d 791, 792 (Ct. App. 2010) (“[W]hen you receive a defamatory e-mail over the internet and simply hit the forward icon on your computer, sending it on someone else[,] ... you cannot be held liable for the defamation.” (emphasis omitted)); Citron, supra note 1.

  133. [133]. See, e.g., Vazquez v. Buhl, 90 A.3d 331, 334, 338–44 (Conn. App. Ct. 2014) (holding that an article urging viewers to read a defamatory report, and containing a hyperlink thereto, is protected from liability under § 230).

  134. [134]. Zeran v. Am. Online, Inc., 129 F.3d 327, 330–31 (4th Cir. 1997) (presenting the legislative goals of § 230); see also Carafano v., Inc., 339 F.3d 1119, 1122 (9th Cir. 2003) (same); Jones v. Dirty World Ent. Recordings LLC, 755 F.3d 398, 407 (6th Cir. 2014) (adding that § 230 “keep[s] government interference in the medium to a minimum” (alteration in original) (quoting Zeran, 129 F.3d at 330)).

  135. [135]. See Telecommunications Act of 1996, Pub. L. No. 104-104, § 509, 110 Stat. 56, 137 (codified as amended at 47 U.S.C. §230 (2018)); text accompanying supra note 103.

  136. [136]. 47 U.S.C. § 230(c)(2); e360Insight, LLC v. Comcast Corp., 546 F. Supp. 2d 605, 607
    –08 (N.D. Ill. 2008) (holding that under subsection (c)(2) the defendant, an e-mail service provider, is not liable for blocking the plaintiff’s spam to the defendant’s customers). But see Goldberg & Zipursky, supra note 5, at 326–29 (contending that subsection (c)(2) is akin to Good Samaritan legislation, which modified negligence law so that it does not discourage benevolent conduct); Zipursky, supra note 5, at 30–35 (analyzing the legislative history of § 230, concluding that it “plainly constitutes an Internet version of the traditional Good Samaritan statute. To incentivize voluntarily protecting those who are at peril of injury, it negates the common law principle that voluntarily protecting others creates an affirmative duty where
    none existed before.” (footnote omitted)); Benjamin C. Zipursky, Thinking in the Box in Legal Scholarship: The Good Samaritan and Internet Libel, 66 J. Legal Educ. 55, 60–61 (2016) (asserting § 230 does not impose an affirmative duty on users to censor obscene or defamatory material).

  137. [137]. Barrett v. Rosenthal, 146 P.3d 510, 517 (Cal. 2006).

  138. [138]. Id.

  139. [139]. Zeran, 129 F.3d at 333.

  140. [140]. Id.

  141. [141]. Barrett, 146 P.3d at 529 (referring to the benefits of the internet enumerated in 47 U.S.C. § 230(a)(3)).

  142. [142]. Id.; see also Sharp-Wasserman, supra note 130, at 230 (explaining that individual users’ risk aversion might also incentivize censorship).

  143. [143]. Sharp-Wasserman, supra note 130, at 231.

  144. [144]. Barrett, 146 P.3d at 525, 529.

  145. [145]. Id. at 526.

  146. [146]. Id.

  147. [147]. Id. at 527 n.19.

  148. [148]. See, e.g., Jones v. Dirty World Ent. Recordings LLC, 755 F.3d 398, 408­–09, 413 (6th Cir. 2014) (“[The immunity] applies only to the extent that an interactive computer service provider is not also the information content provider of the content at issue.... [The Sixth Circuit] adopt[s] the material contribution test to determine whether a website operator is ‘responsible ....’”); Fair Hous. Council of San Fernando Valley v., LLC, 521 F.3d 1157, 1167–68 (9th Cir. 2008) (“[A] website helps to develop unlawful content, and thus falls within the exception tosection 230, if it contributes materially to the alleged illegality of the conduct.”); Doe v. City of New York, 583 F. Supp. 2d 444, 449 (S.D.N.Y. 2008) (“When [defendant] added his own allegedly tortious speech to the third-party content he forwarded, he fell out of
    the statute’s protections.” (citing Blumenthal v. Drudge, 992 F. Supp. 44, 50 (D.D.C. 1998))); Blumenthal, 992 F. Supp. at 50 (“Section 230(c)(1) would not immunize [defendants] with respect to any information [they] developed or created entirely by [themselves] ....”); Vazquez v. Buhl, 90 A.3d 331, 344 (Conn. App. Ct. 2014) (“[T]he meaning of ‘development in part,’ as defined in case law interpreting the language of § 230(f)(3), covers conduct ranging from ‘material contribution’ to ‘solicitation’ of the information at issue.... It is immaterial whether the defendant amplified, endorsed, or adopted the defamatory statements, because the defendant played no role in their composition ... [and] was not an ‘information content provider.’”); Shiamili v. Real Est. Grp. of N.Y., Inc., 952 N.E.2d 1011, 1020 (N.Y. 2011) (holding that the website operator merely re-posted defamatory statements about the plaintiff and that the “added headings and illustration do not materially contribute to the defamatory nature of the third-party statements,” so the immunity applied); Zipursky, supra note 5, at 50–51 (“[If] the republisher has added independent content ... there can be liability that is not for information content supplied by a third party, but supplied by the defendant independently.”).

  149. [149]. Barrett, 146 P.3d at 527 n.19.

  150. [150]. Id. at 529 (Moreno, J., concurring).

  151. [151]. Id. at 530.

  152. [152]. Citron, supra note 1.

  153. [153]. See Mike Isaac, Nude Photos of Jennifer Lawrence Are Latest Front in Online Privacy Debate,
    N.Y. Times (Sept. 2, 2014), [].

  154. [154]. See supra Section II.B.1.

  155. [155]. See, e.g., Zipursky, supra note 5, at 52 (“[I]t would be a particularly bizarre policy judgment to eliminate the republication rule for the Internet, but for no other medium.”); Brittan Heller, Note, Of Legal Rights and Moral Wrongs: A Case Study of Internet Defamation, 19 Yale J.L. & Feminism 279, 286 (2007) (“[W]hat is impermissible in the real world should not be permitted in the virtual world ....”); Jae Hong Lee, Note, Batzel v. Smith & Barrett v. Rosenthal: Defamation Liability for Third-Party Content on the Internet, 19 Berkeley Tech. L.J. 469, 488 (2004) (“The Internet ... is probably not so unique as to require the formulation of a truly novel approach to defamation liability.”); Sanchez, supra note 74, at 302 (“[Critics] argue that immunizing online reproduction while punishing identical offline reproduction makes little sense.” (footnote omitted)); Melissa A. Troiano, Comment, The New Journalism? Why Traditional Defamation Laws Should Apply to Internet Blogs, 55 Am. U. L. Rev. 1447, 1465, 1468–69 (2006) (“[D]efamatory speech should not be protected in some instances just because the defamer disseminated the message through one medium, but then not protected when the same speech is transmitted through a different medium.”).

  156. [156]. Lyrissa Barnett Lidsky, Silencing John Doe: Defamation & Discourse in Cyberspace, 49 Duke L.J. 855, 894 & n.204 (2000).

  157. [157]. Sanchez, supra note 74, at 313 (“Large media entities can mitigate the danger of suit by maintaining defamation insurance, employing in-house counsel, and including litigation expenses in their budget.”).

  158. [158]. Id. at 309.

  159. [159]. Id. at 310.

  160. [160]. See Wash. Post Co. v. Keogh, 365 F.2d 965, 968 (D.C. Cir. 1966) (“The threat of being put to the defense of a lawsuit ... may be as chilling to the exercise of First Amendment freedoms as fear of the outcome of the lawsuit itself ....”); Sharlene A. McEvoy, “The Big Chill”: Business Use of the Tort of Defamation to Discourage the Exercise of First Amendment Rights, 17 Hastings Const. L.Q. 503, 505 (1990) (“Although ... ‘the probability of an adverse judgment is small,’ the price of
    ... [defamation lawsuits] can be very high .... [D]iscouraging even the hardiest souls from exercising their first amendment rights.” (citation omitted)); Sanchez, supra note 74, at 312 (“[T]he cost of defending a lawsuit may have as much of a deterrent effect on speech as the fear of large jury awards.”).

  161. [161]. See Liberty Lobby, Inc. v. Dow Jones & Co., 838 F.2d 1287, 1292–93, 1298 (D.C. Cir. 1988) (“The common law of libel has long held that one who republishes a defamatory statement ‘adopts’ it as his own, and is liable in equal measure to the original defamer.” (citation omitted)).

  162. [162]. Sanchez, supra note 74, at 310–11.

  163. [163]. Id. at 310–12, 315.

  164. [164]. Id. at 310–12.

  165. [165]. But see Troiano, supra note 155, at 1479–81 (arguing that determining whether something is defamatory “is ... a matter of common sense,” and that “bloggers [can] easily control what [content] to omit based on [its apparent] defamatory nature”).

  166. [166]. See, e.g., Restatement (Second) of Torts § 578 illus. 3–4 (Am. L. Inst. 1977) (applying the rules to individual non-professional republishers); see also Hoover v. Peerless Publ’ns, Inc., 461 F. Supp. 1206, 1209 (E.D. Pa. 1978) (same).

  167. [167]. Sanchez, supra note 74, at 315.

  168. [168]. Id. at 315–16.

  169. [169]. This conceptual distinction is a novel contribution to legal literature, and therefore cannot be supported by any external authority.

  170. [170]. See infra Section II.B.3.i.

  171. [171]. See infra Sections II.B.3.ii–.iii.

  172. [172]. See infra notes 173–85 and accompanying text.

  173. [173]. Bürgerliches Gesetzbuch [BGB] [Civil Code], § 823, para. 1, [] (Ger.).

  174. [174]. Oberlandesgericht Dresden [OLG] [Higher Regional Court of Dresden], Feb. 7, 2017, 4 U 1419/16, sec. II.2.b (Ger.) (“Ist dem ‘Teilen’ für sich genommen keine über die Verbreitung des Postings hinausgehende Bedeutung zuzumessen.”); Oberlandesgericht Frankfurt [OLG] [Higher Regional Court of Frankfurt], Nov. 26, 2015, 16 U 64/15, sec. II (Ger.) (same).

  175. [175]. OLG Dresden, Feb. 7, 2017, 4 U 1419/16, para. II.2.b (Ger.); OLG Frankfurt, Nov. 26, 2015, 16 U 64/15, sec. II. (Ger.).

  176. [176]. OLG Dresden, Feb. 7, 2017, 4 U 1419/16, para. II.2.b (Ger.).

  177. [177]. Medienmitteilung des Bezirksgerichts Zürich, Bezirksgericht Zürich (May 29, 2017),
    Zuerich/GG160246.pdf []; see also Agence France-Presse, Man Fined by Swiss Court for ‘Liking’ Defamatory Comments on Facebook, Guardian (Nov. 27, 2017, 6:31 PM), [].

  178. [178]. Bezirksgericht Zürich, supra note 177; France-Presse, supra note 177.

  179. [179]. Crookes v. Newton, [2011] S.C.R. 269, 271, 294–95 (Can.).

  180. [180]. Id. at 294 (emphasis omitted).

  181. [181]. Visscher v Mar Union of Austl [No. 6] (2014) NSWSC 350, ¶¶ 28–31 (Austl.).

  182. [182]. Id. ¶¶ 28–29.

  183. [183]. Id. ¶¶ 30–31.

  184. [184]. Bolton v Stoltenberg [2018] NSWSC 1518, ¶ 171 (Austl.).

  185. [185]. Id.

  186. [186]. See infra notes 187–213 and accompanying text.

  187. [187]. PCA 1239/19 Shaul v. Nidaily Commc’n, Ltd., Nevo Legal Database (Jan. 1, 2020) (Isr.).

  188. [188]. Id. at para. 12.

  189. [189]. Id. at paras. 14–15.

  190. [190]. Id. at paras. 1, 15.

  191. [191]. Id. at paras. 16–18.

  192. [192]. Id. at para. 19.

  193. [193]. Id. at paras. 20–25.

  194. [194]. Id. at paras. 23, 43, 46.

  195. [195]. Id.

  196. [196]. Id.

  197. [197]. Id.

  198. [198]. Id.

  199. [199]. Id.

  200. [200]. Id. at paras. 24, 45.

  201. [201]. Id. at para. 24 (citing § 19(1), Prohibition of Defamation Act, 1965, SH 240 (Isr.)).

  202. [202]. Id. at paras. 26–30 (summarizing CivA (DC TA) 35757-10-16 Nidaily Commc’n, Ltd. v. Shaul, paras. 57­–63, 72–78, 85, Nevo Legal Database (Jan. 16, 2019) (Isr.)).

  203. [203]. Id. at paras. 43–46.

  204. [204]. Id. at para. 73.

  205. [205]. Id. at paras. 53–55.

  206. [206]. Id. at paras. 56–58.

  207. [207]. Id. at para. 46.

  208. [208]. Id. at para. 31.

  209. [209]. A South African court held that when users are “tagged” they can remove the tags, and by failing to do so they knowingly associate themselves with the respective message and can be liable for its publication. See Isparta v. Richter 2013 (6) SA 529 (HC) at para. 35 (S. Afr.).

  210. [210]. Bolton v Stoltenberg [2018] NSWSCR (Eq) 1518, ¶¶ 171, 176 (Austl.).

  211. [211]. Id. ¶ 171 (quoting Google Inc v Duffy [2017] SASCFC 130, 133 (Austl.)).

  212. [212]. Id. ¶ 176.

  213. [213]. Id.

  214. [214]. Crookes v. Newton, [2011] S.C.R. 269 (Can.).

  215. [215]. Id. at 270.

  216. [216]. Id. at 286.

  217. [217]. Id.

  218. [218]. Id. at 291.

  219. [219]. Id. at 292.

  220. [220]. See infra notes 223–32 and accompanying text.

  221. [221]. See supra Section II.B.2.

  222. [222]. Barrett v. Rosenthal, 146 P.3d 510, 513 (Cal. 2006); see also id. at 529 (“Plaintiffs are free under section 230 to pursue the originator of a defamatory Internet publication.”).

  223. [223]. Allen, supra note 71, at 84; see also Sanchez, supra note 74, at 318 (explaining that the “[d]efamed ... retain[s] the right to take action against the originator of the defamatory speech”).

  224. [224]. Hickey v. St. Martin’s Press, Inc., 978 F. Supp. 230, 236–37 (D. Md. 1997); see also Blue Ridge Bank v. Veribanc, Inc., 866 F.2d 681, 689–90 (4th Cir. 1989) (holding that a publisher may be liable for republication by another if “it is the natural and probable consequence”); Hatfill v. Foster, 401 F. Supp. 2d 320, 343 (S.D.N.Y. 2005), rev’d on reconsideration, 415 F. Supp. 2d 353 (S.D.N.Y. 2006) (same); Tunca v. Painter, 965 N.E.2d 1237, 1260–63 (Ill. App. Ct. 2012) (holding that “a defamer is liable for damages caused by repetitions that were reasonably foreseeable, or the natural and probable consequence of [the] original statement”); Restatement (Second) of Torts § 576(c) (Am. L. Inst. 1977) (“[A] publication ... is [the] legal cause of ... harm[s] resulting from ... repetition ... if ... the repetition was reasonably to be expected.”).

  225. [225]. Pritchard v. Van Nes, 2016 CanLII 686 (Can. B.C. S.C.).

  226. [226]. Id. at paras. 2–3.

  227. [227]. Id. at para. 78 (quoting Raymond E. Brown, The Law of Defamation in Canada 349 (2d ed. 1994)).

  228. [228]. Id. at para. 83.

  229. [229]. Id. at para. 84.

  230. [230]. Id.

  231. [231]. Id. at para. 87.

  232. [232]. PCA 1239/19 Shaul v. Nidaily Commc’n, Ltd., para. 70 Nevo Legal Database (Jan. 1, 2020) (Isr.).

  233. [233]. Ronen Perry & Elena Kantorowicz-Reznichenko, Income-Dependent Punitive Damages, 95 Wash. U. L. Rev. 835, 846 (2018).

  234. [234]. Liability can be greater than harm only in the case of fault-based liability, with no risk of application error, where the potential injurer can avoid liability altogether by not being at fault.Liability that exceeds the harm caused under a perfectly applied negligence rule would not lead to overdeterrence. Alan D. Miller & Ronen Perry, The Reasonable Person, 87 N.Y.U. L. Rev. 323, 339 n.71 (2012) (citing Robert Cooter & Ariel Porat, Should Courts Deduct Nonlegal Sanctions from Damages?, 30 J. Legal Stud. 401, 403, 405–08, 412–13, 415–16 (2001)).

  235. [235]. Perry & Kantorowicz-Reznichenko, supra note 233, at 846; Miller & Perry, supra note 234, at 346; see also Robert D. Cooter, Three Effects of Social Norms on Law: Expression, Deterrence, and Internalization, 79 Or. L. Rev. 1, 16 (2000) (providing examples of when the potential injurer will internalize costs); John C.P. Goldberg, Twentieth-Century Tort Theory, 91 Geo. L.J. 513, 545–46 (2003) (providing a hypothetical illustrating when the potential injurer will internalize costs); Keith N. Hylton, Punitive Damages and the Economic Theory of Penalties, 87 Geo. L.J. 421, 421 (1998) (defining “optimal deterrence”); A. Mitchell Polinsky & Steven Shavell, Punitive Damages: An Economic Analysis, 111 Harv. L. Rev. 870, 873 (1998) (explaining how to achieve appropriate deterrence).

  236. [236]. Perry & Kantorowicz-Reznichenko, supra note 233, at 846.

  237. [237]. See J. Shahar Dillbary, Tortfest, 80 U. Chi. L. Rev. 953, 954, 960 (2013); Alon Harel & Assaf Jacob, An Economic Rationale for the Legal Treatment of Omissions in Tort Law: The Principle of Salience, 3 Theoretical Inquiries L. 413, 422 (2002); Assaf Jacob, Dilution of Liability and Multiple Tortfeasors in the Context of Liability for Unrequested Precautions, 108 Mich. L. Rev. First Impressions 12, 13 (2009); William M. Landes & Richard A. Posner, Joint and Multiple Tortfeasors: An Economic Analysis, 9 J. Legal Stud. 517, 526–28 (1980); Saul Levmore, Waiting for Rescue: An Essay on the Evolution and Incentive Structure of the Law of Affirmative Obligations, 72 Va. L. Rev. 879, 937–38 (1986).

  238. [238]. See Dillbary, supra note 237, at 954–55 (providing a similar example).

  239. [239]. Id. at 955, 958, 960; Harel & Jacob, supra note 237, at 414, 422, 429–30, 432.

  240. [240]. Harel & Jacob, supra note 237, at 434–37.

  241. [241]. Dillbary, supra note 237, at 955.

  242. [242]. Harel & Jacob, supra note 237, at 414–15.

  243. [243]. “Under ... [the] comparative negligence [doctrine], courts ... apportion damages [between the] parties in the proportion that their negligence contributed to the accident.” See Robert D. Cooter & Thomas S. Ulen, An Economic Case for Comparative Negligence, 61 N.Y.U. L. Rev. 1067, 1075–79 (1986) (presenting the different versions of this doctrine).

  244. [244]. Guido Calabresi, The Costs of Accidents: A Legal and Economic Analysis 158 (1970).

  245. [245]. Under the contributory negligence doctrine, “the plaintiff’s negligence, however slight,” completely bars recovery. See Cooter & Ulen, supra note 243, at 1072–74 (discussing how the contributory negligence doctrine has fallen in and out of favor).

  246. [246]. Restatement (Third) of Torts: Apportionment of Liab. § 22 (Am. L. Inst. 2000) (“When two or more persons are ... liable for the same harm and one of them discharges the liability of another ... the person discharging the liability is entitled to recover indemnity in the amount paid to the plaintiff [under certain circumstances] ....”).

  247. [247]. Dillbary, supra note 237, at 962 (quoting J. Shahar Dillbary, Apportioning Liability Behind a Veil of Uncertainty, 62 Hastings L.J. 1729, 1756–69 (2011)).

  248. [248]. Dillbary also shows that if A can prevent the harm for $60 and B can prevent it for $10, and the harm is apportioned between them 75%–25%, “both will take precaution[s]” (because $75>$60 and $25>$10), leading to over-investment in precautions. Id. at 962–63; see also Harel & Jacob, supra note 237, at 422, 430 (discussing the possibility of “over-investment in precautions”). This is essentially a coordination problem: If the potential injurers can decide among themselves which one will bear the entire cost, the problem will be solved.

  249. [249]. William M. Landes & Richard A. Posner, The Economic Structure of Tort Law 198–201 (1987).

  250. [250]. Harel & Jacob, supra note 237, at 447.

  251. [251]. Tort law distinguishes between “misfeasance,” namely taking affirmative action that harms others, and “nonfeasance,” namely failing to prevent harm by passive inaction. Traditionally, pure omissions could not lead to liability. W. Page Keeton, Dan B. Dobbs, Robert E. Keeton & David G. Owen, Prosser and Keeton on the Law of Torts § 56 (W. Page Keeton ed., 5th ed. 1984).

  252. [252]. Harel & Jacob, supra note 237, at 428–29.

  253. [253]. Id. at 428–30.

  254. [254]. Id. at 415–16, 424, 426. Harel and Jacob also contend that where the harm is caused by a combination of an act and omissions, liability is imposed only on those who act because they are usually a few compared to those who fail to act. Id. at 423–24.

    The . . . exemption from liability for omissions is a way of carving a simple, practical rule to distinguish between . . . cases in which an agent can be easily selected and provided with sufficient incentives (typically, cases of acts) and cases in which . . . a . . . [liability dilution problem exists] (typically, cases of omissions).


  255. [255]. See Restatement (Second) of Torts § 324 (Am. L. Inst. 1965) (“One who, being under no duty to do so, takes charge of another ... [is liable] to the other for any bodily harm caused to him by ... the failure of the actor to exercise reasonable care to secure the safety of the other ....”).

  256. [256]. See, e.g., id. § 314A (imposing a duty on common carriers, innkeepers, and possessors of premises open to the public, “to protect [clients] against unreasonable risk[s] of physical harm” created by third parties); Harel & Jacob, supra note 237, at 426 (discussing examples).

  257. [257]. See, e.g., Pacht v. Morris, 489 P.2d 29, 31–32 (Ariz. 1971) (holding that a driver who struck a horse and failed to move it or notify a police officer he encountered on the way, was liable for injuries sustained by the plaintiffs whose vehicle collided with the dead horse); Harel & Jacob, supra note 237, at 437 (discussing this category of cases).

  258. [258]. See Harel & Jacob, supra note 237, at 429, 432, 436. Harel and Jacob also mention cases of “natural salience,” where there is only one potential rescuer. Id. at 441–42. For example, the case of “render[ing] assistance to any individual found at sea in danger of being lost.” Id. (quoting 46 U.S.C. § 2304 (1994)).

  259. [259]. Id. at 426, 437.

  260. [260]. Id. at 426, 429.

  261. [261]. Id. at 429.

  262. [262]. Dillbary, supra note 237, at 958, 960; Harel & Jacob, supra note 237, at 420, 451.

  263. [263]. See supra note 56 and accompanying text.

  264. [264]. Dillbary, supra note 237, at 959, 965, 997.

  265. [265]. Harel & Jacob, supra note 237, at 430.

  266. [266]. Dillbary, supra note 237, at 997.

  267. [267]. Id. at 997–98.

  268. [268]. Perry & Kantorowicz-Reznichenko, supra note 233, at 849 (discussing illicit gains in economic analysis).

  269. [269]. Dillbary, supra note 237, at 999.

  270. [270]. Id. at 1002 (emphasis omitted).

  271. [271]. See also id. at 1002–03 (discussing a concrete numerical example).

  272. [272]. Richard W. Wright, Allocating Liability Among Multiple Responsible Causes: A Principled Defense of Joint and Several Liability for Actual Harm and Risk Exposure, 21 U.C. Davis L. Rev. 1141, 1187 & nn.157­–58 (1988).

  273. [273]. See Dillbary, supra note 237, at 1003 (explaining that “since the damage is constant
    ... , liability will be diluted as the number of tortfeasors increases,” and that “[i]f enough tortfeasors join the activity, no one will take precaution”).

  274. [274]. Id. at 995 (presenting a similar numerical example without explaining how the incentives to take precautions are affected).

  275. [275]. N.Y. Times, Facebook, [].

  276. [276]. N.Y. Times (@nytimes), Twitter, [

  277. [277]. See supra Section II.B.

  278. [278]. The administrative cost of an action against a republisher may decrease to some extent as the number of claims increases due to factual and legal similarities between the claims; but it will still be significant, and the aggregate cost of all claims will be much larger than the cost of a single action.

  279. [279]. The exceptions and caveats are discussed in Part IV infra.

  280. [280]. See supra Section III.A.2.

  281. [281]. Perry & Zarsky, Who Should Be Liable?, supra note 6, at 165–66.

  282. [282]. See id. at 165–66, 168, 171, 173 (discussing the availability and nature of de-anonymization processes in different jurisdictions); Perry & Zarsky, Online Anonymous Speech, supra note 6, at 214–16, 218–19, 222–23, 227­­–28 (same).

  283. [283]. Perry & Zarsky, Who Should Be Liable?, supra note 6, at 166.

  284. [284]. Id. at 166 & n.23 (“See Doug Lichtman and Eric Posner, Holding Internet Service Providers Accountable, 14 S Ct Econ Rev 221, 234 (2006) (explaining that sophisticated wrongdoers can ‘conceal their tracks by routing messages through a convoluted path that is difficult for authorities to uncover’); Raymond Placid & Judy Wynekoop, Tracking Down Anonymous Internet Abusers: Who Is John Doe?, 85 Fla Bar J 38, 39 (2011) (discussing [tools] ... that can mask anonymous [user]s’ IP addresses). In the related context of online anonymous copyright infringement, a federal district court explicitly admitted that ‘the technology that enables [wrongdoing] has outpaced technology that prevents it.’ Hard Drive Productions, Inc v. Does 1-90, 2012 WL 1094653, *7 (ND Cal) (denying a discovery request to identify anonymous online users in a copyright infringement case).” (fourth alteration in original)).

  285. [285]. Id. at 166 & n.24 (“In fact, this was one of the reasons for denying a John Doe subpoena in the copyright infringement case of VPR Internationale v. Does 1-1017, 2011 WL 8179128, *2 (CD Ill) (‘The list of IP addresses attached to VPR’s complaint suggests, in at least some instances, a similar disconnect between IP subscriber and copyright infringer. The ISPs include a number of universities, such as Carnegie Mellon, Columbia, and the University of Minnesota, as well as corporations and utility companies.’).”).

  286. [286]. Id. at 166 & n.25 (“See, for example, Carolyn Thompson, Bizarre Pornography Raid Underscores Wi-Fi Privacy Risks, (NBC News, Apr 24, 2011), archived at (describing cases in which homeowners were initially accused by federal agents for downloading child pornography but it later came to light that other parties had connected to the homeowners’ wireless routers to commit these offenses).”).

  287. [287]. Id. at 166 & n.26 (“Zeran, 129 F3d at 329 n 1. The cost of information retention is correlated with the amount of daily traffic and the required duration of retention. More importantly, retention laws should not infringe basic rights. On April 8, 2014, the European Court of Justice held that the EU Data Retention Directive, Directive 2006/24/EC, which required telecom companies to store user data for up to two years, was invalid because it infringed on the right to privacy and the right to the protection of personal data. Digital Rights Ireland Ltd v Minister for Communications, Marine & Natural Resources, Case C-293/12, 2014 ECJ CELEX LEXIS 238, *19–20 (Court of Justice 2014).”).

  288. [288]. Id. at 166–67 (“For example, the Supreme Court of Virginia recently examined the ‘territorial limits of [its] subpoena power.’ It vacated a John Doe subpoena issued at the request of a Virginia carpet-cleaning business to a California-based business-rating website (Yelp), which published anonymous users’ negative reviews of the plaintiff, because the statements were published outside its jurisdiction.” (quoting Yelp, Inc. v. Hadeed Carpet Cleaning, Inc., 770 S.E.2d 440, 444 (Va. 2015) (alteration in original) (footnote omitted))).

  289. [289]. Id. at 167. For example, the plaintiff might need to request disclosure in a foreign forum and under its laws.

  290. [290]. Id.

  291. [291]. Id.

  292. [292]. Id.; Perry & Zarsky, Online Anonymous Speech, supra note 6, at 232.

  293. [293]. See Barrett v. Rosenthal, 146 P.3d 510, 529 (Cal. 2006).

  294. [294]. Id. at 529–30 (Moreno, J., concurring).

  295. [295]. See Kyle D. Logue, Solving the Judgment-Proof Problem, 72 Tex. L. Rev. 1375, 1375, 1379 (1994); S. Shavell, The Judgment Proof Problem, 6 Int’l Rev. L. & Econ. 45, 45–46 (1986); Comment, The Case of the Disappearing Defendant: An Economic Analysis, 132 U. Pa. L. Rev. 145, 157–59 (1983).

  296. [296]. Doug Lichtman & Eric Posner, Holding Internet Service Providers Accountable, 14 Sup. Ct. Econ. Rev. 221, 230 (2006).

  297. [297]. See Int’l Telecomm. Union, Measuring Digital Development: Facts and Figures 2019, at 2 (2019), [] (“In developed countries, most people are online, with close to 87 per cent of individuals using the Internet.”).

  298. [298]. Assaf Hamdani,Who’s Liable for Cyberwrongs?,87 Cornell L. Rev. 901, 910–11 (2002).

  299. [299]. Lichtman & Posner, supra note 296, at 234.

  300. [300]. Cf. Lidsky, supra note 156, at 868–72 (explaining that ISP liability secures fuller compensation to victims of online defamation).

  301. [301]. Perry & Zarsky, Online Anonymous Speech, supra note 6, at 234–36 (discussing possible solutions to the problem of judgment-proof defendants).

  302. [302]. Id. at 234.

  303. [303]. Id. at 234, 237–50.

  304. [304]. See supra Section II.B.2.ii.

  305. [305]. See supra notes 131–33 and accompanying text.

  306. [306]. See supra note 1 and accompanying text.

  307. [307]. See, e.g., 18 U.S.C. § 2511(1)(c)–(d) (2018) (providing that one who intentionally discloses to another or uses “the contents of any wire, oral, or electronic communication, knowing or having reason to know that the information was obtained through [illegal] interception” commits an offense).

  308. [308]. Bartnicki v. Vopper, 532 U.S. 514, 535 (2001) (“[A] stranger’s illegal conduct does not suffice to remove the First Amendment shield from speech about a matter of public concern.”).

  309. [309]. Batzel v. Smith, 333 F.3d 1018, 1030–31 (9th Cir. 2003).

  310. [310]. Barrett v. Rosenthal, 146 P.3d 510, 515 (Cal. 2006).

  311. [311]. In the relatively recent case of Glennon v. Rosenblum, 325 F. Supp. 3d 1255, 1261 (N.D. Ala. 2018), the defendant posted a false defamatory story about the plaintiff on a website dedicated to the shaming of adulterous women, using a picture that the plaintiff owned and used to promote her business. The story attracted a substantial amount of attention online, and later through mainstream media. Id. at 1262. The main cause of action used by the plaintiff was copyright infringement. Id. at 1263–65. Scholars similarly argued that each publication of Jennifer Lawrence’s illegally obtained “selfies” constituted a copyright violation. Citron, supra note 1.

  312. [312]. 47 U.S.C. § 230(e)(2) (2018) (“Nothing in this section shall be construed to limit or expand any law pertaining to intellectual property.”).


Professor of Law and Director, Aptowitzer Center for the Study of Risk, Liability, and Insurance, University of Haifa.

I am grateful to John Goldberg and Alon Harel for valuable comments on an earlier draft, and to the editors of the Iowa Law Review for their exceptional editorial work.