On 2 February 2016, the European Court of Human Rights (ECtHR) delivered its first post-Delfi judgment on the liability of online service providers for the unlawful speech of others. Somewhat puzzlingly, the Court reached the opposite conclusion from that of last summer’s controversial Grand Chamber ruling, this time finding that a violation of Article 10 of the European Convention on Human Rights (ECHR) had occurred through the imposition of liability on the applicant providers. While in principle therefore the judgment is good news for both internet intermediaries and their end-users, the ruling does little to dispel the legal uncertainty that plagues the area: attempting to reverse and head off in the right direction, the Court still finds itself falling over the stumbling blocks it set out for itself last year.
The facts of the case
The case involved a Hungarian self-regulatory body of internet content providers, Magyar Tartalomszolgáltatók Egyesülete (“MTE”), and Index.hu Zrt, one of Hungary’s major online news portals. The two companies had published an opinion piece on their respective websites, criticising the unethical and misleading business practices of a Hungarian real estate company. The piece attracted a number of angry comments posted, under pseudonyms, by the websites’ readers. “People like this should go and sh*t a hedgehog and spend all their money on their mothers’ tombs until they drop dead,” suggested one commentator. Another warned: “Is this not that […] sly, rubbish, mug company again? I ran into it two years ago, since then they have kept sending me emails about my overdue debts”.
In response, the real estate company brought a civil action, claiming an infringement of its right to a good reputation. While the claim as regards the opinion piece itself was dismissed by the Hungarian courts, MTE and Index were nevertheless found to be liable for the dissemination of the defamatory comments of others.
Crucially, the Hungarian courts refused to apply the safe harbour provisions of the EU’s E-Commerce Directive, instead finding that the Hungarian implementation of these provisions limits their reach only to electronic services of a commercial nature, in particular to purchases made via the internet. According to the Hungarian courts, the comments in question were private utterances, made outside the sphere of economic or professional activities or public duties, and therefore ineligible for safe harbour protection.
A developing line of case law: Delfi v Estonia
The issue of the liability of online service providers for unlawful information exchanged through use of their services by third parties first came before the ECtHR with the well-known Delfi case. This concerned the imposition of liability on Delfi, a popular Estonian online news platform, for the anonymous comments of its readers. Delfi contested its liability, claiming protection under Article 10 ECHR.
In that case, both in the initial Chamber judgment in 2013 and in its Grand Chamber last year, the Court found no violation of Delfi’s freedom of expression. In reaching this conclusion, the Court emphasised a number of elements, including the fact that the comments were posted in reaction to an article published by the applicant company, the fact that the applicant company was a professionally managed news portal run on a commercial basis and the moderate sanction of only EUR 320 in damages imposed upon it by the Estonian courts. The Court also highlighted the nature of the user-generated comments at stake, which were qualified by the ECtHR as hate speech and speech that directly advocated acts of violence. On this basis, the Court decided that they were “manifestly illegal”, making the measures taken by Delfi (which included both a notice-and-take-down regime and a word-based filtering system) insufficient. In particular with regard to notice-and-take-down, the Court concluded that:
“If accompanied by effective procedures allowing for rapid response, this system can in the Court’s view function in many cases as an appropriate tool for balancing the rights and interests of all those involved. However, in cases such as the present one, where third-party user comments are in the form of hate speech and direct threats to the physical integrity of individuals, as understood in the Court’s case-law […], the Court considers […] that the rights and interests of others and of society as a whole may entitle Contracting States to impose liability on Internet news portals, without contravening Article 10 of the Convention, if they fail to take measures to remove clearly unlawful comments without delay, even without notice from the alleged victim or from third parties.”
The nature of the comments was to prove especially decisive to the outcome of MTE v Hungary a few months later.
MTE and Index v Hungary: the ECtHR’s assessment
In MTE, as in Delfi, the ECtHR followed its by now well-trodden route for the assessment of interferences with Article 10 of the Convention: in order to be permissible, an interference must (a) be “prescribed by law”; (b) have one or more legitimate aims as laid out in paragraph 2 of Article 10; and (c) be “necessary in a democratic society” (see para. 46).
The first two of these conditions were easily despatched. The “legitimate aim” justifying an interference with freedom of expression was the protection of the rights of others, in this case, the personality rights of the estate agency (para. 52). A small hiccough presented itself with regard to the “prescribed by law” criterion, given that – arguably – the Hungarian understanding of the E-Commerce Directive’s safe harbour provisions misinterprets and misapplies them. The ECtHR nevertheless dismissed this concern, stating that the existence of the provisions of the Hungarian Civil Code on personality rights made it foreseeable, to a reasonable degree, for large online service providers, that they could, in principle, be held liable under domestic law for unlawful comments of third-parties (para. 51).
As a result, the bulk of the analysis, as is usually the case, concerned the final question of whether the interference was “necessary in a democratic society”. In this regard, given that the right to the protection of reputation is a right which is protected by Article 8 of the Convention as part of the right to respect for private life (para. 57), the case was interpreted as concerning a conflict between competing Convention rights: on the one hand the providers’ freedom of expression under Article 10 of the Convention and, on the other, the estate agency’s privacy rights under Article 8 of the Convention (para. 58). Given that, as a matter of principle, both these rights deserve “equal respect”, the case required recourse to the so-called principle of a “fair balance” (para. 58-59).
Crucially, before proceeding with the balancing exercise, the Court made an observation that set the tone for the rest of the judgment: according to the Court, the comments in question in the case at hand were not manifestly unlawful. Certainly, they constituted neither hate speech nor incitement to violence. Additionally, while the domestic courts had found that the case concerned the violation of the personality rights of the plaintiff company, the Court noted that, under its own case law, only natural persons may hold personality rights. A legal person may have commercial reputational interests, which may be worthy of protection, but these are devoid of any moral dimension (para. 65-66). As a result, the comments at hand merely qualified as offensive and vulgar (para. 63-64).
On this basis and by reference to its previous balancing case law (in addition to Delfi, Von Hannover v Germany (no. 2), Axel Springer v Germany and Couderc v France were drawn on for inspiration), the Court identified five factors as relevant for the assessment of whether a “fair balance” has been struck by the Hungarian courts in the material case (para 68-69). These were the following:
- the context and content of the impugned comments;
- the liability of the actual authors of the comments;
- the measures taken by the applicants and the conduct of the injured party;
- the consequences of the comments for the injured party;
- the consequences for the applicants.
Each of these criteria was analysed individually in detail:
- The context and content of the comments. As regards the context of the comments, the Court noted that the article in response to which they were posted concerned a matter of public interest and could not in itself be said to have provoked them. In addition, while one of the applicants was a large commercial online news portal, the other was not, meaning that its publication of professional content was unlikely to provoke heated online discussions. As regards the content of the comments, the Court repeated its refusal to qualify them as defamatory, and instead labelled them value judgments and opinions brought on by personal frustration. Moreover, although the comments were offensive and one of them was vulgar, the Court noted that style constitutes part of an expression and is therefore protected alongside it. In the material case at hand, the style used, although of a low register, was common in communications on internet portals, thus mitigating its negative impact (para. 72-77).
- The liability of the actual authors of the comments. The Court noted that the national courts did not sufficiently consider the liability of the authors of the comments. Instead, they were satisfied that the liability belonged with the applicant companies without embarking on any proportionality analysis, as these had, in contravention of the Hungarian Civil Code, “disseminated” defamatory statements. Yet, according to the ECtHR, the applicants’ conduct in providing a platform for the expression of others was a journalistic activity of a particular type, while, in line with the Court’s previous case law (see in particular Jersild v Denmark), the punishment of a journalist for assisting in the dissemination of statements made by another person should not be envisaged without strong justification (para. 79).
- The measures applied by the applicant company and the conduct of the injured party. The applicants had had a number of measures in place for the prevention of defamatory comments on their platforms: both applicants had included a disclaimer in their general terms and conditions stipulating that the authors of comments – rather than the applicants – were to be the ones liable for those comments; they prohibited the posting of comments injurious to the rights of third parties; the second applicant had set up a team of moderators performing partial follow-up moderation of comments posted on its portal; and, finally and most importantly, they both had a notice-and-take-down system in place, whereby anybody could indicate unlawful comments to the service provider so that they may be removed. The Court held that there was no reason why this notice-and-take-down system could not be seen as sufficient. Indeed, while the domestic courts had held that filtering was necessary in addition, the Court stated that this would amount to “requiring excessive and impracticable forethought capable of undermining freedom of the right to impart information on the Internet” (para. 82).
- The consequences of the comments for the injured party. Re-emphasising the difference between the reputation of an individual and the commercial reputational interests of a company, the ECtHR observed that the domestic courts had failed to evaluate whether the comments had reached the requisite level of seriousness and whether they had actually caused real prejudice. In any case, the Court observed that there were already ongoing inquiries in Hungary at the material time into the plaintiffs’ business practices, meaning that the comments in question were unlikely to have any additional significant impact on the attitude of consumers towards them (para. 84-85).
- The consequences for the applicants. Finally, the Court noted that, while the applicant companies were only obliged to pay courts fees and no awards were made for non-pecuniary damages, it could not be excluded that the finding against them might produce a legal basis for further legal action resulting in the imposition of damages. The decisive question, therefore, when assessing the consequence for the applicants was not the absence of payable damages, but the manner in which internet portals can be held liable for third-party comments. According to the Court, such liability may have foreseeable negative consequences on the comment environment of an internet portal, for example by impelling it to close the commenting space altogether. This may result in a considerable chilling effect on online freedom of expression (para. 86). The lack of any consideration whatsoever by the Hungarian courts for freedom of expression – or indeed, for balancing of any kind – was found to be particularly worrisome.
In the final analysis, the Court confirmed the Grand Chamber conclusion in Delfi according to which, if accompanied by effective procedures allowing for a rapid response, notice-and-take-down-systems can function in many cases as an appropriate tool for balancing the rights and interests of all those involved in a given intermediary liability dispute. While in Delfi however this rule was found to be inapplicable, as the contested comments constituted hate speech and direct threats to the physical integrity of others, thus allowing Contracting States to impose liability on internet news portals when they fail to take measures to remove clearly unlawful comments without delay, even without notice from the alleged victim or from third parties, no such utterances were found to be at issue in MTE, making the imposition of a stricter standard unjustifiable. Instead, the rigid stance of the Hungarian courts was found to reflect a notion of liability which effectively precludes the appropriate balancing between the competing rights according to the criteria laid down in the Court’s case law. As an ultimate result, a violation of the two internet platforms’ freedom of expression was established.
In ancient Greece, the oracle of Delphi was notorious for speaking in riddles. While appropriate for the priestesses of apocryphal religions, this approach is perhaps less suitable for one of Europe’s highest courts – nevertheless the ECtHR seems to have taken the name of its leading case in the area of intermediary liability a little too much to heart: with two judgments that directly contradict each other now handed down one only a few months after the other, a clear line of case law is nowhere to be found.
So, while the MTE Court does a brave job of hiding the fundamental incompatibilities, paying lip service to respecting precedent, the criteria relied upon in that case and in Delfi are notably different: although both apply balancing for their resolution, the Delfi court only considered 4 criteria, while MTE increased these to 5½, adding the factor of the consequences of the comments for the injured party and considering, as part of its first condition, not only the context, but also the content of the comments. Although these divergences can of course be put down to an evolving understanding of a complex issue by the Court (after all, slightly different criteria were also considered in Von Hannover (No. 2) and in Axel Springer), they are particularly noteworthy given the essentially identical nature of the two cases.
More importantly, entirely different weight is given to the factors considered in the analysis of the identified criteria. In Delfi, the fact that the applicant company was “a professionally managed Internet news portal run on a commercial basis which sought to attract a large number of comments on news articles published by it” was given great consequence, as was the substantial degree of control it had over those comments and its economic interest in their posting – yet these facts are hardly mentioned in MTE, despite being equally true of at least the first applicant in that case as well. Instead, the protection the second applicant attracts is treated as extending to the first, so that identical consequences attach to both platforms.
Similarly, while in Delfi a fine of EUR 320 was deemed perfectly proportionate, in MTE the absence of any award for damages whatsoever was found to be excessive, the principle of liability itself being interpreted as an excessive threat to freedom of expression in its own right.
The similarities between the facts of the two cases having been ignored or dismissed as irrelevant, the MTE Court instead focuses intently on the one difference that separates them: the nature of the user-generated comments. As in Delfi so in MTE, the contested comments had been labelled by the national courts as defamatory. The ECtHR altered this designation, though in each case in a different direction: in Delfi the comments were qualified as hate speech and speech that directly advocates for acts of violence and in MTE as merely offensive. Certainly, differences between the comments in the two cases are discernible, with some of the Delfi comments in particular standing out as clearly anti-Semitic. Yet, if this was the turning point, a more detailed analysis would have been helpful. After all, as the dissenting Delfi judges had already pointed out, it is telling that illegality that is professed to be so manifest should be qualified so differently by the national courts and by the ECtHR. And, while the Court notes that comments such as those at issue in MTE are commonplace online, any frequenter of internet discussion forums might well point out that that would sadly hold equally true of even the worst of the Delfi comments – was normal online behaviour therefore the actual criterion applied? The jumbling up of multiple comments of very different individual natures in both cases doesn’t help in this regard.
Regardless, reading the two judgments together, the main take-away seems to be a calibration of permissible intermediary liability rules depending on the type of content involved: hate speech and incitement to violence may be more sharply dealt with than merely offensive comments without fear of human rights’ violations. Where the line lies between these two categories has perhaps always been a tricky question and a problem the online context cannot be expected to fix, but only to exacerbate. What remains entirely unclear however is what lies in between: what should the fate of other types of illegality be? What if the MTE comments had, for example, qualified as defamation after all, while still failing to constitute hate speech in the eyes of the Court? The answer to this question remains entirely obscure.
More importantly, the elephant that set up residence in Delfi remains firmly lodged in place in MTE, with the new judgment doing nothing but shine an extra spotlight on it: according to both judgments, in cases of hate speech or other unprotected expressions, notice-and-take-down may be deemed insufficient and providers may be ordered by state authorities, without any fear of contravening Article 10 of the ECHR, to take measures to remove clearly unlawful comments without delay, even without notification from others. This is first established in Delfi, and, while everything else in the substance of MTE goes in the opposite direction, the principle itself is re-affirmed in that judgment as well, the difference in the nature of the comments concerned being put forth by the Court as the reason the two cases are allowed to diverge.
But if that is the case, how is the provider of services used to post hate speech or otherwise manifestly unlawful content to know that it is there, given that the content did not originate with itself and it must remove it before receiving notification from somebody else? This practical consideration remains unaddressed, while, if anything, the MTE judgment emphases the impossibility of the position in which the Estonian courts and the ECtHR had previously placed Delfi: a demand of filtering, MTE pronounces, “amounts to requiring excessive and impracticable forethought capable of undermining freedom of the right to impart information on the Internet.” This is indubitably true – but in Delfi the provider had implemented filtering measures and even these were deemed insufficient by the Court. In the final event, Delfi was in fact forced to deal with the problem by setting up a dedicated team of human moderators, an interference arguably far greater than the machine oversight of automatic filtering and certainly equally representative of “excessive and impracticable forethought”. Does limiting the effects of that ruling only to hate speech make a difference? It should be recalled that jumpy internet service providers concerned for their own liability do not make the best judges of illegal content, while the decision as to what qualifies as hate speech is, far from manifest, a difficult one for even the best-trained lawyers. Human moderation is, finally, not a measure all providers will be able to afford, while, under the Delfi rule, if it fails to correctly identify “manifestly unlawful content”, liability remains a threat – in which case, the closure of online commentating spaces that MTE warns against remains the only option and one that will inevitably affect not only the freedom to provide information that amounts to hate speech, but all information. Indeed, neither filtering nor human moderation can be contained to only hate speech either, both necessarily involving the examination of the totality of content on the provider’s services and accordingly having a deleterious effect on freedom of expression of non-hate speech as well. The insulation the Court seems to rely on that permits filtering for one type of content while disallowing it for another does not exist.
Why would the Court undermine its own case law so blatantly? It is tempting to put these inconsistencies down to internal ECtHR disagreements. The names of the judges sitting in the Chamber provide tantalising intimations: both Judges Sajó and Tsotsoria, whose dissenting opinion in Delfi so adroitly attacked all its many flaws, sat in the Chamber that considered MTE. They clearly disapproved of Delfi, but seem to have attempted to do the best they could with the limiting precedent it imposed upon them. The concurring opinion of Judge Kūris is another curious clue in this regard: apparently aware of the strangeness of this surreptitious turn in the case law, the judge warns that the MTE judgment “although it may now appear to some as a step back from Delfi AS, will prove to be merely further evidence that the balance to be achieved in cases of this type is a very subtle one.” It is perhaps to be expected that in its search for this subtle balance the ECtHR would have to deliver some misfires – hopefully, however, it will eventually realise that it was Delfi and not MTE that represents the most worrying “step back”.
Regardless, perhaps the problem that is made most obvious by both cases is the flawed understanding that the national courts of EU Member States seem to have of settled EU law. Both Estonia and Hungary are EU Member States. Are the domestic courts of the EU really so badly informed as to the conditions of the E-Commerce safe harbours? Why would the Hungarian implementation of the E-Commerce Directive restrict its application in a way in no way envisioned in the original text? While Delfi concerned a case that started in 2006, i.e. before the CJEU had had the opportunity to shed light on certain unclear areas, the MTE case was decided by the Hungarian Constitutional Court as recently as 2014. It is of course not the job of the ECtHR to fix internal EU miscommunications. The EU legislator itself however should take heed: clearly, regardless of how things develop on the European human rights front, greater guidance is necessary on the modalities of EU intermediary liability law. Recent communications from the Commission have hinted at the potential further development of EU intermediary liability law. These cases are evidence of how much this is needed.