Skip to Main Content
The Liability of Internet Intermediaries The Liability of Internet Intermediaries


The Liability of Internet Intermediaries The Liability of Internet Intermediaries

The scope of liability for defamation 8.04


The publication requirement 8.04


Publication by offline intermediaries 8.08


Joint tortfeasorship 8.29


Application to internet intermediaries 8.37


Platforms 8.39


Hosts 8.62


ISPs 8.84


Gateways 8.96


Summary 8.102


Limitations upon secondary liability 8.109


Innocent dissemination 8.112


Safe harbours 8.125


Exhaustion 8.127


Voluntary disclosure 8.134


Abuse of process 8.146


Freedom of expression 8.149


Alternative remedies 8.153

8.01 Fuelled by the scale, ubiquity, and persistence of internet content, and by the public’s appetite for scandal and intrigue, online defamatory statements possess unparalleled potential to harm a person’s reputation indelibly. Necessarily, when such statements are promulgated, it is an internet intermediary which processes, stores, and transmits them. This chapter considers when such a party is liable to pay damages to the claimant.

8.02 This question is, in one sense, far from new: similar concerns have attended the arrival in England of less advanced tools for mass communication since at least 1476.1 Partly in response to those concerns, tort law has long recognised the potential for innocent messengers, distributors, and other secondary parties to face monetary liability for defamation on the basis that, although they are not the primary authors of defamatory material, they have ‘taken part’ in its publication. Simultaneously, the concept of publication has developed built-in limits which exonerate services that supply only the means or facilities of publication, while various statutory enactments have further limited secondary liability.

8.03 This chapter does not attempt an exhaustive study of the English law of defamation or its history, but instead analyses the development and application of doctrines of secondary liability in disputes involving internet intermediaries. Section 1 traces the evolution of ‘publication’ in actions involving offline intermediaries and considers the outer limits of secondary and joint liability for defamation. Section 2 analyses recent actions against internet intermediaries, concluding that prima facie liability now depends upon actual or implied authorisation, approval, or acquiescence in the defamatory words. Section 3 considers statutory limitations upon the liability of intermediaries, the general effect of which is to encourage complainants to resolve disputes directly with the author, editor, or primary publisher of the impugned material. Section 4 considers complementary non-monetary orders such as identity disclosure, alternative dispute resolution, and discursive remedies. Safe harbours are considered separately in chapter 12.

8.04  Elements. An essential element of any claim in defamation is that the defendant has published or participated in the publication of defamatory material.2 This requires the claimant to prove two matters: first, that the material was communicated to at least one other person; and, second, that the defendant was in law responsible for the act of communication.3

8.05  Means of communication. Communication can occur using any means through which it is possible to impart information, whether in a conventional analogue medium—such as speech or printed text—or an electronic one—such as a computer network or broadcast.4 The existence of the communication requirement is consistent with the modern view that the gist of defamation lies not in the mere saying or writing of words, but in their actual conveyance to others, in whose eyes the claimant’s reputation is lowered.5

8.06  Responsibility for publication. The traditional basis of liability for defamation is that each person who ‘takes part’ in a publication of defamatory material is, subject to any available defences, a primary wrongdoer and liable to compensate the claimant, regardless of the degree of participation.6 In well-settled areas—for example, communications by the author,7 editor,8 trade publisher, or printer9 of material—publication is rarely in doubt. However, judicial definitions of ‘taking part’ provide very limited guidance in determining whether a party is responsible for communicating defamatory material in borderline cases, instead offering only vague conclusions, frequent contradictions, and an unhelpful tendency towards metaphor and simile.

8.07 For this reason, adducing clear indicia of secondary publication is a task of considerable difficulty. The following sections consider cases involving offline intermediaries, which offer a source of guidance and possible analogies when considering the problems raised by internet publication.

8.08  Incremental development. At common law, the liability of real-world service providers has closely paralleled judicial refinement of the classes of person regarded as publishers. This process has, in turn, been informed by a wide range of historical and political factors, from the suppression of duelling10 and the regulation of political and religious speech,11 to jurisdictional competition between common law, Star Chamber, and the ecclesiastical courts.12 In broad terms, the boundaries of ‘publication’ have corresponded at various times to prevailing social and political policies relating to the dissemination of information.

8.09  Changes in dissemination technology. Although a comprehensive account of these factors is beyond the scope of this work, it proves instructive to consider two aspects of this history: first, the striking manner in which the concept of publication has accommodated rapid changes in the technology used to print and disseminate defamatory material; and second, the development of a fault requirement for secondary disseminators. This section discusses four classes of offline intermediaries: (1) operators of printing presses; (2) innocent disseminators of printed libels; (3) owners of property on which defamatory matter is exhibited; and (4) postal and telegraphy services. These cases are the source of most analogies used in internet defamation cases.

8.10  Seditious libel. The difficulty of acquiring and concealing early printing presses made them natural targets for the suppression of unwanted political speech.13 Criminal prosecutions for seditious libel provided one of several options for regulating gatekeepers of printing technology.14 For example, R v Knightly was a Star Chamber prosecution against a Member of Parliament who had allowed reformists to use his hidden press for the purpose of printing libellous pamphlets.15

8.11  Emergence of strict liability. At this time, both the operator and proprietor of the press were strictly liable for the resulting publications, regardless of whether they had seen, produced, or authorised their contents. R v Clerk provides an extreme example, where a printer’s servant was convicted of criminal libel, though he could not read what was being printed and his only involvement was to ‘clap down’ the press at the direction of his employer.16 His ignorance and limited contributions were ‘entirely immaterial’, since to publish did not require malice—merely that the defendant caused printing to occur. Political expediency, in short, demanded absolute liability.

8.12 The scope of publication mirrored advances in printing technology and practice. In Baldwin v Elphinston,17 the act of printing a libel onto paper was considered evidence of publication, since the defamatory material was delivered to the print compositor and workmen who operated the presses. A century later, technology had evolved to the point where secondary parties no longer needed to be involved, and the inference was abandoned in Watts v Fraser.18 These changes made it increasingly difficult to prove publication, and seditious libel—like the systems of censorship, taxation, and print licensing that preceded it—became ineffective as a means of controlling private printing. The combined effect of unfavourable precedent, changes in public opinion, and wider access to printing presses was to force governments to target intermediaries in other ways, as previous regulatory strategies ‘became inadequate, defunct, or otherwise obsolete.’19

8.13  Joint liability of printers. Both printers and proprietors of newspapers were strictly liable for publishing others’ defamatory material.20 In Zenger’s Case, Lancey CJ justified secondary liability by conjuring the spectre of an illiterate intermediary, who could be employed ‘to publish the most virulent papers with the greatest security’.21 Later, in Dixon v Enoch,22 the Court accepted that separate civil actions might lie against the printer, trade publisher, and proprietor of the Pall Mall Gazette regardless of their knowledge of the defamatory matter it contained.

8.14  Development of fault-based liability. Circulators and distributors of defamatory material may also be publishers. Early cases imposed strict liability. R v Harris, for example, records the conviction of a bookseller at Guildhall for selling a blasphemous book, despite his ignorance of its contents.23 By the close of the nineteenth century, distribution liability had undergone a dramatic transformation: it required fault. The seminal case was Emmens v Pottle,24 in which the defendant newsvendors were circulators of a newspaper which defamed the claimant. The jury found the defendants blameless, since they were neither aware of the libels nor negligent in their ignorance.

8.15  Reasoning in Emmens v Pottle. Dismissing the claimant’s appeal, Lord Esher MR explained that a secondary disseminator, although prima facie liable as a publisher, will be exonerated if it exercises reasonable care yet lacks knowledge of the defamatory character of material. This reflected two differences between primary and secondary disseminators: first, the original author, printer, or proprietor brought the libel into existence and thereby assumed some responsibility for it (whereas the same was not necessarily true of secondary disseminators); and, second, extending strict liability to disseminators would carry oppressive social consequences, since

the result would be that every common carrier who carries a newspaper which contains a libel would be liable for it, even if the paper were one of which every man in England would say that it was not likely to contain a libel. To my mind the mere statement of such a result shews that the proposition from which it flows is unreasonable and unjust.25

8.16  Negligent publication by intermediaries. The meaning of negligent publication was considered further in McLeod v St Aubyn,26 where the alleged publication was the appellant handing to a librarian an unread copy of a newspaper in which certain scandalous material appeared. The Privy Council accepted that the appellant was not a publisher of the newspaper but a ‘mere agent and correspondent of it’.27 Failing to read the newspaper before handing it on was not negligent because the appellant, unlike a printer or publisher, ‘never intended to publish’ the material. Although this was a criminal prosecution for contempt of court, McLeod is consistent with the general principle that fault is required to publish words (unlike for their composition): if a defamatory statement is communicated accidentally or inadvertently by a party, there will not be publication.28

8.17  Supermarkets. Similarly, in Bottomley v F W Woolworth & Co Ltd, a supermarket was not a secondary publisher of articles in an American magazine which it distributed on consignment.29 Although the supermarket did not employ anyone to read what it sold, there was no reason to suspect that the magazine contained libels and its dissemination was not negligent. The same result might be expected against a retailer or, it is suggested, an online marketplace.

8.18  Libraries. Conversely, in Vizetelly v Mudie’s Select Library Ltd,30 the proprietors of a circulating library that lent out a defamatory book were liable for publishing that material. Smith LJ held that the defendants’ ignorance of the libel was for want of care: the library did not employ anyone to examine books, did not take heed of publishers’ circulars, and had admitted that it was cheaper to be sued for libels than examine each book.31 This outcome is difficult to reconcile with McLeod, Emmens, and Woolworth,32 where the disseminators were held to much lower standards of care.

8.19  News agencies. The twentieth century saw growing reluctance to impose liability on commercial distributors. This reluctance is particularly evident—though ultimately not triumphant—in Goldsmith v Sperrings Ltd,33 where a prominent businessman alleged that he had been defamed by three articles appearing in Private Eye. In addition to suing the proprietor, editor, and wholesale distributor of the magazine, he commenced proceedings against 37 secondary distributors, including the defendant, a Southampton news agency. It was not alleged that they had done anything wrong other than distribute Private Eye. The majority adopted the traditional view, which went undisputed by counsel, that distributors of the magazine were prima facie liable, subject to the defence of innocent dissemination. On this basis, their Lordships refused to strike out the proceeding as an abuse of process.

8.20  Onus of proof. In a strong dissent, Lord Denning MR argued that the onus of proof should be reversed, so that distributors are not liable unless the claimant proves they have actual or constructive notice of the libel. Although Private Eye had a reputation for controversy, this was not enough to fix the defendant with knowledge of the likelihood of a libel. Disseminators of newspapers

are nothing more than conduit pipes in the channel of distribution. They have nothing whatever to do with the contents. They do not read them . . . Common sense and fairness require that no subordinate distributor . . . should be held liable . . . unless he knew or ought to have known that the newspaper . . . contained a libel.34

8.21 This approach alludes to Lord Sumner’s metaphor of ‘mere conduits’. Even prima facie liability, it was said, would risk stifling freedom of the press and its ‘channels of distribution’.35 Lord Denning’s appeal to principles of fairness reflects an implicit claim that carriers who are oblivious to what they are charged with carrying and incapable of knowing otherwise are blameless for going about their business. Without a reasonable possibility to detect wrongdoing, the carrier could not avert it and would probably not be a least-cost avoider.36 This is especially true of a tort which does not require intent to defame or knowledge of a statement’s falsity for liability to be imposed: a carrier could otherwise face liability despite reasonably believing a statement to be true or justified.

8.22  Acquiescence to exhibition of defamatory matter. Liability can arise simply by exhibiting defamatory matter, whether or not it was authored by the exhibitor.37Byrne v Deane provides the well-known example of a golf club proprietor who was liable for publishing a defamatory poster which an unknown patron had affixed to its wall.38 The Court reasoned that publication could be inferred from the club’s failure to remove the material where it: (1) had actual knowledge of the material; (2) exerted control over the material (in the sense of the ability and authority to remove it); and (3) knew that the material would, if not removed, be communicated to patrons of the club who saw it. In those circumstances, Greene LJ held that:

The proper inference . . . is that [the proprietors] were consenting parties to its continued presence on the spot where it had been put up. That being so . . . they must be taken to have consented to its publication to each member who saw it.39

8.23  The reasoning inByrne v Deane. Liability in Byrne was founded on an inference of ‘consent to publication’ arising from the defendants’ culpable acquiescence in the libel. In that case, the inference arose from the exhibiting intermediary’s exercise of control over the facilities and premises through which publication occurred, coupled with its awareness of that publication and failure to remove it within a reasonable period. Greene LJ explained the defendants’ control in the following terms:

the affixing of this notice to the walls of the defendants’ property was in fact a trespass, and they were entitled as proprietors to remove the trespassing article from the walls. It was a matter which fell right outside the rules of the club; it was not authorized and, therefore, it appears to me that they had ample power, notwithstanding the position and rights of the committee, to remove something from their property the presence of which could not be justified under the rules.40

8.24  Basis for inferring consent to publication. Care should be taken before the Byrne indicia are elevated to the status of universal principles. The overall question to be answered is: has the intermediary expressly or impliedly consented to publication of the defamatory matter? If other conduct of the intermediary is inconsistent with there being consent to publication, then it would be arguable that knowledge and control might not be enough to impose liability. The indicia articulated in Byrne were simply factors from which such consent could be inferred on the facts of that case, and may not be determinative in other cases.

8.25  Authorisation of publication. To the extent that the conduct of the club proprietor in Byrne was also an example of ‘taking part’ in publication, this suggests that participation is strongly related to an intermediary’s knowledge, control, and foresight of wrongdoing.41 That approach is difficult to reconcile with the traditional requirement that publication be wilful, which arguably precludes liability for publications by omission. Such cases are therefore better understood as instances of secondary liability by authorisation.42

8.26  Presumption of publication. The traditional view is that parties who deliver defamatory articles are prima facie liable for publications to the recipient. However, they escape liability upon proof that at all times they lacked knowledge that the contents were defamatory. In Day v Bream, the porter of a London coach office was not liable for delivering bundles of defamatory handbills to inhabitants of the building, because the defendant could ‘shew his ignorance of the contents’.43

8.27  Rebutting the presumption. Patteson J accepted that delivery raised a prima facie presumption of publication, but this was rebutted where the defendant conveyed the parcels innocently, ‘in the course of his business without any knowledge of their contents’. It is unclear whether ‘the contents’ of handbills refers to the words used, their defamatory character, or their actionability. Mitchell takes a broad view, arguing that ‘knowledge that his action would injure the claimant’s reputation was crucial’;44 on this view, the disseminator would escape liability without at least awareness that the words used were defamatory. Arguably, this does not go far enough, since it fails to consider the availability of defences which render a defamatory statement justified; however, because it will ordinarily be impossible for a delivery agent to read, much less assess the veracity of words contained in a package, the effect will often be the same.

8.28  Innocent dissemination. Like other qualifications upon dissemination liability, the result in Day v Bream sits uncomfortably with the traditional view that the test for publication does not depend on the defendant’s mental state. If communication is constituted by delivery, it is difficult to see how the deliverer’s mental state can transform that act into something other than a publication. Day v Bream is therefore better characterised as an early example of the defence of innocent dissemination.45

8.29  Secondary liability. Just as a defamation may be carried out jointly by A and B as co-authors or joint distributors,46 it may also be perpetrated where A procures or joins in a common design with B to publish defamatory material. Under the general principles of joint tortfeasorship examined in chapter 5, A will be liable despite not herself communicating the material. This liability is secondary rather than primary because it derives from, and intrinsically depends upon, B’s independent wrong of defamation by primary publication.

8.30  Relationship between primary and secondary liability. While there are no conceptual barriers to applying this approach in defamation cases, in practice publication has proved sufficiently flexible that secondary liability rules have normally been unnecessary to catch procurers, conspirators, and the like. Instead, there are deemed to be a series of publications effected by parties who, while acting jointly, also ‘take part’ as primary common law publishers.47

8.31  Origins. Some nineteenth-century cases refer to the liability of one who authorises another to publish defamatory material.48 An authoriser is said to be a publisher of the material, which suggests that her liability is primary. However, because the authoriser does not herself engage in the act of communication, this is properly characterised as an example of secondary liability deriving from the authorised party’s primary act of communication.

8.32  Modern recognition. Although the doctrinal origins of authorisation liability for defamation are relatively straightforward, the concept has not been widely applied by modern courts. This introduces a degree of uncertainty as to whether this form of secondary liability will be recognised and applied to internet intermediaries, and precisely how its elements should be articulated. Nevertheless, it is submitted that the doctrine follows from ordinary principles of tortious secondary liability.

8.33  Elements. The cases considering authorisation liability indicate that for a secondary party (A) to be jointly liable for defamation with a primary publisher (B), four elements must be satisfied: (1) A must have approved of a publication by B and furnished material in support of it; (2) B’s publication of the material must be defamatory (and unjustified); (3) the material published by B must be the same, or substantially the same, as that approved by A.

8.34  Approval of published matter. In R v Cooper, the defendant was liable for approving a defamatory article published by a newspaper, where there was ‘substantial identity’ between what he authorised and what the newspaper later published.49 The test was simply whether the defendant ‘approved of’ something which was defamatory and furnished the necessary material.

8.35  Implied approval. Likewise, Byrne’s emphasis on ‘consent’ to a publication may be characterised as an example of secondary liability arising from an implied approval of the published material, where the approval is inferred from silence coupled with knowledge and control. More recently, Tugendhat J considered it a live issue whether employees who had spoken allegations to the defendant had thereby ‘authorised’ him to publish the material to a journalist.50

8.36  Requests to publish. Similarly, a request to publish defamatory material can be sufficient for liability, but not a ‘mere wish or hope’—the purpose being to prevent a party from deploying ‘the safe shelter of intermediate agents.’51 Again, however, these cases are better treated as examples of joint tortfeasance, since an explicit request may be procurement. Ultimately, the absence of modern cases on secondary liability suggests that it is an unnecessary gloss on publication, which is already sufficiently broad to encompass anyone who intentionally takes part by a request, permission, or other material contributory act.

8.37  Overview. Defamation litigation against internet intermediaries has centred on two main issues: first, whether the defendant is a ‘publisher’ of defamatory material authored by a third party but cached, stored, or transmitted using its services; and second, whether the defendant’s activities attract the operation of a defence, statutory immunity, or limitation.

8.38 This section addresses the first question in light of how the English authorities have refined the concept of publication in internet defamation actions. It commences by examining claims against platform operators, before moving to parties further removed, such as hosts, ISPs, and gateways.52 It identifies several emerging limits to the scope of publication which fully insulate network-layer intermediaries from liability and provide partial protection to application-layer services. The issue of defences is considered in section 3, while safe harbours and other limitations on injunctive remedies are discussed in chapters 12 and 13.

8.39  Primary and secondary publication. Considering the volume of internet material published online, surprisingly few cases address platforms’ liability for third party statements. The majority of claims in this category instead concern material originating from website operators themselves—frequently journalists or bloggers—or their agent or employee. These claims do not concern us here because they are determinable by established principles of primary and vicarious liability. Other claims raise issues of internet publication incidentally, such as where the same material is published both in print and online.53

8.40  Factual and technical context. The few claims against true secondary publishers involve user-created content posted to weblogs and discussion fora.54 The factual and technical distinctions between platforms, and their varying levels of involvement in the authorship, selection, or conveyance of defamatory matter, underscore the importance of treating publication as a question of fact that examines the conduct of each defendant in the particular circumstances of the case.

8.41  Common issues. In these cases, two issues commonly arise: whether the information has been sufficiently published at all; and whether the intermediary has sufficiently participated in publication to impose prima facie liability. A third issue relating to the republication of archived material has been largely displaced by statute.

8.42  Threshold requirements. To prove publication on a platform, the claimant must now establish that the defamatory statement has been intelligibly perceived by a substantial number of people within the jurisdiction besides the claimant.55 This extension of the traditional publication threshold does not owe its existence to any unique property of the internet, but rather to section 6 of the Human Rights Act 1998 and the need for courts to balance the claimant’s right to reputation against the defendant’s freedom of expression, alongside increasingly important policies concerning the cost and availability of judicial procedures for determining what often turn out to be insubstantial claims. This is achieved by precluding trivial claims and other disproportionate demands on judicial resources.

8.43  Proving local publication. Substantial local publication may sometimes be inferred from search engine visibility,56 but normally requires proof of access rather than the mere potential for access. In Loutchansky, Gray J refused to infer publication of a headline story without evidence of visitors, even where the website in question received 12.5 million monthly visits.57 Limited publication may be inferred to ‘followers’ of a weblog (or, by analogy, a Twitter account),58 members of a website ‘of specialist interest’,59 those who comment upon the impugned article,60 and, perhaps, RSS and social media subscribers.

8.44  Print and online media. In short, the question of website publication is answered in much the same manner as publication in print. Just as substantial publication can no longer be presumed from the mere fact that material was printed by its author, it cannot be presumed from material being made available on the internet.61 Echoing the transition from Baldwin to Watts in the nineteenth century, this reflects the growing maturity of the internet as a publication medium and the recognition that there may be dark corners which go largely unexamined by anyone.

8.45  Publication via social networks. A quartet of cases has considered publication on social networks. First, Applause Store Productions Ltd v Raphael concerned litigation between two former friends for defamatory statements made by one against the other on Facebook.62 The words complained of appeared in fake Facebook profiles which contained various libels on the claimant and his business. Although Facebook was not alleged to be a publisher, it did remove the material voluntarily and gave disclosure under a Norwich Pharmacal order.63 Publication was assumed.

8.46 Second, in Elsbury v Talbot the Court accepted that defamatory tweets posted to Twitter could be actionable publications.64

8.47 Third, Cairns v Modi awarded damages for tweets alleging match-fixing by the claimant, a well-known cricketer. Although only about 65 people had seen the messages, the Court inferred substantial publication to the defendant’s followers.65 In upholding this award, the Court of Appeal noted the tendency of social networks to make scandalous stories ‘“go viral” more widely and more quickly than ever before’. This was a legitimate factor to be taken into account when assessing both the existence and extent of liability.66

8.48 Fourth, in Richardson v Facebook UK Ltd, the claimant alleged that she had been defamed by a Facebook profile which purported to have been created by her but was in fact published by an unknown imposter.67 Facebook had twice removed the profile following earlier complaints by her. The claimant also sought disclosure of information to identify the primary publisher. Although the action was dismissed on the basis that the claimant had sued the incorrect defendant,68 Warby J added that there was no realistic prospect that the defendant would be held responsible for the publications, and would have also struck out the claim on that basis.69

8.49  Publication of user-created content. Courts have struggled to articulate coherent principles for assessing when a website operator is a publisher of defamatory material that has been authored and uploaded by users of the website. Only three English cases appear to deal with such material.70

8.50  Discussion fora. In Metropolitan International Schools Ltd v DesignTechnica Corp,71 the first defendant operated an online consumer electronics discussion forum in which postings were made by unknown third parties. It failed to remove these postings, which criticised the claimant’s distance learning courses as fraudulent. In entering default judgment, Eady J noted that his order was probably unenforceable against the American operator, but concluded that a claim lay against it in principle.72

8.51  User reviews. The second case, McGrath v Dawkins, concerned reviews and comments posted to the claimant’s book product page on The Court assumed publication in the claimant’s favour but struck out the claims against Amazon because they were defeated by the storage safe harbour.74

8.52  Postings on a multi-user weblog. Third, in Kaschke v Gray an allegedly defamatory article was contributed by the first defendant and automatically published on the homepage of a political weblog created and maintained by the second defendant, Mr Hilton. Hilton occasionally removed, edited, or ‘promoted’ posts, but did not monitor the weblog and was unaware of the offending material until notified by the claimant, whereupon he removed it.75 Hilton conceded that there had been a publication, but argued that he did not participate in it and, in any event, a defence applied. Stalden J refused to strike out the claim, but dealt with only the second of these arguments, instead assuming that Hilton was a publisher.76

8.53 It is suggested that this conclusion is far from obvious. While Hilton provided the facilities of publication and had some control over what was posted to the homepage of the website, he was not aware of the defamatory posting or any prospect that it would be conveyed to others.77 The facts are therefore readily distinguishable from Byrne. The only reasoning consistent with the Court’s finding is that his dissemination of the posting was somehow negligent in a way that might preclude reliance on Emmens and McLeod. However, Kaschke offers no indication of what the applicable standard of care might be. Paradoxically, to infer publication so readily from the capacity or historical tendency to moderate material that is not the material complained of discourages website operators from developing and using the systems best capable of removing harmful content.

8.54  Authorisation of publication by users. As an application for summary judgment, Kaschke’s refusal to rule out publication does not signal a return to strict liability. However, it creates uncertainty about the test for platform publication. Although the availability of safe harbours and defences may make the precise test a question of little practical significance, it is suggested that the cases are best understood as creating a fault-based liability rule premised upon the defendant’s implied authorisation of publication.

8.55The test for pre-notification publication. Prior to notification, the Emmens standard applies: operators could normally not be publishers if they were unaware of the words used and acted as a reasonable platform operator would—unlike, perhaps, one who sought out unreliable content, adopted reckless editorial and moderation policies, or failed to exclude users with a known history of abuse.

8.56The test for post-notification publication. Following notification, the Byrne standard applies: the question is whether by his conduct the website operator authorised, consented to, or assumed responsibility for the conveyance of the relevant words. This could be evidenced by intentional actions taken to promote or highlight the material, but it would be rarely inferred from silence without actual knowledge that the words are actionable, control over the content, and a reasonable opportunity to act.

8.57 In applying well-established principles, such an approach offers greater certainty and consistency with the test for offline publication. Additionally, it protects diligent website operators, encourages efficient investments in content moderation systems, and discourages intermediaries from soliciting or acquiescing in obviously tortious postings with impunity.

8.58 Finally, although several cases have considered defamatory statements published on social networks, none has directly considered whether the social network is itself a publisher. It is suggested that the principles considered in section 2.2 should apply by analogy, since social platforms are essentially providing a passive hosting service for users’ material. Some support for this approach may now be found in the obiter remarks of Warby J in Richardson.78

8.59  Editors of websites cataloguing defamatory statements. Where a website operator conducts himself so as to solicit or encourage defamatory statements, and then to edit and catalogue those statements on a website, he is considered an editor thereof and may be liable as if he published the statements directly. In Brett Wilson LLP v Persons Unknown, the defendants were the anonymous operators of a successor to the website ‘Solicitors from Hell’.79 The website purported to operate as a catalogue of complaints made by dissatisfied members of the public against various members of the legal professions. On a subpage listed in the ‘complaints’ section, the defendant had published a letter of complaint purportedly sent by an anonymous client of the claimant law firm, which contained various defamatory statements. In the circumstances, Warby J had little hesitation in concluding that the operators of the website were at least editors (if not authors) of the material and were therefore to be treated as falling outside section 10(1) of the Defamation Act 2013.80

8.60  Abrogation of multiple publication rule. Traditionally, English law regarded a separate cause of action as arising with each successive communication of a defamatory statement by a website.81 This position, which significantly disadvantaged intermediaries (who faced endless liability where an article was kept online in a digital archive),82 has now been abolished in the United Kingdom by statute.83

8.61  Manner of republication. The effect of the new single publication rule is that the limitation period for a cause of action for defamation will normally run from the date of the first publication. However, later publications must not be in a manner that is materially different to the first publication, or they may be treated as fresh publications. Relevant factors where a print publication is subsequently published online include the level of prominence given to the statements contained in the publication, and the extent of subsequent publication.84

8.62  Overview. Early authorities suggested that, subject to defences, hosts were prima facie publishers of material once put on actual or constructive notice of its defamatory character. However, more recent cases appear to ask whether the host is secondarily liable for authorising publication in the manner recognised in Byrne v Deane. This section argues that network-layer facilitation by a host does not amount to publication whether before or after receiving actual notice of the defamatory words, until the host’s conduct (including a failure to act) would reasonably be regarded as an assumption of responsibility for or acquiescence in the primary wrongdoing.

8.63  Ratio. Godfrey decided that a Usenet host is a publisher where: (1) material is uploaded to equipment under its control; (2) the material is ‘on its face defamatory’; (3) the host has the capacity to remove that material and fails to do so within a reasonable period; (4) having been put on notice of its defamatory character.85

8.64  Facts. The defendant (‘Demon’) operated public newsgroup servers to which third parties could post messages that were stored automatically and made accessible to other newsgroup users for a fixed period. The case arose out of a ‘squalid, obscene and defamatory’ posting made by an anonymous user who impersonated the claimant (a practice known as ‘frogery’ or ‘fraping’).86 Four days after being uploaded, the claimant notified the defendant of the posting and requested its removal. The defendant did not do so, and the post remained accessible until its scheduled expiry ten days later. The claimant only sought damages for publications occurring during this interval.

8.65  Imposition of strict liability. In the circumstances, Morland J concluded that the defendant’s continued storage and transmission of the posting amounted to publication. The defendant was treated analogously to a traditional distributor, such as a bookseller or circulating library. Because it ‘deliberately chose’ to store postings and could ‘obliterate’ them at any time, Demon’s conduct went beyond mere ownership of the storage equipment (which, by implication, would not be sufficient) and constituted prima facie publication—even without knowledge—since the material was defamatory ‘on its face’:87  

At common law liability for the publication of defamatory material was strict. There was still publication even if the publisher was ignorant of the defamatory material within the document. Once publication was established the publisher was guilty of publishing the libel unless he could establish . . . that he was an innocent disseminator.88

8.66  The analogy with Byrne v Deane. Morland J appears to have understood cases like Day v Bream and Emmens as deciding that knowledge was relevant to innocent dissemination rather than publication. The Court adapted the ‘proper inference’ test from Byrne, suggesting a further analogy between an occupier’s physical control over premises and a web host’s technical control over stored messages. However, this explanation could not justify liability before the point of notification, unless one takes the factually implausible view that a service provider assumes general responsibility for all stored messages. This approach also suffers from the basic problem that it treats publication as occurring without any positive act or authorisation by the putative publisher. It is therefore, with respect, wholly unprincipled.

8.67  Defamation ex facie. The distinction between messages defamatory ‘on their face’ and those whose meaning is less obvious is problematic in several respects. First, it is difficult to apply, requiring a host to read material and form a qualitative judgement about its meaning. This is both impracticable and ignores possible defences such as justification.

8.68 Second, where statements are published by an automatic technical process, it is nonsensical to delimit publication liability by reference to a test that presumes human examination. Manual review would rarely (perhaps never) be required in other contexts, as Day and Byrne assumed.

8.69 Finally, even if the distinction were workable, it would be inappropriate in an impersonation case such as Godfrey, where the statements appeared to originate from the claimant. Taken at face value, such statements could not be defamatory, since they would, on that assumption, have been consented to by their apparent author.89 Such statements can, of course, be highly defamatory, despite appearances90—a conclusion at odds with the approach in Godfrey.

8.70  Bunt v Tilley. Later cases have tended to distinguish or confine Godfrey. For example, in Tilley Eady J interpreted Godfrey on the narrow basis that it was concerned with a single posting of which the host had specific knowledge, before going on to propose a test with a significantly higher threshold for ISPs.91 Two other cases have directly addressed hosting liability, preferring an authorisation-based publication standard derived from Byrne.

8.71  Davison v Habeeb. In Davison v Habeeb, the fifth defendant (‘Google’) facilitated the publication and storage of defamatory articles on its hosted weblog service, an ‘enormous burgeoning Babel’ to which content was posted by its 400 million users in 2011 at a rate of some 250,000 words per minute. Google did not create, select, or approve Blogger content, but ‘merely provides the tools for users to operate their sites’.92 Google refused to delete the material without a court order on the basis that it did not know whether any of the allegations against the claimant were true.93 Despite striking out the claim as an abuse of process,94 HHJ Parkes QC reasoned that there was an arguable case that Google was a publisher of the material—at least after being notified by the claimant—because ‘at some point’ following notification it must be regarded as having consented to or acquiesced in publication, by analogy with the club operator in Byrne.

8.72  The relevance of knowledge. Like Godfrey, Davison fails to clarify whether publication requires actual knowledge. Notice was ‘of cardinal importance’, presumably reflecting the mental elements recognised in Emmens and Byrne,95 but the precise moment of publication remains unclear. It is suggested that the approach should be the same as for website operators: until the point of notice, Google was an innocent disseminator at common law, since it had exercised reasonable care despite being ignorant of the defamatory posting. After sufficient notice was given, it would no longer be an innocent disseminator, but due to its passivity nor would it necessarily be a publisher. The question is then whether Google can be inferred to have authorised or acquiesced in the publication, or otherwise by its conduct assumed responsibility for the words used; in almost all circumstances, the answer is likely to be ‘no’, but the possibility cannot be ruled out without considering the circumstances of publication and the intermediary’s response in detail.

8.73  Tamiz v Google Inc. This was the approach preferred by the Court of Appeal in Tamiz v Google Inc, where the issue was whether Google was a publisher of defamatory comments posted to another Blogger weblog.96 In that case, the Court rightly overturned the trial judge’s conclusion that Google could never be a publisher, even after receiving notice and failing to act after a reasonable time. Applying Byrne, the essential issue was whether Google could ‘be inferred to have associated itself with, or to have made itself responsible for, the continued presence of [the tortious] material on the blog’.97 This would presumably involve considering how realistic such an inference was having regard to the scale of third party publications and the relationship between Google, bloggers, and commenters. Although it seems doubtful whether such an inference is open on the facts of Tamiz (or indeed any case involving Blogger), these were essentially ‘matters for argument’ rather than summary judgment. This approach lacks the simplicity of a bright-line rule, but provides adequate certainty for intermediaries without conferring absolute immunity, which might retard genuine efforts to remove tortious content.98

8.74  Assumption of responsibility. The comparison in Davison and Tamiz between Blogger and the noticeboard in Byrne is open to two criticisms. First, unlike the defendant in Byrne, Google expressly disclaims any personal endorsement of users’ statements. In Davison, HHJ Parkes QC concluded that Google ‘appears to assume a degree of responsibility’ for content on the basis of prohibiting various classes of material in its Blogger Contents Policy. However, there is no necessary inconsistency between reserving the contractual right to remove offensive or unlawful content, and the refusal to assume a duty to remove tortious material. In any case, the policy expressly preserves Google’s right not to take action for violations,99 and users promise to bear sole responsibility for their content.100 Unless Google takes some specific action to approve or promote a specific posting, it is therefore difficult to see how merely operating the platform can be regarded as an assumption of responsibility for users’ statements generally.

8.75  Practicality of enforcement. Second, Google possesses no practicable means to investigate each complaint, determine the applicable law, and verify the accuracy and lawfulness of posted material. Google’s noticeboard is ‘almost infinitely huge’: if all Blogger postings were printed and pinned to an imaginary golf club wall, its length would stretch three times around the earth and grow by roughly the distance from London to Oxford each day, weighed down by the postings of millions.101 Even HHJ Parkes QC conceded that it is ‘unrealistic’ to expect removal of all material alleged to be defamatory. Unlike the proprietors of a small clubhouse in Byrne, it is inherently unlikely that Google ever consents to any particular posting, tortious or not.

8.76  Policy considerations. The observations of Lord Esher MR in Emmens v Pottle merit renewed attention: to impose prima facie liability upon common carriers would be ‘unreasonable and unjust’, and impose far too heavy a burden.102 However, for Google simply to acquiesce to all complaints would make ‘significant inroads into freedom of expression’, since far from all allegations are well founded.103 Yet that is exactly what Davison appears to require, since—in the absence of a scalable way to determine whether content is tortious—hosts (especially those less well resourced than Google) will inevitably err on the side of caution.

8.77  Use of metaphor. In Tamiz, Eady J accepted a less onerous metaphor, treating Google as the owner of a vast wall on which graffiti had been festooned overnight;104 while it could paint over or clean the wall, its failure to do so would not ordinarily entail consenting to the content of statements inscribed by the vandals. The fact that Google also ‘built’ the wall and failed to install surveillance cameras was no more relevant than to the liability of the unfortunate property owner.

8.78 The Court of Appeal disagreed: because Google developed the platform and intended it to be used to post messages (unlike, presumably, the owner of a freshly painted wall) on terms it specified, Blogger more closely resembled a noticeboard than a defaced wall. The Court emphasised Google’s volition in choosing to supply tools and operate a service for the dissemination of messages: in setting its terms and controlling its ultimate functionality, Google was more than a transmitting conduit, though not necessarily a participant.105

8.79 Metaphor is apt to mislead. The debate about whether Blogger is ‘like a noticeboard’ or ‘like a graffiti wall’ is ultimately less important than the assessment of its actual contribution to the publication in question. While it is true that Google’s contribution goes beyond transmission to include storage of the defamatory words, the reality is that it still supplies, at most, an automated medium for publication. Its outward relationship to material does not change after notification, except where it secures the removal of the material.

8.80  Passivity of host or platform. A neutral host of defamatory material authored by third parties remains essentially passive—indeed, it is normally this passivity, characterised by a failure to act at all, about which claimants complain. As Eady J remarked in Tamiz:

It is not easy to see that [Google’s] role, if confined to that of a provider or facilitator before [receiving notice], should be automatically expanded thereafter into that of a person who authorises or acquiesces in publication....It takes no position on the appropriateness of publication one way or the other.106

8.81  The need for participation. On this view, a necessary but insufficient condition of publication is that the intermediary plays an ‘active’ role in approving or participating in publication. A neutral and passive role, in which the words are conveyed without human intervention, will be insufficient regardless of notice. This is consistent with the principle that a defendant is not liable in tort for involuntary or unintended actions.107 The Court of Appeal seems to endorse this approach in Tamiz (CA), with the qualification that a platform might eventually cease being ‘purely passive’ at least a reasonable period after notification.108 Similarly, acquiescence might be inferred from ignoring a valid court order holding that the user’s words are actionable.

8.82  Emerging themes. Coherence and consistency in this emerging area of law demand that like services are treated alike. Godfrey is an outlier, and there are good arguments that it has already been abandoned. With the decision in Tamiz (CA), a pattern appears to be emerging which treats platform operators and hosts as non-publishers until notification and, if the inference of acquiescence is open, prima facie publishers from a reasonable period thereafter.

8.83 The basis for that inference is far from clear, and the distinction between ‘passive’ and participatory hosts remains mired in confusion and metaphor. While it would presumably be a matter of evidence, it is suggested that such an inference would rarely arise for automated services such as Blogger, which few would regard as consenting to words published by their users. To the extent they assumed otherwise, Davison and Tamiz are difficult to reconcile with the treatment of search engines as non-publishers because they are ‘entirely automatic’ services that do not rely on human intervention.109 Fortunately, the true ratio of Davison rests on abuse of process, while Tamiz (CA) recognised no more than the possibility of inferring acquiescence in an appropriate case, based on the traditional Byrne criteria.

8.84  Distinction between ISPs and hosts. Although Godfrey is sometimes improperly described as an authority on ISP liability,110 it is important to recall that that case concerned material uploaded to the defendant’s newsgroup servers and stored in a persistent forum under its operational control, whereas an ISP provides transient routing facilities to access material stored by others.111 The ephemeral nature of this role means that intermediaries who merely provide network connectivity will not, without more, be liable as publishers of defamatory material transmitted at the request of their subscribers.112

8.85  The facts in Bunt v Tilley. Bunt v Tilley is the most recent authority on ISP liability.113 Relevantly, the claimant alleged that three English ISPs had published defamatory newsgroup postings transmitting them to message boards hosted by third parties. Eady J granted an application to strike out the claims against the ISPs, concluding that they were not publishers of the defamatory material. Instead, they supplied a passive medium of communication without assuming any general responsibility for subscribers’ statements.

8.86  Reasoning. Eady J’s approach to publication was to identify the specific causal contribution of each defendant: ‘to focus on what the person did, or failed to do, in the chain of communication’.114 Unlike Morland J in Godfrey, Eady J considered that publication entails a strong mental element, consisting of ‘knowing involvement’ in the relevant words:

it is essential to demonstrate a degree of awareness or at least an assumption of general responsibility, such as has long been recognised in the context of editorial responsibility....[T]here must be knowing involvement in the process of publication of the relevant words. It is not enough that a person merely plays a passive instrumental role in the process.115

8.87 This partly explains why editors and commercial publishers are presumed from their position to intend to publish such libels, as Lord Morris reasoned in McLeod: they assume responsibility for the words that are published.116 Similarly, it might reconcile the strict liability of printers and fault-based dissemination liability recognised in Emmens and Vizetelly: the latter are only ‘knowingly involved’ in publication where they knew or ought to have known of the defamatory words. By contrast, an intermediary which supplies a ‘passive medium of communication’ is not a publisher, since it does not assume any responsibility for the messages and has no knowledge of their contents. It is, in short, no distributor of the messages at all.

8.88  Other network-layer conduits. It follows that transmitting ISPs—and other network-layer conduits—can rarely if ever be described as publishers. While expressing caution about using imperfect analogies to describe new technical phenomena, Eady J accepted Collins’ suggestion that such intermediaries are analogous to postal services and telephone carriers,117 providing ‘a means of transmitting communications without in any way participating in that process’.118

8.89  Lack of notice. On the facts of Tilley, there was no evidence that any ISP had received proper notice from the claimant or ‘sanctioned any publication with knowledge’, even though only one ISP suspended the subscriber’s connection upon learning of the proceedings. Accordingly, no defendant had participated ‘in any meaningful sense’ in publication.

8.90  The ‘knowingly involved’ test. Although the approach in Tilley correctly recognises the important technical distinctions between hosts and mere conduits, to rely on the criterion of knowing involvement raises a number of difficulties.

8.91Blurring of conduct and mental state. First, publication is concerned with the defendant’s causal contribution to a communication of defamatory words through his or her conduct; a subjective attitude towards the words being conveyed can have no bearing on whether or not that conduct has effected a publication, though it may supply an excuse for publishing as an innocent disseminator. Using knowledge to distinguish ‘passive’ from ‘active’ participants is, at best, unhelpful, because both ultimately contribute equally to the diffusion of tortious material. If, as Eady J suggested, prior awareness of defamatory material is what elevates involvement into ‘knowing involvement’, this test adds nothing. If the intermediary’s contribution depends on its mental state, this risks distorting both concepts.

8.92Ambiguous threshold of liability. Second, Tilley fails to clarify the threshold of knowing participation. While it seems uncontroversial that granting permission is sufficient while innocent transmission is not,119 the precise point at which an ISP will move from passive medium to active participant remains undefined. Because transmission is transitory, merely informing an ISP of a notorious subscriber’s propensity to post defamatory material would not necessarily be enough to make it participate in subsequent publications, which are distinct.

8.93Overlap with joint liability. Further, the relationship between ‘knowing involvement’ and joint tortfeasance is unclear. It remains an open question whether an ISP could be liable for authorising or procuring a publication without ‘knowingly participating’ in it. Equally unclear is what action must be taken by an ISP who has been notified of defamatory material; disconnection of the user would usually be a disproportionate response, but the ISP would have no power to remove material which is hosted elsewhere. At best, it could block access to its other subscribers, a possibility discussed in chapter 14. This issue did not arise in Tilley because none of the ISP defendants had received proper notice, and the point was not raised by the claimant.

8.94Encrypted communications. Finally, if participation requires acts which are directed to taking responsibility for the communication of specific words, this has the consequence that no liability can attach if it is impossible to read the defamatory statement. For example, if the words are encrypted, in a foreign language, or otherwise obfuscated, an ISP cannot be ‘knowingly’ involved in their transmission. This is the problem of words not being ‘defamatory on their face’ in a different guise.

8.95  Use of metaphor. The approach in Tilley represents a significantly higher threshold of publication than was recognised in Godfrey, and sets clear fault-based limits on the scope of primary liability. However, its analogy with postal services is unhelpful because their immunity derives from specific statutory provisions rather than general principles of publication. A better analogy is with the porter in Day v Bream, who was exonerated for innocently delivering sealed packages in the ordinary course of business. Sealed packages have a more natural counterpart in the data packets delivered by ISPs.

8.96  Search engines. Like other application-layer services, search engines which actively participate in a publication can be liable as publishers, but not those which merely facilitate it. In DesignTechnica, the claimant (in addition to suing the website operator) alleged that Google was a publisher of defamatory ‘snippets’ which appeared in the third and fourth search results for the claimant. Google refused to remove the offending material from its search index. Google argued that it was, at most, a ‘mere facilitator’ of the defamation, since it had no control over the words displayed—which were selected automatically120—and could influence neither the keywords entered by its users nor the results they chose to access. Alternatively, Google relied on the statutory defence of innocent dissemination and safe harbours.121

8.97  Reasoning in DesignTechnica. Eady J held that Google was not a publisher of the search snippets. Following Tilley, the test is ‘whether the relevant Internet intermediary was knowingly involved in the publication of the relevant words’.122 Prior to notification, Google lacked knowledge of the words appearing in the snippets; thereafter, Google had not authorised, approved, or acquiesced in their publication simply by failing to remove them. The decisive factor was the lack of ‘human input’ by Google into its indexing and search functions, which were carried out by a web-crawling ‘robot’ rather than the conscious deliberations of human agents. Their display was automatically triggered by the public’s search queries rather than Google—a fact Eady J thought was ‘fundamentally important’.123 This reasoning was endorsed by Sharp J in Budu v British Broadcasting Corporation, where a website operator was held not to be liable for publication of words in a search snippet which, when isolated from their context in a larger article, might convey a different and defamatory meaning. The Court reasoned that such snippets were produced by robots and merely served to direct internet users to the original webpage.124

8.98  Relevance of automation. The automated quality and large scale of Google’s operations meant that no authorisation or intent to publish could be presumed from a mere failure to exclude a particular snippet from the index, even though it had actual knowledge. Google therefore stood in a different position to a library whose cataloguer extracted defamatory snippets from a book. There the library must specifically select the material for inclusion and the words to be used, and can therefore be liable under the repetition rule. Google was simply a passive medium of communication, like the ISP in Tilley. Presumably, the same logic would apply where the defamatory matter is contained in a URL rather than a snippet.125

8.99  Control over published material. DesignTechnica reflects the general principle that, to be publishers, secondary disseminators must satisfy a mental element.126 Failure to act after acquiring knowledge may support an inference that the search engine authorised, approved, or acquiesced in the publication, but before such an inference can arise it must be possible and reasonable to remove the offending material.127 Google could not prevent a particular snippet from appearing in search results—which might respond to similar queries or reference mirrored copies of material—without disproportionately affecting other search results. However, this reasoning seems slightly puzzling given that Google could simply de-index the URLs complained of by the claimant—indeed, that is the approach expected in other contexts, such as data protection and copyright.128

8.100  Alternative means of control. Another operative factor may have been the relative ease of including a robots.txt file on the DesignTechnica website, which could simply instruct Google’s crawler (and those of other search engines) not to index the contents of that website.129 One potential difficulty is that, although default judgment was entered against the foreign website operator in DesignTechnica, it appeared unlikely that any order would be enforceable against it. In those circumstances, there was little likelihood of the Robots Exclusion Protocol being applied, given that neither the claimant nor Google had direct control over its use by the website operator.

8.101  Potential future liability. The difficulty of removing the defamatory content took the facts outside Byrne, where removal was as simple as removing a sheet of paper. Hosts are not like search engines.130 However, DesignTechnica leaves open the possibility that a search engine which can easily remove material may become a publisher after failing to do so within a reasonable time. This possibility has been taken up with a certain degree of enthusiasm by recent decisions in Australia, New Zealand, Hong Kong,131 and Canada.132 In A v Google New Zealand Ltd,133 the High Court of New Zealand refused to grant summary dismissal on the basis that the defendant search engine was not a common law publisher of defamatory search snippets, but struck out the action on other grounds. In Bleyer v Google Inc LLC, the Supreme Court of New South Wales followed Design Technica and Tamiz (CA) in concluding that the same result would be reached under the common law of Australia.134

8.102 Conversely, in Trkulja v Google Inc LLC (No 5) the Supreme Court of Victoria concluded that it was open to a jury to conclude that Google was a publisher and stood in an analogous position to a newsagent who sold a newspaper containing a defamatory article.135 To the extent that Trkulja imposed prima facie liability upon Google for mere operation of its indexing algorithm, and before it had actual knowledge of the defamatory publication, Trkulja is—with respect—inconsistent with persuasive authority and clearly wrong. In Duffy v Google Inc, the Supreme Court of South Australia followed Trkulja and concluded that Google was a publisher of snippets, at least where it failed to remove them within a reasonable period of notification.136 In that case, Blue J treated Google as being in the same position as a host of defamatory material; this approach is open to question, since it ignores important differences between these network and application layer intermediaries.

8.103  Enforcement role of intermediaries. These developments are heavily indebted to the liability rules developed in disputes involving offline intermediaries and previous generations of communications technology, reflecting their iterative and analogical nature as ‘indicia of the law’s capacity for growth and adaption’.137 Modern courts continue to apply them, but place greater emphasis upon fault and the availability of statutory safe harbours. The collective effect of these limitations is to resist the impetus for intermediaries to assume a role as ‘content police’ on the internet, absolving them from monetary liability if they fail pre-emptively to verify, monitor, or remove defamatory material authored by others.

8.104  Scope of publication. Several propositions may be derived from this brief survey of defamation claims against internet intermediaries. First, the scope of publication is becoming narrower. Where early authorities tended to impose prima facie liability, more recent authorities offer limited or total immunity. Application-layer platforms are unlikely to be publishers if they play a merely passive or instrumental role in conveying the defamatory words.

Similarly, network-layer intermediaries who do no more than transmit, index, or aggregate material without knowledge will not be publishers, provided that they take no position on the appropriateness of content, and act neutrally and passively. Only if the intermediary intervenes to promote or edit content, or unreasonably fails to act, could this amount to publication by authorising, approving, or acquiescing in the words used. However, some authorities in other common law jurisdictions take a stricter approach in which notification has a more decisive role.

8.105  Distinguishing publication from facilitation. Although ‘straightforward common law principles’ of publication have been adapted to conduct by internet intermediaries,138 it remains difficult to articulate a clear criterion to explain when a secondary party ‘takes part’ in publishing defamatory material, as distinct from merely facilitating or passively conveying it. The tendency to adopt different metaphors in similar cases has produced irreconcilable results and unpredictability in the principles applied (see Table 8.1). Despite a certain level of confusion, the underlying principle appears to be that non-publisher intermediaries neither assume responsibility for the defamatory words nor authorise their communication. Authorisation may be inferred from the combination of knowledge, control, and inaction in limited circumstances, while assumption of responsibility may be inferred from conduct suggesting editorial control.

Table 8.1
Outcomes in publication cases
GodfreyHost of UsenetPrima facie liability subject to EmmensOwner of noticeboardYes
TilleyRetail ISPsKnowing involvement; not mere passive rolePostal service or telephone companyNo
DesignTechnicaSearch engineHuman input; not passive mediumLibrary catalogue distinguishedNo
KaschkeWebsite operatorCapacity to monitor or editN/AYes
Host of BloggerAuthorised or acquiesced in publicationOwner of noticeboard
Owner of graffiti wall
GodfreyHost of UsenetPrima facie liability subject to EmmensOwner of noticeboardYes
TilleyRetail ISPsKnowing involvement; not mere passive rolePostal service or telephone companyNo
DesignTechnicaSearch engineHuman input; not passive mediumLibrary catalogue distinguishedNo
KaschkeWebsite operatorCapacity to monitor or editN/AYes
Host of BloggerAuthorised or acquiesced in publicationOwner of noticeboard
Owner of graffiti wall
Table 8.1
Outcomes in publication cases
GodfreyHost of UsenetPrima facie liability subject to EmmensOwner of noticeboardYes
TilleyRetail ISPsKnowing involvement; not mere passive rolePostal service or telephone companyNo
DesignTechnicaSearch engineHuman input; not passive mediumLibrary catalogue distinguishedNo
KaschkeWebsite operatorCapacity to monitor or editN/AYes
Host of BloggerAuthorised or acquiesced in publicationOwner of noticeboard
Owner of graffiti wall
GodfreyHost of UsenetPrima facie liability subject to EmmensOwner of noticeboardYes
TilleyRetail ISPsKnowing involvement; not mere passive rolePostal service or telephone companyNo
DesignTechnicaSearch engineHuman input; not passive mediumLibrary catalogue distinguishedNo
KaschkeWebsite operatorCapacity to monitor or editN/AYes
Host of BloggerAuthorised or acquiesced in publicationOwner of noticeboard
Owner of graffiti wall

8.106  Difficulty of bringing claims. Second, these common law limits upon the scope of publication, together with the statutory and other limitations considered in section 3, act as strong deterrents of claims against intermediaries who host, cache, or transmit defamatory material. By and large, these claims fail—either because the defendant is not sufficiently involved in publication to face prima facie liability, or because another defence applies. This may partly account for the paucity of such claims.

8.107  Policy considerations. Third, English courts exhibit reluctance to engage with policy arguments which favour limits on intermediary liability. This reflects an approach that regards internet publication as a matter to be fitted within the existing doctrinal framework, rather than as one element of a larger problem of internet content regulation. With few exceptions, courts have not considered whether remedies against secondary publishers would interfere with users’ and intermediaries’ fundamental rights.139

8.108  Chilling effects on innovation. A further risk of adopting liability rules which cast the net of prima facie liability too widely is that they may deter innovative forms of content moderation. Anticipated liability discourages investment in automated content analysis and moderation tools, since they have a tendency to be used by claimants (with varying levels of success) to show that the defendant possesses knowledge of and control over material that eventuates to be defamatory. Although recent limits on publication are welcome developments, the courts’ failure to address these arguments largely ignores the important communications policies which underlie limits on secondary publishers’ liability at common law.

8.109  Upstream and downstream limitations. To regard an internet intermediary as a non-publisher may be thought of as an upstream limitation on liability, in that it defeats any claim at a preliminary stage without having to raise a positive defence or contest the meaning and imputations of a statement.140 Even if an intermediary is responsible as a publisher, several downstream limitations exist which carry over the policy of protecting innocent conduits from liability for tortious statements made by others.

8.110  Overview of defences. First, a service provider may avail itself of the defence of innocent dissemination within the meaning of section 1 of the Defamation Act 1996 (UK) (‘1996 Act’). Second, the statutory safe harbours available under regulations 17–20 of the E-Commerce Regulations may apply; these are the subjects of chapter 12. Third, the defences of voluntary disclosure and non-primary publication enacted by the Defamation Act 2013 may apply.

8.111  Underlying policies. Unlike a plea of justification, which relies on external characteristics of the statement itself, these defences protect intermediaries based on their identity, mental state, and relationship to the primary wrongdoer. In this respect, they more closely resemble hybrid defences such as fair comment or publication on a matter of public interest, which involve considerations relating both to the statement and the status of its maker. The defences considered in this section reflect a policy of encouraging claimants to exhaust their remedies against primary wrongdoers, though they suffer from imprecision, fragmentation, and a failure to address more fundamental concerns over cost and procedure in defamation actions.

8.112 Section 1(1) of the 1996 Act provides that:

In defamation proceedings a person has a defence if he shows that—


he was not the author, editor or publisher of the statement complained of,


he took reasonable care in relation to its publication; and


he did not know, and had no reason to believe, that what he did caused or contributed to the publication of a defamatory statement.

8.113  Statutory purpose. The impetus for statutory reform emerged from concerns that the common law defence, which the Act ‘superseded’ (but did not expressly abolish),141 was inadequate to protect intermediaries.142 Section 1 was intended to modernise and expand the common law defence,143 but its three conjunctive requirements offer narrower protection to internet intermediaries, since notice of allegedly defamatory matter is usually sufficient to defeat it.

8.114  Status of the defendant. The first requirement excludes any defendant who played a part in composing the words complained of, such as their authors or editors, and any ‘commercial publisher’ whose business is to issue material to the public or a section of it.144

8.115  Professional website publishers. In DesignTechnica, Eady J stated in obiter that section 1(1)(a) prevents commercial website operators from relying on the defence, since their businesses involve ‘issuing material to the public’.145 Such an approach would have the unintended effect of excluding any intermediary who disseminates a defamatory statement in the course of an information business. By contrast, Godfrey concluded that the defendants ‘incontrovertibly can avail themselves of section 1(1)(a)’, despite the fact that the ISP was in the business of storing and providing access to Usenet materials to a section of the public; namely, its subscribers.

8.116 In Amazon, the Court held that a defence of innocent dissemination could apply where the website did not ‘originate’ material and merely supplied the system or service through which publication occurred.146 To similar effect, Tamiz would have treated Google as outside the definition of ‘commercial publisher’, because material was issued to the public by Blogger’s users and not Google itself.

8.117  Examples of secondary disseminators. This view of paragraph (1)(a) is clearly more consistent with the text and purpose of section 1, which was intended to apply to new media147 and protect ‘a conduit which has enabled another person to publish defamatory material’.148 It is also subservient to subsection (3), which gives non-exhaustive examples of activities deemed innocent dissemination; among them are ‘providing any...service by means of which the statement is...made available in electronic form’149 and ‘provid[ing] access to a communications system’ without ‘effective control’ over its users. Only if the intermediary goes beyond these activities—eg by composing, selecting,150 or editing material—might it fail to satisfy the condition. The contrary view would render section 1 completely ineffective at protecting providers of commercial internet services.

8.118  Nature of the duty. Second, secondary publishers must exercise due care in relation to a publication. This involves considering the extent of their responsibility for the statement, the nature or circumstances of its publication, and any previous conduct or character of the primary wrongdoer.151

8.119  Automatic and unmoderated publication. Although there is little case law interpreting this duty, it appears difficult to satisfy where the secondary publisher acts passively. In DesignTechnica, Eady J found it ‘difficult to comprehend’ how Google could satisfy paragraph (1)(b) when publication occurred automatically without human input.152 HHJ Moloney QC refused to strike out the claim in Amazon for this reason, though noted the possible counter-argument that pre-moderation was not reasonable on a ‘vast website’.153

8.120  Passing on complaints. In Tamiz, Eady J thought it not ‘outside the bounds of a reasonable response’ for Google not to investigate content published on Blogger—because there it did pass complaints to the primary author.154

8.121  Relevance of terms of service. In Tilley the ISPs’ contractual terms, which prohibited subscribers from publishing defamatory statements, were evidence of reasonable steps, though they took no steps to enforce them.155 Given that Google’s policies also prohibit defamatory content, Tilley and DesignTechnica are difficult to reconcile. Ultimately, once an intermediary is given notice, failing to act is ‘an insuperable difficulty’ to establishing reasonable care156—a view supported by extrinsic materials.157

8.122  The required mental state. Third, the secondary publisher must not know or have reason to believe that its conduct contributes to the publication of the defamatory statement. These words ‘import the concept of recklessness’158 rather than constructive knowledge; they operate on what the defendant actually knew. However, a good faith belief that a defamatory statement is justified (eg because true or privileged) will not satisfy paragraph (1)(c).159 The defendant must be an ‘unwitting contributor’ to publication and have ‘no idea of the defamatory nature’ of material.160

8.123  Examples. In Godfrey and Tamiz, the defence was unavailable once the defendant had been notified. At that point, even with an assurance of truth from the author, the intermediary would have cause to believe that its services were contributing to a relevant publication.161 However, in Tilley, two of the defendant ISPs were held to satisfy paragraph (1)(c) because they never received proper notice.162

8.124  Comparison with common law defence. The combined effect of paragraphs (a)–(c) is to exclude reliance upon the statutory defence after an intermediary receives valid notice from the claimant. This means that it adds little to the common law protections for secondary disseminators considered in section 1.2.

8.125  Overview. The European safe harbours described in chapter 12 apply, in principle, to defamation actions against internet intermediaries that qualify as information society service providers. Of the three safe harbour activities, mere conduit and caching are unlikely to provide protection beyond the common law. Hosting is the most important safe harbour for defamation claims, but its limits are unclear and defined by reference to knowledge, which discourages voluntary content moderation while rewarding reflexive removal.

8.126  Examples. Safe harbours have generally shielded intermediaries from monetary liability in defamation claims (see Table 8.2). The details of the safe harbours and their application to defamation cases are discussed in chapter 12.

Table 8.2
English defamation cases involving safe harbours
CaseDefendantPublisher?Mere conduit?Storage/caching?Result
GodfreyHost of UsenetYesN/A — prior to RegulationsLiable
TilleyRetail ISPsNoN/A (Yes)N/A (Yes)Not liable
DesignTechnicaSearch engineNoNoNoNot liable
KaschkeWebsite operatorYesNoYesNot liable
DavisonHost of BloggerYesNoYesNot liable
TamizPossiblyNoYes (at trial)Not liable
CaseDefendantPublisher?Mere conduit?Storage/caching?Result
GodfreyHost of UsenetYesN/A — prior to RegulationsLiable
TilleyRetail ISPsNoN/A (Yes)N/A (Yes)Not liable
DesignTechnicaSearch engineNoNoNoNot liable
KaschkeWebsite operatorYesNoYesNot liable
DavisonHost of BloggerYesNoYesNot liable
TamizPossiblyNoYes (at trial)Not liable
Table 8.2
English defamation cases involving safe harbours
CaseDefendantPublisher?Mere conduit?Storage/caching?Result
GodfreyHost of UsenetYesN/A — prior to RegulationsLiable
TilleyRetail ISPsNoN/A (Yes)N/A (Yes)Not liable
DesignTechnicaSearch engineNoNoNoNot liable
KaschkeWebsite operatorYesNoYesNot liable
DavisonHost of BloggerYesNoYesNot liable
TamizPossiblyNoYes (at trial)Not liable
CaseDefendantPublisher?Mere conduit?Storage/caching?Result
GodfreyHost of UsenetYesN/A — prior to RegulationsLiable
TilleyRetail ISPsNoN/A (Yes)N/A (Yes)Not liable
DesignTechnicaSearch engineNoNoNoNot liable
KaschkeWebsite operatorYesNoYesNot liable
DavisonHost of BloggerYesNoYesNot liable
TamizPossiblyNoYes (at trial)Not liable

8.127  The statutory defence. Section 10 of the Defamation Act 2013 provides:

A court does not have jurisdiction to hear and determine an action for defamation brought against a person who was not the author, editor or publisher of the statement complained of unless the court is satisfied that it is not reasonably practicable for an action to be brought against the author, editor or publisher.

The effect of this section is to remove all subject matter jurisdiction to hear defamation actions against secondary parties, unless it is ‘not reasonably practicable’ to proceed against any of the specified primary parties. This limitation applies to both offline intermediaries, such as booksellers, and all internet services who are not primary publishers.163

8.128  Underlying policy. This provision drastically curtails the responsibility of service providers, who no longer face liability unless the claimant can first clear the threshold of proving impracticability. This expresses a clear policy of encouraging complainants to resolve disputes directly with primary wrongdoers (and limited relational secondary wrongdoers, such as employers) rather than pursuing secondary disseminators. In doing so, section 10 directs claimants’ attention to primary wrongdoers before seeking to shift losses and enforcement costs onto secondary parties. However, it suffers from several ambiguities which are yet to be resolved by the courts.

8.129  Meaning of ‘author’, ‘editor’, and ‘publisher’. First, section 10 does not apply to actions against a person who is the ‘author’, ‘editor’, or ‘publisher’ of the statement complained of. According to section 10(2), an author is the originator of the statement, an editor is a person ‘having editorial or equivalent responsibility’ for the statement or the decision to publish it. This would include a website operator who edits submissions from users, as Warby J held in Solicitors from Hell.164 Finally, a publisher is a commercial person whose business is issuing the material to the public, in the course of which the statement was issued.

8.130  Intermediaries who are not primary publishers. Section 10(3) of the 2013 Act sets out a number of exclusions which clarify that certain intermediaries are not to be regarded as authors, editors, or commercial publishers of statements whose publication they facilitate. These examples may be applied by analogy to other classes of intermediaries.165 First, the mere processing, making copies of, distributing, or selling the statement in any electronic medium is insufficient. Second, so is operating or providing any equipment, system, or service by means of which the statement is retrieved, copied, distributed, or made available in electronic form. Third, to operate or provide access to a communications system by means of which the statement is transmitted, or made available, by a person over whom the intermediary has no effective control, is likewise insufficient.166 These examples clarify that intermediaries who do no more than facilitate access to a statement will be taken not to be its author, editor, or publisher, and seem intended to avoid the kind of debate that clouded section 1(1)(a) of the 1996 Act.

8.131  Meaning of ‘not reasonably practicable’. Second, section 10 does not apply where the Court is satisfied that it is not ‘reasonably practicable’ to bring an action against the author, editor, or publisher of the material. However, the threshold of impracticability is unclear. If material is posted anonymously, a claimant might yet obtain voluntary disclosure—or, failing that, a Norwich Pharmacal order—from the intermediary and thereby proceed against the primary party. This would, presumably, be reasonably practicable; a lower threshold of practicability would deprive this limitation of its intended effect. In Solicitors from Hell, Warby J doubted whether a bare plea of impracticality arising from the anonymity of the author of the material would suffice, but did not resolve the question.167

8.132  Claims for injunctive relief. Third, although section 10 excludes any ‘action for defamation’, it is unclear whether it prohibits joinder of intermediaries for the purpose of injunctive relief—though joinder solely for that purpose may well be refused if the claimant could already obtain a remedy against the primary author.168 However, it is strongly arguable that section 10 would not preclude a claim against an intermediary for Norwich Pharmacal disclosure of the identity of the person or persons responsible for an allegedly defamatory publication (since such a claim would not be an ‘action for defamation’).

8.133  Pre-emptive action by intermediaries. Fourth, because section 10 does not displace the traditional rules on publication and safe harbours, intermediaries may still need to remove content (particularly if anonymous) on the assumption that action against the primary tortfeasor may be unavailable. Section 10 is built on the foundation that no claim for defamation should be brought against an intermediary unless and until it turns out to be impractical to proceed against the primary publisher. However, if an intermediary does not act on a complaint made about defamatory content and it later eventuates that action against the primary publisher is not reasonably practicable, then it is an open question whether the intermediary could face liability for acquiescing to publication from the date of the original notification, or only from the date it was notified that the conditions under section 10 had been satisfied. It is suggested that the latter approach is more consistent with the statutory purpose of section 10.

8.134  The statutory defence. Section 5 of the 2013 Act applies where the operator of a website is sued for defamation but was not the person who ‘posted’ the defamatory statement. This defence applies unless it is not possible for the claimant to identify the person responsible and ‘gave the operator a notice of complaint’ about the statement, to which the operator failed to respond as required by regulations.

8.135  Effect of the defence. Where material is posted with attribution, this confirms that the intermediary need do nothing: the claimant’s grievance lies with the primary author. Where it is posted anonymously, website operators must voluntarily disclose the poster’s identity to retain their immunity. Having done so, they then remain protected even after acquiring actual knowledge of actionable material.

8.136 Identity disclosure is discussed in chapter 4. It forms a vital part of any effective system of internet enforcement. The new defence of voluntary disclosure primarily benefits application-layer intermediaries who are considered publishers of user-created content, since existing case law exonerates many gateways, ISPs, and hosts at the publication stage. Section 5 also creates several uncertainties.

8.137  Onus of proof. First, like section 1 of the 1996 Act, section 5 may be less attractive to service providers than passive inactivity, since the claimant bears the onus of proving publication, while the defendant would have to raise and prove each element of the defence.169

8.138  Meaning of ‘website operator’. Second, the term ‘website operator’ is not defined by the Act. It is unclear, but sensible, that the defence should extend to publications of a statement occurring by means of the website version of a platform which is also accessible via a mobile application (such as Facebook). Also unclear is whether the defence would apply to exonerate the operators of other interactive services which are not delivered through a web browser (eg online multi-player games in which statements can be made by players). Further complications arise regarding in relation to blogging and cloud services, for which there may be multiple potential operators. It is suggested that a broad approach should apply; to limit the defence arbitrarily risks redundancy in an age of apps and platforms.

8.139  Public interest anonymity. Third, unlike Norwich Pharmacal orders, section 5 makes no provision for public interest anonymity. Presumably this is deliberate. However, a defence of voluntary disclosure should not lead to websites being encouraged to disclose the identity of people whose anonymity it is in the public interest to preserve. Widespread disclosure without judicial supervision would make it much harder for people to write anonymously about controversial but lawful topics, and may expose activists in developing countries to harm.

8.140  Risk of unnecessary disclosure. Fourth, section 5 is likely to encourage automatic disclosure by risk-averse website operators who fear the absence of the defence if they fail to comply, even if the merits of primary claim are weak or the maker of the statement has an obvious defence. Unlike the judicially mediated Norwich Pharmacal procedure, this invites opportunistic allegations of defamation to discover primary authors’ identities cheaply; yet to give disclosure may unjustifiably invade their privacy if there is no arguable, let alone actionable, wrong.

8.141  Impossibility of identification. Fifth, the threshold of ‘possible’ identification is unclear. It suggests a higher standard than ‘reasonably practicable’, yet if the operator possesses the information (which section 5 presupposes, since without it the operator could disclose nothing), then it will be ‘possible’ to obtain it via a Norwich Pharmacal application and section 5 will have no application. Conversely, if judicial disclosure has been sought and failed, then voluntary disclosure is unlikely to be appropriate.

8.142  Extent of identification. Sixth, the degree of identification is unclear. An IP address might be all that the website operator possesses, so the claimant is still likely to need to seek further Norwich Pharmacal relief against, at a minimum, the tortfeasor’s ISP (who could not avail itself of the defence, but would not need to in light of Tilley).

8.143  Lack of safeguards against abuse. Seventh, section 5 does not address the problem of ‘casual threats of litigation’170 being made against website operators in order to compel withdrawal of unsavoury but non-tortious material by people who wish to remain anonymous for other reasons.171 The prospect of defending costly proceedings is at least as chilling when threats are made to primary authors rather than intermediaries, since authors are typically one-shot defendants who lack the specialist advice or resources necessary to defend claims. For this reason, section 5 fails to solve the problem of intimidation, and risks being turned into a vehicle for reputation management. Disclosure and takedown may prove even more chilling than pure takedown, unless the cost of defending an unmeritorious action can be substantially reduced.

8.144  Defences available to the primary publisher. Eighth, the notice of complaint only has to specify why the statement is defamatory. This ignores possible defences available to the original poster, which might mean that the statement, although defamatory, is not actionable. Disclosure might therefore be expected on the basis of true statements which involve no tort, which may lead to the surprising result that section 5 does not apply but the website operator has a defence under the storage safe harbour.

8.145  Moderation of content. Finally, section 5 discourages content moderation by failing to protect a website operator who amends user-created content in good faith. Most of these issues could be avoided with appropriate guidance in the regulations. Despite its many problems, section 5 reflects the desirable policy of encouraging complainants to resolve disputes directly with primary wrongdoers, rather than pursuing monetary remedies against secondary publishers.

8.146  No real or substantial tort. Abuse of process offers a basis for striking out actions against intermediaries where there has been no real or substantial tort. As the Court of Appeal held in Jameel v Dow Jones & Co Inc, defamation proceedings will be dismissed where there is no realistic prospect that they would yield any legitimate advantage that outweighs their cost and expense—in other words, where ‘the game is not worth the candle’.172 Although once rejected in Goldsmith as a question for Parliament,173 this principle now follows from section 6 of the Human Rights Act 1998 and the overriding objective, which oblige courts to protect the rights of defendants in actions which are not necessary and proportionate to protecting the claimant’s reputation.174

8.147  Application to internet intermediaries. The Jameel doctrine has been applied in several cases involving internet intermediaries. For example, in Kaschke, proceedings were not proportionate because the original posting had been removed and a right of reply already published three years earlier; a claim against the website operator was therefore unlikely to vindicate the claimant any further.175

8.148 In Davison, the claimant’s action against Google was disproportionate considering the limited extent of publication, especially when she had remedies against the primary authors.176 The same was true of the action in Tamiz, where—even though the comments were visible for a longer period—they had since ‘receded into history’ and caused minimal damage.177 Accordingly, even if claims technically lie against service providers, it would not often be a proportionate use of the Court’s resources or compatible with defendants’ and users’ rights to allow them to proceed.

8.149  Removal of lawful internet publications. The relationship between removal and freedom of expression is complex. In Reynolds v Times Newspapers Ltd, Lord Hobhouse remarked: ‘There is no human right to disseminate information that is not true.’178 Although this statement must be qualified—certain untruths, such as satire, honest opinion, and responsible journalism, may be lawfully published—it reflects the obvious premise that removing material interferes with authors’ and readers’ freedom of expression only to the extent that non-tortious information is affected. Article 10(2) of the Convention acknowledges this by permitting derogations ‘for the protection of the reputation or the rights of others’. The relevant question is therefore to what extent secondary liability rules encourage the removal of non-tortious statements. This may occur in several ways. It is suggested that courts and practitioners should be mindful of these risks when formulating and responding to claims for removal of allegedly defamatory material.

8.150  Low threshold for notification. First, despite Tamiz, the prima facie burden on claimants remains low—to notify an intermediary (usually electronically) of defamatory material, giving details of its actionability. Intermediaries are not obliged to consider both sides of the argument, let alone conduct their own investigations into the merits of any possible defence, since this is costly and impracticable in large volumes.

8.151  Limited investigation of defences. Second, even were they so inclined, intermediaries will almost never possess the evidence necessary to raise substantive defences to a defamation claim since they are not, by definition, involved in the composition of defamatory material.179 This may prevent relevant limits on claimants’ rights from being pleaded and cause non-tortious content to be removed. Once an intermediary has been notified, the most sensible course is usually to remove, since it may be unclear whether the notification confers sufficient knowledge to exclude protection. This risk is especially harmful to bloggers, whose reportage frequently serves valuable public interests.

8.152  Risk aversion. Third, intermediaries typically wish to avoid a protracted dispute, since there may be substantial costs even if the claimant fails at trial. Even well-founded but disproportionate claims have the potential to interfere with freedom of expression.

8.153  Effect of the internet. The internet has dramatically increased the extent to which defamatory statements are communicated, consumed, and remembered. These developments are summarised in chapter 1 and below.

8.154Low cost of dissemination. First, it makes possible communications between individuals and other members of the public ‘at virtually no cost’,180 which makes defamatory material more common and widespread.

8.155Speed and extent of dissemination. Second, because material is easily accessed and propagated, the ‘poison tends to spread far more rapidly’ online.181

8.156Longevity of material. Third, its permanence is prolonged thanks to search engines, caches, and archives.182

8.157Professional defendants. Fourth, claimants may be disadvantaged in claims against intermediaries who are, through their online activities, ‘repeat players’ in defamation litigation.183

8.158Anonymous publications. Fifth, traditional remedies will often be futile where the defamatory words are published anonymously or from a place where service providers enjoy stronger protections. Jurisdictional competition for their business and incompatible national speech values make this form of regulatory arbitrage inevitable.184

8.159  The role of the common law. Although not all these factors are unique to the internet, the common law has always been concerned that new technologies which assist the spread of defamatory material should be brought within its control. As Brett J (who would later decide Emmens) cautioned in a decision recognising publication by telegraph:

It was never meant by the legislature that these facilities for postal and telegraphic communication should be used for the purpose of more easily disseminating libels.185

8.160 Without meaningful ways to target offshore and anonymous internet material, it may be impossible for claimants to obtain redress—particularly if damages against the enabling service provider are excluded by the doctrines discussed in section 3. These factors make it imperative for courts to develop alternative responses to internet defamation.

8.161  Advantages of removal. In many cases, prompt removal of defamatory material offers satisfactory redress without the need for further litigation.186 Indeed, the effectiveness of removal is one reason why the number of claims against internet intermediaries remains low despite self-evident growth in the quantity of defamatory internet publications. If a claimant seeks only a retraction or apology, proceedings against an intermediary offer few additional benefits—a fact acknowledged in Davison and Kaschke—but carry higher costs and potential for adverse publicity.187

8.162  Disadvantages of removal. However, removal is imperfect: it may be unavailable from platforms and gateways that rely on primary authors to withdraw defamatory material (as in Tamiz); when it is available, removal tends to produce over-enforcement characterised by ‘spill-over’ removal of non-tortious content. Further, many passive intermediaries refuse to remove defamatory content without a court order, on the reasonable supposition that a court is usually in the best position to determine whether material is tortious. However, the cost of pursuing a defamation complaint to this stage is prohibitive and may frustrate legitimate claims.

8.163  Restrictions on access to defamatory material. Where it is not possible to remove defamatory material by an order directed at the primary wrongdoer or host, it may be possible to restrict access to that material by means of an order requiring an ISP to attempt to block access to it, or one requiring a search engine to remove the material from search results. These remedies are discussed further in chapters 14 and 16.188

8.164  Identification of primary authors. Section 5 of the Defamation Act 2013 creates a defence of disclosure, which makes it easier for claimants to identify, contact, and, if necessary, obtain injunctive relief against the primary authors of material. This approach protects freedom of expression without depriving claimants of a remedy; it solves a basic problem of high transaction costs between intermediaries and primary authors, allowing the latter to assume the risk that information is tortious and so make the decision whether to leave impugned materials online. While this approach is imperfect for the reasons noted previously, it reflects a fairer balance between the rights of authors, intermediaries, and claimants. Indeed, the disclose-and-indemnify approach should arguably be extended to other torts.

8.165  Corrections. A fourth possibility is to require storage intermediaries to publish corrections or disputations without monetary liability and thereby avert much of the injury caused by defamation without harming freedom of expression. These amendments could only be published by the website operator, host or cache of material (because conduits could not easily detect and update it during transmission), but could occur with or without the primary author’s cooperation. As Mullis and Scott have argued,189 defamation should recognise discursive remedies which give defamed claimants what they really want without the need for protracted and costly court procedures; namely, an opportunity to reply with their side of the story and to have the true facts published in place of (or alongside) the false ones.

8.166  Advantages of discursive remedies. While a discursive response could not provide full vindication in all cases, it would offer a cheap, speedy way for claimants to correct or dispute the veracity of statements concerning them without unnecessarily interfering with freedom of expression, innovation, or neutrality. Once a discursive remedy has been obtained, little purpose is served by a claim against an internet intermediary—except to enforce the discursive remedy itself. The availability of non-monetary injunctive relief is considered further in chapters 14, 15, and 16.


See Colin Lovell, ‘The “Reception” of Defamation by the Common Law’ (1962) 15 Vanderbilt Law Review 1051, 1062 (describing the arrival of the Caxton press in London).


See Broughton’s Case (1583) 1 Moo KB 141; 72 ER 493.


See Edwards v Wooton (1608) 12 Co Rep 35; 77 ER 1316; Jones v Davers (1653) 1 Cro Eliz 497; 78 ER 747; Barrow v Lewellin (1792) 1 Hob 62; 80 ER 211.


Al Amoudi v Brisard [2007] 1 WLR 113, 121–2 (Gray J).


Pullman v Hill & Co [1891] 1 QB 524, 527 (Lord Esher) (a man cannot publish to himself); Powell v Gelston [1916] 2 KB 615, 619 (Bray J) (publication ‘is the foundation of the action’).


David Price and Korieh Duodu, Defamation Law, Procedure & Practice (3rd ed, 2004) [3–02].


Bond v Douglas (1836) 7 C & P 626.


Watts v Fraser (1835) 7 C & P 369.


Goldsmith v Sperrings Ltd [1977] 1 WLR 478, 487 (Lord Denning MR).


Lord Darcy v Markham (1792) 1 Hob 120, 121; 80 ER 270 (attempt to provoke a duel using slanderous words).


See Holdsworth, A History of English Law (1926) vol 8, 336.


See Palmer and Thorpe’s Case (1583) 4 Co Rep 20a (Lord Coke); Roscoe Pound and Theodore Plucknett, Readings on the History and System of the Common Law (3rd ed, 1927) 71–2, 480.


Van Vechten Veeder, ‘The History and Theory of the Law of Defamation’ (1904) 4 Columbia Law Review 33, 45.


Other options included treason, scandalum magnatum, heresy, and the systems of royal and statutory print licensing: see Philip Hamburger, ‘The Development of the Law of Seditious Libel and the Control of the Press’ (1985) 37 Standard Law Review 661, 666–73.


(1588) 31 Eliz 1263.


(1744) 1 Barn KB 304, 305; 94 ER 207.


(1775) 2 Blackstone W 1037, 1038; 96 ER 610, 611 (De Grey CJ).


(1837) 7 Ad & El 223, 232; 112 ER 455, 460 (Lord Denman CJ).


Hamburger, n 14, 761.


Cf Lobay v Workers and Farmers Publishing Association [1939] 2 DLR 272, 275 (Taylor J) (where the defendant printer knew its equipment was being misused and failed to do anything about it).


R v Zenger (1735) 9 Geo II 675, 677 (emphasis added).


(1871) 8 LR 394, 399 (Wickens VC).


(1680) 32 Charles II 925, 930. See also R v Curll (1727) 17 St Tr 154.


(1885) QBD 354 (‘Emmens’).


Emmens, 357 (Lord Esher MR) (Cotton LJ agreeing). See also Emmens, 358 (Bowen LJ).


[1899] AC 549 (‘McLeod’).


McLeod, 562 (Lord Esher MR).


Huth v Huth [1915] 3 KB 32. Cf Weld-Blundell v Stephens [1920] AC 956 (negligent publication).


(1932) 48 TLR 521 (‘Woolworth’); approved in Tamiz v Google Inc [2013] EWCA Civ 68, [26].


[1900] 2 QB 170 (‘Vizetelly’).


Vizetelly, 175–6 (Smith LJ). See also Vizetelly, 178 (Vaughan Williams LJ), 179 (Romer LJ).


See also Weldon v ‘The Times’ Book Company (Ltd) (1911) 28 TLR 143.


[1977] 1 WLR 478 (‘Goldsmith’).


Goldsmith, 487 (Lord Denning MR) (emphasis added).


Goldsmith, 488 (Lord Denning MR).


Cf Information Society Directive recital (59) (which speaks of the intermediary ‘best placed’ to avoid an infringement of copyright or related rights): see chapter 14, section 3.1.


Buckley v Wood (1591) 33 & 34 Eliz 888, 891; 4 Co Rep 14b. See also Hersey’s Case (1572) 77 ER 1378; 12 Co Rep 103, 104 (damages awarded for exhibition of a malicious bill).


[1937] 1 KB 818 (‘Byrne’).


Byrne, 830 (Greene LJ).


Byrne, 837 (Greene LJ).


See also Urbanchich v Drummoyne Municipal Council (1991) Aust Tort Rep 81–127 (public authority liable for defamatory posters exhibited in bus shelters by members of the public).


See section 1.3.


Day v Bream (1837) 2 M & R 54, 56; 174 ER 212 (Patteson J).


Paul Mitchell, The Making of the Modern Law of Defamation (2005) 107 (emphasis added).


Later cases appear to take this view: see Ridgway v Smith & Son (1890) 6 TLR 275; Mallon v W H Smith & Son (1893) 9 TLR 621.


Glanville Williams, Joint Torts and Contributory Negligence (1951) 10.


See, eg, Lord Porter et al, Report of the Committee on the Law of Defamation (1948) (Cmd 7536) 29.


See also Webb v Bloch (1928) 41 CLR 331, 364 (Isaacs J).


(1846) 8 QB 533, 536 (Lord Denman CJ), 537 (Coleridge J).


Hays plc v Hartley[2010] EWHC 1068 (QB), [14] (Tugendhat J) (‘Hays’).


Parkes v Prescott (1869) LR 4 Exch 169, 177, 179 (Montague Smith J).


Because defamation requires the conveyance of human-readable information, it implicates higher-layer services rather than physical-layer intermediaries.


See, eg, Atlantis World Group of Companies NV v Gruppo Editoriale L’Espresso SPA [2008] EWHC 1323 (QB), [19], [42] (Gray J) (parallel publication of article in magazine and website).


For discussion of hosting platforms such as Google’s, see paras 8.71 to 8.83.


Jameel, 966 (Lord Phillips MR). See also para 8.146.


Steinberg v Pritchard Englefield [2005] EWCA Civ 288, [21] (Sedley LJ).


Loutchansky v Times Newspapers Ltd [No 2] [2001] EMLR 876, [14]–[15], [20] (Gray J).


Davison, [24] (HHJ Parkes QC).


Trumm v Norman [2008] EWHC 116 (QB), [36]–[37] (Tugendhat J).


Kaschke, [95] (Stadlen J).


Al Amoudi v Brisard [2007] 1 WLR 113, 123 (Gray J).


[2008] EWHC 1781 (QB).


See chapter 4, section 2.5.


(Unreported, High Court of Justice, 10 March 2011).


[2010] EWHC 2859 (QB), [20], [41] (Tugendhat J); [2012] EWHC 756 (QB), [122]–[123] (Bean J).


[2012] EWCA Civ 1382, [27] (damages only).


[2015] EWHC 3154 (QB) (‘Richardson’).


The Facebook platform was not operated or controlled by the defendant, Facebook UK Ltd, but rather by Facebook Inc (a Delaware corporation) and Facebook Ireland Ltd: Richardson, [28]. This case illustrates the importance of selecting the correct defendant against whom to issue proceedings: see chapter 3, section 2.4 for further discussion.


Richardson, [43], [48] (Warby J).


In a fourth case, Carrie v Tolkien [2009] EWHC 29 (QB), the defendant posted a critical comment to the claimant’s weblog, which was allowed to stand for some 22 months, leading to the unsurprising conclusion that the claimant had consented to the defamation: at [17] (Eady J). As an acquiescence case involving primary authors rather than intermediaries, it is distinguishable.


[2009] EWHC 1765 (QB) (‘DesignTechnica’). See section 2.4 in relation to search engines.


DesignTechnica, [46] (Eady J).


[2012] EWHC B3 (QB) (‘Amazon’).


See chapter 12, section 5.


[2010] EWHC 690 (QB), [27]–[28], [84]–[85] (Stalden J) (‘Kaschke’).


Kaschke, [40]–[41] (Stalden J).


Kaschke, [102]–[103] (Stalden J).


See paragraph 8.48.


[2015] EWHC 2628 (QB), [2]–[7] (Warby J) (‘Solicitors from Hell’).


Solicitors from Hell, [20]–[23] (Warby J).


See Duke of Brunswick v Harmer (1849) 14 QB 185; Lewis v King [2004] EWCA Civ 1329, [29]–[30].


See Times Newspapers Ltd [Nos 1 and 2] v United Kingdom (European Court of Human Rights, Fourth Section, 10 March 2009) [45]–[49].


Defamation Act 2013 s 8.


Defamation Act 2013 s 8(5).


[2001] QB 201, 209, 212 (Morland J) (‘Godfrey’).


‘Frogery’ referred to the practice of impersonating a newsgroup user for the purpose of harming their online reputation. Today, ‘frape’ (a portmanteau of ‘Facebook rape’) is equivalent: see Rick Moen, ‘Obfuscation Mitigation (Lexicon)’ (11 December 2008) <>.


Godfrey, 203 (Morland J).


Godfrey, 207 (Morland J).


In reality, the claimant also made ‘puerile, unseemly and provocative’ postings ‘to tempt people to overstep the mark and defame the Plaintiff so that he can sue’: Godfrey v Demon Internet Ltd [1999] (Unreported, High Court of Justice, Morland J, 23 April 1999), [14]–[16].


See, eg, Bryce v Barber (Unreported, High Court of Justice, 26 July 2010) (awarding damages for ‘Facebook rape’).


See section 2.3.


[2011] EWHC 3031 (QB), [17] (HHJ Parkes QC) (‘Davison’).


Davison, [11]–[15] (HHJ Parkes QC).


See section 3.5.


Davison, [42] (HHJ Parkes QC).


[2012] EWHC 449 (QB), [38]–[39] (Eady J) (‘Tamiz’).


Tamiz v Google Inc [2013] EWCA Civ 68, [34] (Richards LJ) (Lord Dyson MR and Sullivan LJ agreeing) (‘Tamiz (CA)’). The Court of Appeal otherwise upheld Eady J’s approach as essentially sound: at [23].


The experience of absolute immunity in the United States is cautionary: cf Communications Decency Act 1996 (US) 47 USC § 230; Mark Lemley, ‘Rationalizing Internet Safe Harbors’ (2007) 6 Journal on Telecommunications & High Technology Law 101, 112–13; Brian Leiter, ‘Cleaning Cyber-Cesspools: Google and Free Speech’ in Levmore and Nussbaum (eds), The Offensive Internet: Speech, Privacy, and Reputation (2010) 155, 172–3.


Google Inc, ‘Blogger Content Policy’ (2012) <>.


Google Inc, ‘Google Terms of Service’ (1 March 2012) <>.


This assumes that the average blog post comprises roughly one A4 page, stacked in columns of four sheets.


Emmens, 357 (Lord Esher MR) (Cotton LJ agreeing). See paragraph 8.15.


Davison, [45] (HHJ Parkes QC).


Tamiz, [10], [38] (Eady J).


Tamiz (CA), [18], [24], [33]–[35] (Richards LJ).


Tamiz, [38] (Eady J).


National Coal Board v J E Evans & Co (Cardiff) Ltd [1951] 2 KB 861, 874–5 (Cohen LJ).


In Tamiz, five weeks elapsed from notification to removal, which was ‘somewhat dilatory’ and left room for such an inference: Tamiz (CA), [35] (Richards LJ). It is illustrative of the difficulties involved that Google Inc received effective notice only when the claim form was issued: at [10].


See section 2.4. It is not easy to see why Google’s choice to supply the Blogger service should lead to potential liability as a publisher of blog postings, but not its choice to offer a search service which lists defamatory results.


See, eg, Loutchansky v Times Newspapers Ltd [Nos 4 and 5] [2002] QB 783, 813 (Lord Phillips MR); James Tumbridge, ‘Defamation—The Dilemma for Bloggers and Their Commenters’ [2009] European Intellectual Property Review 505, 506.


The likely reason for Godfrey being misclassified as a case about ISP liability is that the defendant carried on business simultaneously as a provider of internet access and website hosting services. However, it was in its capacity as a website host and not an ISP that Demon published the defamatory statements.


Cf Tamiz, [39] (Eady J) (describing ISPs as ‘closely analogous’ to the Blogger platform). It is suggested that this goes no further than noting that both are intermediaries who are not necessarily publishers.


[2007] 1 WLR 1243 (‘Tilley’).


Tilley, 1249 (Eady J).


Tilley, [21]–[23] (Eady J) (emphasis in original).


[1899] AC 549, 562 (Lord Morris) (noting that ‘[a] printer and publisher intends to publish, and so intending cannot plead as a justification that he did not know the contents’).


See Matthew Collins, The Law of Defamation and the Internet (2nd ed, 2005) [15.38], [15.43].


Tilley, 1246 (Eady J).


Tilley, [21] (Eady J).


See Stephen Spencer, ‘Anatomy of a Google Snippet’ (Search Engine Land, 18 March 2010) <>.


See section 3.1.


DesignTechnica, [36] (Eady J) (emphasis in original) (citing Tilley, [23]).


DesignTechnica, [11]–[13], [50]–[51], [53] (Eady J). Although Google now supplies suggested search queries, it is submitted that the same result would follow, since the suggestions are equally automatic.


[2010] EWHC 616 (QB), [75] (Sharp J).


Cf DesignTechnica, [51], [55], [57]–[58], [124] (Eady J). It is possible that the URL of a hyperlink might contain sufficient words to convey the false imputation, as in the case of a link which read <>.


Emmens, 357 (Lord Esher MR); DesignTechnica, [64]–[65] (Eady J).


DesignTechnica, [49]; citing Tilley, [21].


See chapter 6, section 2.4 (copyright); chapter 10, section 3.2 (data protection); and chapter 16, section 1.1 (voluntary de-indexing).


DesignTechnica, [58] (Eady J).


DesignTechnica, [55] (Eady J).


Dr Yeung, Sau Shing Albert v Google Inc [2014] HKCFI 1404 (search engine); Oriental Press Group Ltd v Fevaworks Solutions Ltd [2013] HKCFA 47 (forum).


Crookes v Wikimedia Foundation Inc (2011) 337 DLR (4th) 1.


[2012] NZHC 2352, [51]–[52], [60]–[62] (Abbott AJ). Cf Murray v Wishart [2014] NZCA 461; [2014] 3 NZLR 722.


[2014] NSWSC 897, [83]–[85] (McCallum J).


[2012]VSC 533, [18]–[31] (Beach J) (‘Trkulja’).


[2015] SASC 170, [204]–[210], [221]–[222] (Blue J). The Court later awarded AUD$100,000 in general damages: Duffy v Google Inc (No 2) [2015] SASC 206.


R C Donnelly, ‘History of Defamation’ [1949] Wisconsin Law Review 99, 123, 124.


DesignTechnica, [113] (Eady J).


Cf DesignTechnica, [42], [44] (Eady J).


By analogy with Ronald Dworkin, ‘Foreword’ in Ivan Hare and James Weinstein (eds), Extreme Speech and Democracy (2009) 5, 8.


DesignTechnica, [70] (Eady J).


Lord Mackay, Reforming Defamation Law and Procedure: Consultation of Draft Bill (July 1995) [2.5], [2.6].


Cf New South Wales Law Reform Commission, Report 75—Defamation (1995) [9.9] <> (concluding that reform is better left to the courts, since legislative intervention could ‘stultify the development of the law’).


See 1996 Act ss 1(2), 17(1) (‘publisher’ not to be given its common law meaning).


DesignTechnica, [80] (Eady J).


Amazon, [40]–[41] (HHJ Moloney QC).


Mackay, n 142, [2.3].


Hansard, Defamation Bill, House of Lords (2 April 1996, Lord Chancellor) col 214.


1996 Act s 1(3)(c).


An example might be readers’ comments in a newspaper (whether published in print or online).


1996 Act s 1(5).


DesignTechnica, [74] (Eady J).


Amazon, [44] (HHJ Moloney QC).


Tamiz, [47] (Eady J); aff’d Tamiz (CA), [42]–[43] (‘a generous view’ but not wrong).


Tilley, [63]–[66] (Eady J).


Godfrey, 206 (Morland J).


Mackay, n 142, [2.4].


Milne v Express Newspapers [2005] 1 WLR 772, 788 (May LJ) (in the context of s 4(3)).


Hansard, Defamation Bill, House of Lords (2 April 1996, Lord Williams) col 216.


Hansard, Defamation Bill, House of Lords (2 April 1996, Lord Mackay LC) col 214.


Tamiz (CA), [44]–[46] (Richards LJ).


Tilley, 1261 (Eady J).


Parliamentary Research Paper 12/30, Defamation Bill No 5 2012–13 (28 May 2012) 20.


Solicitors from Hell, [23] (Warby J).


2013 Act s 10(3).


2013 Act s 10(3)(c), (e).


Solicitors from Hell, [24].


Davison, [69] (HHJ Parkes QC).


Graham Smith, ‘What the Defamation Bill Means for the Internet’ (Inforrm’s Blog, 17 May 2012) <