Abstract
The latest Grand Chamber decision of the European Court of Human Rights in Sanchez v. France makes the previous Delfi test absolutely unpredictable. This article explains why the uncertainty now concerns almost every single aspect of this test, and why case law hardly offers any guidance on the most basic questions. It is argued that with Sanchez, the Strasbourg case law on liability for the speech of others online officially descended into chaos without a proper sense of direction. Grand confusion about Sanchez now has the potential to threaten legal certainty introduced by EU law, as illustrated by its application in Zöchling v. Austria. Despite the Court's proclaimed deference to national law and increased use of the subsidiarity principle, it is striking that democratically adopted European legislation about digital services has been ignored for so long in Strasbourg. Sanchez now raises a serious prospect that the ECtHR is on a collision course with the EU's newly adopted legislation, the Digital Services Act, that builds on the last 20 years of rules.
Introduction
One of the central questions of the regulation of the Internet is who can be held responsible if something goes ‘wrong’ online and how.
Imposing strict liability upon intermediaries who facilitate discussions would lead to strict control of individual speakers and their expressions. Such editorial control is incompatible with the idea of the Internet as a space that enables individuals to meet and discuss online without seeking any permissions to publish. Not imposing any liability upon intermediaries, on the other hand, submits potential victims to the mercy and whims of intermediaries over how they handle such non-editorial speech.
For the European legal tradition, neither is acceptable because it either empowers the state or private power too much at the expense of the fundamental rights of individuals. The ECtHR has thus so far embraced the Internet as a new medium with speakers who do not have editors, while also refusing to accept that victims of online abuse are left powerless. 2
Navigating the middle between the two extreme positions is proving increasingly difficult for Strasbourg judges. After all, the heightened levels of hate and vitriol in society are most visible online. Such developments cannot leave cold anyone conscious of the recent European history. In addition, the societal debate increasingly favours quick, simple and radical measures instead of slow, complex and limited responses.
However, we must be careful not to overreact by slowly undermining the very foundation of the system we are trying to defend. Freedom of expression is not only a freedom for ‘good times’. If such freedom cannot withstand the tides of extreme weather, it is not worth much as ‘one of the essential foundations of [democratic] society’ or ‘one of the basic conditions for its progress and for the development of every man’. 3
This article is about one instance where we believe the Grand Chamber of the European Court of Human Rights has got it wrong. Sanchez ruling offers a quick, simple and radical solution: hold politicians liable, including in criminal law, for failing to remove hateful comments of others, even if they did not incite or were not aware of them. Unfortunately, this solution is also very easy to misuse against those who use their freedom of expression to fight hate by persuading, campaigning or otherwise holding fearmongers to account. There are alternatives to Sanchez. One of them, which is slower, more complex and limited but is less prone to abuse, is represented by the model offered by the EU Digital Services Act, a law adopted several months before Sanchez was handed down.
This article proceeds as follows. It first places the Sanchez v. France judgment into the perspective of the broader pre-existing case law and legislative landscape. Next, it analyses the key elements of the judgment, especially how it looks at the question of knowledge. Finally, we articulate seven key concerns that follow from these developments.
Background
Long before the European Court of Human Rights was called upon to decide the first cases, the EU legislatures adopted rules about when providers can be held liable for others.
Legislative landscape
In 2000, the European Union adopted an all-encompassing horizontal set of rules in the E-Commerce Directive (ECD). 4 These rules created a set of liability exemptions to clarify when providers of certain digital services cannot be held liable for other people's content or actions. Much of the application layer of digital intermediaries, such as social media, marketplaces or discussion forums, thus started benefiting from legal clarity: if they do know about the illegality of specific content, they cannot be held liable together with third parties as long as they swiftly act upon becoming aware of such illegality. 5 Thus knowledge-based choreographies, under which victims usually notify illegal content and providers evaluate it and take it down, became the norm. In addition, the framework prohibited indiscriminate general monitoring of people and content. Thus the imposition of obligations on providers to generally verify content, vet its creators or generally know about everything that happens on their services was outlawed. The goal of the rules was to support freedom of expression and privacy online and embrace decentralized non-editorial content. The EU and the United States inspired many other countries to start granting conditional immunity – liability exemptions that require at least providers’ knowledge of others’ actions to expose them to liability for those actions. 6
Twenty years and numerous cases and controversies later, the legal framework was updated in 2022 via the Digital Services Act (DSA). 7 The liability exemptions were preserved. To improve the enforcement of the rights of victims, the DSA introduced new self-standing due diligence obligations that mostly try to discipline how companies run their notice-based content moderation systems and design their services. They also create an ecosystem of actors that help victims to enforce their rights and encourage standardization and responsible use of automation. In addition, for a subset of companies – mid-sized platforms, but mostly for very large online platforms – the DSA creates a new regime of risk management. 8 Thus, 20 years later, the EU legislature continues on the course of keeping the liability of digital services limited but rather focuses on how to increase accountability of such companies around how they make the decisions and design their systems. Such a systems-oriented approach exposes them to some a priori legal obligations, but not to liability for individual pieces of illegal content.
This targeted and tiered approach is only possible because companies are a priori viewed as not liable for what they distribute. 9 Had they been strictly liable before they obtained knowledge, they would need to vet all content from all users upfront all the time. This would basically make decentralized non-editorial content very costly and thus unlikely. Thus specific knowledge of illegality is one of the key differentiators between being a mere intermediary that stores content and being potentially culpable along with perpetrators. Recently, the Court of Justice of the European Union (CJEU) even increased the bar of knowledge to that of manifest illegality in order to discourage overblocking of innocent lawful content in YouTube. 10
From Delfi v. Estonia to Sanchez v. France
The starting point of the Strasbourg (and Luxembourg) case law seems to be the specificity of the Internet. As mentioned above, for the ECtHR, digital services facilitating user-generated content must be subject to ‘graduated and differentiated’ regulation that differs from the regulation of editorial media. 11 The first Grand Chamber case, Delfi v. Estonia (2015), 12 concerned the situation where comments under news articles contained hate speech and other ‘clearly unlawful speech’. 13 The authors of these comments were anonymous, and once the comments were posted, only the platform could remove them. It was common ground between parties that any liability imposed on the platform for user-generated content constituted an interference with its rights under Article 10. 14
In its decision, the Grand Chamber set a broad margin of appreciation for Member States. It devised a balancing test and ultimately expressed that professional news websites can be held liable under national law if they failed to scrutinize content in the comments even before they are notified about such comments by victims. The judgment seemed to have implied that the Council of Europe's Member States, many of whom were also EU members, would be free to impose monitoring obligations on platforms concerning third-party content.
The follow-up case law of the Court did a lot to clarify and refine the approach, although Delfi did not entirely disappear as a framework of reference. 15 The general approach of the Court was rewritten by numerous Chamber decisions that applied the Delfi test. The most important of them is MTE v. Hungary (2016). 16 In this case, the Court acted much more prescriptively to give effect to Article 10 rights. It endorsed as ‘ECHR-necessary’ the liability model (but not the details), adopted by both the ECD and DSA, which is based on the idea of shared responsibility for the harms and wrongs in the digital ecosystem. For platforms hosting content, this shared responsibility is known as the notice and takedown (or action) choreography.
In MTE, the Chamber of the Fourth Section of the Court rejected far-reaching monitoring duties due to their chilling effect on the freedom of expression on the Internet. It ruled that ‘the notice-and-take-down-system could function in many cases as an appropriate tool for balancing the rights and interests of all those involved’. 17
In contrast to Delfi, the comments in MTE did not amount to speech inciting hate and violence. More importantly, the users were able to modify their comments once they were published. The authors of comments thus had more agency to do the right thing on their own. In addition, the for-profit criterion that weighed against the provider in Delfi, remained unaddressed. 18 In Delfi, the case concerned a large, professionally managed Internet news portal that ran on a commercial basis, published news articles of its own and invited its readers to comment on them; the Grand Chamber expressly stated that it did not concern other fora on the Internet, such as social media platforms. 19
The test established in Delfi to determine the lawfulness of facilitator liability under Article 10 was subsequently reformulated in MTE as the following criteria:
the context of the comments, (ii) the measures applied by the applicant to prevent or remove the comments, (iii) the liability of the actual authors of the comments as an alternative to the facilitator's liability, (iv) the consequences of the domestic proceedings for the applicant, and (v) the consequences of the comments for the injured party.
20
MTE soon became the general approach that was endorsed by numerous subsequent cases, including those filed on the basis of Article 8, such as Pihl v. Sweden (2017),
21
Tamiz v. United Kingdom (2017)
22
and Høiness v. Norway (2019).
23
In all these cases, the applicants tried to argue that a State was in breach of its positive obligation under Article 8 as it failed to protect their right to reputation by not holding intermediaries liable for third-party comments. Although just Chamber judgments, all three cases support the reading that Delfi's broadened margin of appreciation applies only when speech includes hate or similarly clearly unlawful speech. The commercial character of the platforms constantly seems to be de-emphasized and Delfi's role of eliciting comments in the context of its professional editorial activities is emphasized instead. Then comes Sanchez v. France, which upends all prior case law.
Two Grand Chamber judgments
Grand Chamber decisions are intended to clarify the law. Unfortunately, the European Court of Human Rights (ECtHR) has failed for the second time to provide conceptual clarity about the constraints that human rights law places upon the States’ ability to regulate the most important digital services – those that carry user-generated content.
The first and initial failure of the ECtHR took place in 2015 when the Grand Chamber of the Court decided Delfi v Estonia. That failure was relatively modest compared with the latest one. In Delfi, the Court allowed states to impose liability for damages upon intermediaries, such as newspaper portals, that facilitate online discussions between strangers, even if they were not notified about manifestly illegal hate speech. Yet the Court did not require that the states do so. In other words, in Delfi, the Court only stated that states may hold platforms liable for the speech of their users in compliance with Article 10 of the European Convention on Human Rights (ECHR), but not when they must do so based on Article 8 ECHR.
The Delfi decision lacked conceptual clarity and direction of travel. It was lamented by those who disagreed with the Court's broad deference to national legislatures to limit the protection of speech interests. But it could have been excused as an early misstep caused by an attempt to be very cautious not to prescribe solutions to the hate speech phenomenon from the bench. Moreover, technically, the ruling should not have had much impact in the European Union because the EU law curtailed the choice the Court gave to states.
But the dismay of civil society and general confusion felt after Delfi was so significant that a former president of the ECtHR, Robert Spano, in 2017 wrote an article arguing that the decision is an early start and ‘the case is to some extent unique and unsuitable as a basis for broad interpretive conclusions over and above the facts presented by the case’. 24 He argued that it represents a middle ground between supposed non-regulation and regulation of the Internet. He emphasized that, ultimately, the ruling only speaks to the question of what states may do, because ‘States are not precluded from adopting more stringent forms of intermediary liability when confronted with the gravest forms of negative online comment’. 25 Finally, he argued that, despite the Court's acceptance of civil liability regardless of one's knowledge about the illegal nature of third-party comments as compatible with Article 10, the Court recognized ‘the beneficial value of the notice-and-take-down system as a suitable mechanism for balancing the implicated interests’. 26
Between 2016 and 2023, multiple Chamber judgments confirmed the ‘may’ reading of Delfi. 27 The Court continuously refined and explained its position. By 2023, one could start seeing the contours of the Court's approach: for hate speech and other clearly unlawful speech, the Court gave the States the ability to experiment, but for other cases, imposing liability on digital intermediaries before they acquire knowledge about specific illegality would breach their right to freedom of expression. After all, it was not their own speech for which they were held liable, but that of other people.
Unfortunately, the Grand Chamber judgment in Sanchez v. France (2023) muddied the waters again and arguably made things much worse. 28 It completely flipped the established reading of Delfi and put into question the existence of any consistent legal test underlying the case law. The combination of procedural deference and vague test is putting the ECtHR's approach at risk of colliding with EU law.
According to EU law, many of the services whose operators come before the Court constitute protected hosting services: discussion forums, web hosting services, newspapers hosting comments or operators of social media profiles. They all store other people's information at their request. The inability of the courts on the national level to properly apply the Union framework is the common feature of cases like Delfi and Sanchez. In the former case, the Estonian courts erroneously misinterpreted the concept of neutrality and thus denied the exemption. In the latter case, the French courts failed to appreciate that a user of a digital service can become a host of other people's information, too. 29 Arguably, on proper application of the EU law, based on what is known from the rulings, neither of the two cases should have raised the questions they did. Both facilitators were hosting services that should have enjoyed the liability exemptions vis-à-vis content of others they were not aware of. 30
It is therefore very worrying that in a recent case where the Austrian courts correctly interpreted the EU law on liability exemptions, the ECtHR now intervened to inform national judges that they did it all wrong because they did not go beyond the statutory framework and engage in further balancing. In Zöchling v. Austria, 31 a judgment by a Committee, 32 co-written by two judges from the Sanchez majority, three judges unanimously applied Delfi and Sanchez to demand more proportionality assessment. However, the national law based on EU law was rather clear: for liability purposes, for the time before the platform knows, there is no need to engage in such balancing. Sanchez thus now raises a serious prospect that the ECtHR is on a collision course with the EU's newly adopted legislation, the Digital Services Act, that builds on the last 20 years of rules.
The Strasbourg Court does not display awareness of the context within which it is handing down its decisions. We explain why, despite the technological and legal complexity of the matter, the Court seems to have abandoned its deferential approach and, without examining the democratically adopted legislation on the matter in most of the States, 33 prescriptively plunged into the debate about online liability without much care.
Sanchez v. France: Key elements
On 24 October 2011, Mr Sanchez published a post on his public Facebook wall. On the same day, two users, S.B. and L.R., commented thereon in a manner that the Court considered as ‘clearly unlawful’. 34 One of the comments referred to the partner of Mr Sanchez's political opponent F.P., Ms Leila T. On 25 October 2011, Ms Leila T. became aware of the comments and confronted one of the authors, S.B. Following this confrontation, S.B. deleted his comments and informed Mr Sanchez of his exchange with Ms Leila T. On 27 October 2011, Mr Sanchez posted a ‘warning message’ on his Facebook wall, urging his followers to ‘be careful with the content of [their] comments’. 35 At this point, only L.R.'s comment was still visible. On 23 January 2012, L.R. was interviewed and stated that he had deleted those comments in which Mr Sanchez's political opponent F.P. could recognize himself or be recognized by others. 36 On 28 January 2012, Mr Sanchez stated that he had only become aware of S.B.'s comment after its deletion and that he only found out about L.R.'s comments when summoned for his interview. He stated that the large number of comments he received made it impossible for him to continuously monitor his Facebook wall. 37
The French decisions
On 28 February 2012, the Nîmes Criminal Court convicted Mr Sanchez under Section 23(8) of the Law of 29 July 1881 and Section 93-3 of Law no. 82-652 of 29 July 1982 (‘the 1982 law’). In its judgment, the Nîmes Criminal Court referred to a Constitutional Council decision, 38 explaining that holding a producer criminally liable for online content based on Section 93-3 of the 1982 law was only constitutional if ‘it could be established that the producer had been aware of their content before they were posted, or otherwise where he or she failed to act promptly to delete the comments at issue before becoming aware of them’. 39 A later decision of the Court of Cassation in another case clarified that criminal liability could also be engaged with respect to comments where the producer ‘failed to act promptly to delete the comment at issue upon becoming aware thereof’. 40
The Nîmes Criminal Court rejected Mr Sanchez's submission that he had no knowledge of the impugned comments before their deletion without providing any evidence that would establish Mr Sanchez's knowledge of the comments. 41 Instead, it argued that Mr Sanchez had been ‘responsible for verifying the content of the comments’ on his Facebook wall and that he must have been aware of the greater likelihood of attracting polemical content due to his political activities, reinforcing his duty to monitor the comments he received’. 42 Thus the Nîmes Criminal Court seems to have based its conviction mostly on the idea that Mr Sanchez should have been aware of the comments his initial post had attracted.
The Nîmes Court of Appeal upheld the Criminal Court's judgment. In line with the lower Court's decision, it argued that since Mr Sanchez had decided to allow for user comments, he ‘became responsible for the comments posted’ on his wall. 43 His position as a politician had further required him to be ‘all the more vigilant’. 44 The Court of Appeal addressed the question of whether Mr Sanchez had actual knowledge of the comments, recalling Mr Sanchez's statement that he had been checking his Facebook page every day. 45 However, its reasoning on why Sanchez ‘cannot claim not to have been aware of the remarks’ was largely based on the previously mentioned objective factors (the public and political nature of his account), 46 reducing Mr Sanchez's alleged actual knowledge to an ‘aggravating’ factor. 47 Thus the Nîmes Court of Appeal, despite briefly addressing evidence relevant to assessing Mr Sanchez's actual awareness of the comments, seemed to have followed the line of the Nîmes Criminal Court, holding that Sanchez should have been aware of the impugned comments instead of focusing on evidence establishing beyond reasonable doubt that he was aware of them.
In sum, the French courts seemed to have argued that Mr Sanchez had a duty to monitor the comments he received and that he could be held liable even without actual knowledge, 48 thus endorsing a strict (objective) theory of criminal liability. 49 The French courts focused their discussion on objective factors, such as the public nature of the account and Mr Sanchez's political activity, seemingly to justify this move away from requiring actual knowledge. This approach is in stark contrast with how the CJEU instructs the national courts to review knowledge of other people's comments that would disqualify hosting providers from the liability exemption. The CJEU requires that actual or constructive knowledge concerns specific illegal acts or expressions. 50
The ECtHR's appraisal of Sanchez's knowledge of the comments
The ECtHR Grand Chamber addressed the question of knowledge in different parts of the judgment. It first discussed whether the fact that the ‘point in time from which the producer is deemed to have knowledge of the unlawful remarks’ was not specified in the 1982 law was incompatible with the requirement that the legal basis of the interference be ‘framed in sufficiently precise terms’. 51 It held that this was not the case and that it was permissible to leave the determination of said point to the domestic courts. 52 The Court then turned to the question of whether a system of prior notification was necessary for the law to be sufficiently precise and foreseeable in this respect. It cited Delfi, stating that ‘[t]he lack of a system of prior notification to the producer cannot (…) in itself raise a difficulty in terms of the lawfulness of the interference’. 53
Turning to the steps taken by Mr Sanchez, the Court first stated that the warning message and his conversation with S.B. demonstrated that Mr Sanchez had been aware that his post had attracted problematic comments. 54 Emphasizing that ‘the question whether Mr Sanchez had been aware of the unlawful comments (…) goes to the heart of the matter’, 55 it concluded that while the Nîmes Criminal Court had not endeavoured to assess if Sanchez had actual knowledge of the impugned comments, the Court of Appeal had clarified the matter sufficiently. 56
In addition to the facts considered by the domestic courts, the Court reflected on the number of friends Mr Sanchez had at the relevant time as well as the overall number of comments received in response to his post of 24 October 2011. It held that since the post had only attracted around 15 comments, ‘the question of the difficulties caused by the potentially excessive traffic on a politician's account and the resources required to ensure its effective monitoring (…) clearly does not arise in the present case’. 57
As Mr Sanchez was held criminally liable for both S.B.'s and L.R.'s comments and was ordered to pay civil damages to Leila T., the Court further had to address the question of whether this was justified, given that S.B.'s comment, the only one directly referring to Ms Leila T., was deleted by the author himself within 24 hours. The Court argued that since S.B.'s comments gave rise to further comments ‘echoing the remarks of S.B.’, all comments constituted an ‘ongoing dialogue, forming a coherent whole’. 58 Against this background, it concluded that the Court of Appeal's decision to hold Mr Sanchez liable for both S.B.'s and L.R.'s comments had been neither arbitrary nor manifestly unreasonable. 59
Hate speech under the ECtHR
Another relevant feature of Sanchez is the qualification of the impugned comments as hate speech. The Court examined this when it considered the requirement of necessity in a democratic society.
To determine when it might be necessary in a democratic society to hold an actor on the Internet liable for the comments (in this case for hate speech) posted by third parties, the Court resorted to the precedent in Delfi, 60 requiring it to examine, inter alia, the context and nature of the comments. 61 In addition, it was important to analyse the nature of the comment, the size of the entity, its connection to a profit-making activity and the likelihood that it would attract a large number of comments or would be widely read. 62
Discussing the context of the comments, the Court recognized that there was no universal definition of hate speech,
63
but it resorted to elements laid out in the case of Perinçek
64
to conclude that the comments at issue were indeed unlawful, which required it to consider:
the existence of political or social tensions; (ii) whether the comments could be regarded as a direct or indirect call to violence, or justifying violence, hatred and intolerance and; (iii) the manner the statements were made or their capacity to lead to harmful consequences.
65
The Court also referred to Le Pen v. France
66
to state that remarks capable of arousing a feeling of rejection and hostility towards a community could be deemed excessive and outside of the protection of Article 10.
67
Thus when politicians take a stance on social problems, they should avoid advocating for discrimination or resorting to ‘vexatious or humiliating remarks or attitudes’ that might bring up reactions from the public which could threaten a peaceful social climate and undermine confidence in democratic institutions.
68
Based on this idea, the Court considered that when addressing speech which could incite violence against an individual or a segment of the population, State authorities ‘enjoy a broader margin of appreciation’ to assess the necessity of the interference. 69 The same would apply to expressions which justify hatred based on intolerance. 70
In Sanchez, the Court held that, given the alleged cultural characteristics cited by the authors in their comments, they targeted a defined group of persons, namely Muslims. Secondly, the comments were associated with ‘insulting and hurtful language’. 71 The Court acknowledged the context of a political election and the parties expressing their discomfort, but it considered that precisely in this context, racist and xenophobic discourse could cause even greater harm. 72
Thus the Court considered that ‘when [the comments are] interpreted and assessed in their immediate context’ (meaning, being posted on a politician's Facebook wall during an election campaign), ‘[they] genuinely amounted to hate speech, in view of their content and general tone, together with the virulence and vulgarity of some of the language used’. 73 Consequently, the Court concluded that the comments were ‘clearly unlawful’. 74 The Court did not explain why those hateful comments of third parties should be attributed to Mr Sanchez, who never uttered them or instructed their publication.
Professionals and politicians as facilitators
In Delfi, the Court emphasized that the case involved a large professionally managed Internet news portal run on a commercial basis which was different to other fora on the Internet where third-party comments can be disseminated. 75 Thus previously, in Delfi and MTE, the Court considered what could be called the ‘professional (or commercial) criteria’. 76 This meant that the Court gave significant weight during its balancing exercise to the imposition of certain duties for online intermediaries based on the existence of a commercial structure to comply with such requirements and their expertise in the field. At all times, the Court addressed the role of the intermediary within the context of the comments, particularly their classification as hate speech. 77
In Delfi, the Court pointed out that the professional criteria imply that the entity in question, as it carries out an economic activity, should be well informed on the relevant legislation and case law, as they are able to assess the risks associated with their activities (those being activities which seemed to involve its likelihood to attract comments and engagement). 78
In Pihl, the Court dismissed a claim of denial of justice over a blog post claiming that the applicant was involved in a Nazi party by analysing the context of the comments: ‘the association [that runs the blog] is a small non-profit association, unknown to the wider public, and it was unlikely that it would attract a large number of comments or that the comment about the applicant would be widely read’. 79
In MTE, the Court considered that the imposition of liability over a non-commercial website may have a detrimental consequence on freedom of expression, with a direct or indirect chilling effect. 80 Therefore, whenever the online intermediary was a big enterprise, especially with the capacity to have great engagement, it seems that the professional criteria entered into the analysis of third-party effects.
However, in Sanchez, the role of the applicant as a politician came under scrutiny by the Court when addressing the necessity of the measure. The Court emphasized that Mr Sanchez's Facebook wall was not comparable to the large professionally managed Internet news portal run on a commercial basis as in Delfi. However, it is considered necessary to clarify that duties and responsibilities apply to politicians when they decide to use social media for political purposes. 81 Thus in Sanchez the Court seemed to create a new distinctive criterion.
In particular, the Court considered that ‘even though the applicant's situation cannot be compared to that of an internet news portal, the Court sees no reason to hold otherwise in the present case’. 82 In its discussion that duties and responsibilities may be required from a politician, the Court emphasized the likelihood of inspiring hateful comments, considering that a politician can attract a degree of notoriety and representativeness, especially in an electoral context. 83 The Court did not take into consideration the availability of any resources supporting the operations in the context of his social media activity.
The judgment can be read as introducing a new public persona factor, even in the absence of any commercial structure, since the notoriety and representativeness alone would appear to justify increased criminal liability for third-party comments. Thus even shoestring budget public personas, such as academics, activists or some influencers, could qualify under the new criterion.
Nonetheless, despite attaching greater responsibility to Mr Sanchez because of his status as a politician, the Court indicated that Sanchez did not attract much attention since the post received only about 15 comments. Thus the Court concluded that he could have monitored the content of the comments. 84 Moreover, the Court gave weight to the fact that Mr Sanchez had some skills in operating social media, 85 thus further merging the consideration about his politician status with professional character.
Analysis of the Court's approach: May or must?
The case of Sanchez provided the Court with an opportunity to determine the outer limits of liability for the manifestly hateful comments posted by third parties on a Facebook ‘wall’, especially because prior case law did not delve into the possibility of attributing the criminal liability of a user to a platform. 86 Before addressing the substance, it is important to note the prescriptive tone of the judgment. As noted earlier, in Delfi, the Court tended to act cautiously in the area of third-party liability for hate speech on digital platforms by allowing national experimentation. 87 In contrast, the Sanchez ruling makes it possible to discern a shift towards more prescriptive thinking. However, the Court still made some attempts to appear deferential.
The Court is known to sometimes employ the ‘process-based review’, a mechanism of foregrounding the principle of subsidiarity. There are ample features of this approach in this case. 88 Thus the Court did apply deference in its review towards domestic judicial and legislative processes; however, paradoxically, this led to a more prescriptive tone in the substantive review of the case. 89
The central confusion about the nature of the Court's approach stems from the following statement: The Court first observes that there can be little doubt that a minimum degree of subsequent moderation or automatic filtering would be desirable in order to identify clearly unlawful comments as quickly as possible and to ensure their deletion within a reasonable time, even where there has been no notification by an injured party, whether this is done by the host itself (…) or the account holder [emphasis added].
90
However, the fact that this prescription is couched within the discussion of lawfulness raises the question of its relevance to the more substantive assessment of the Court. Moreover, the Court outlined the principle that the power to interpret and apply domestic law is ‘primarily for the national authorities, notably the courts’. 93 It bears reminding that within the domestic judicial process, the Constitutional Council was understood by the domestic courts as having held that the criminal liability of a user ‘would only be engaged in respect of (…) comments where it could be established that the producer had been aware of their content before they were posted, or otherwise where he or she failed to act promptly to delete the comments at issue before becoming aware of them’. 94
In light of this domestic understanding of appropriate knowledge standards, it is hard to read Sanchez as a clear prescription to the states. Under a process-based review, such an example of domestic judicial dialogue tends to be rewarded with deference, 95 as it indicates that the domestic courts have taken seriously their duty to interpret and apply the law, and ‘to dissipate (…) interpretational doubts’. 96
In a similar vein, within its discussion on lawfulness, the Court noted that domestic law left the matter of the point in time from which the facilitator is deemed to have had knowledge of the unlawful remarks ‘to be decided by the relevant domestic courts on a case-by-case basis’. 97 Within the Court's discussion of the necessity of the interference, a similar sentiment is found in its observations about the quality of the judicial process in reviewing the point in time from which the facilitator is deemed to have knowledge of relevant comments. Thus the Court stated the domestic courts ‘gave reasoned decisions’; they made a ‘reasonable assessment of the facts’; they specifically examined Mr Sanchez's awareness of the unlawful comments; and the Court of Appeal ‘provided further factual clarification’ about whether the applicant had ‘actually known’ about the relevant comments at the time of the investigation. 98
In comparison to the nature of the Court's approach towards knowledge standards, the Court's approach towards the responsibilities of politicians is markedly more prescriptive. However, the extent of the Court's prescription is curtailed by the two features of the discussion regarding the applicant's status as a politician: the fact that aspects of a Delfi test still persist in relation to the applicant's status as a politician; and the use of process-based considerations to support the Court's analysis.
Sanchez is notably the first judgment in the area with a specific subsection entitled ‘existence of responsibility and limits not to be exceeded’. 99 The Court explicitly found it was ‘appropriate to proceed with a proportionality analysis based on the degree of liability that may be attributed to’ a person of ‘notoriety and representativeness’. 100 The fact that the Court attempts to structurally embed the applicant's status as a starting point for the proportionality analysis in Sanchez suggests a desire for a prescriptive approach towards the responsibilities of politicians in the area of third-party liability for online hate speech.
Consequently, the Court distinguished between the duties held by private individuals and local politicians. 101 Again, this distinction indicates a prescriptive approach, as it followed principled reasoning relating to the ‘certain resonance and authority’ of such persons. 102 This stands in contrast to the vague statements of principle pervading the Court's discussion of knowledge standards.
However, the prescription regarding the special responsibilities of politicians must not be overstated. The same applies to the extent to which these became a starting point for the proportionality analysis as a matter of substance. Even where the Court makes clear its decision to approach the question of the applicant's specific liability ‘in light of the “duties and responsibilities” attributable to politicians who make their social media accounts public for political purposes, the Court continues by noting that the applicant ‘has some expertise in the digital services field’. 103
Whatever the factual merits of this observation, 104 it is still relevant to note that the Court appeared unwilling to abandon entirely an examination of the extent to which a platform was professionally managed. 105 Similarly, when distinguishing between the liability that may be attributed to persons of greater notoriety and representativeness, the Court related this distinction to the fact that a national figure has a greater burden partly based on the ‘resources to which he or she will enjoy greater access in order to intervene efficiently on social media platforms’. 106 This suggests some continuity with the position of previous jurisprudence, where the Court's acceptance of third-party liability in this context rested primarily on whether Mr Sanchez was comparable ‘to a large professionally managed Internet news portal run on a commercial basis’. 107
Similarly, the Court occasionally reverted to process-based considerations within its approach towards the responsibilities of politicians on digital platforms, rather than informing its approach through a purely principled basis. Thus the Court found procedural shortcomings in the domestic treatment of the issue when outlining that the domestic courts ‘referred to the applicant's status as a politician and inferred from this that a special obligation was incumbent upon him’. 108 It stated that the Nîmes Court of Appeal should have referred to the fact that the specific duties required from politicians are indissociable from principles upholding their rights ‘in order to strengthen its reasoning’. 109 Moreover, the Court's conclusion on Mr Sanchez's specific liability was underlined by its consideration that ‘the Criminal Court and Nîmes Court of Appeal were best placed to assess the facts in the light of the difficult local context and their acknowledged political dimension’. 110
Despite these process-based mechanisms for applying the margin of appreciation, 111 it still appears that the Court aligned with the position of the domestic courts as a matter of substance. Indeed, the Court agreed that the ‘fact remains that the applicant was using his Facebook account in his capacity as a local councillor and for political purposes’. 112 Consequently, it endorsed the Chamber's finding that the election context could not serve to disguise or minimize the fact that the relevant comments incited hatred on the basis of religion. 113 Ultimately, this reveals a prescriptive approach towards the treatment of responsibilities of politicians on digital platforms under the Convention, albeit shrouded in some deference.
Thus we cannot for a certain answer whether the Court understands its Sanchez decision as another ‘may’ in the path started by Delfi, or as a newly demanded ‘must’. There are several strong reasons, however, why it seems that the Court intends its ruling to be read as a ‘must’, i.e., a mandatory standard to protect the rights guaranteed by Article 8 ECHR.
Seven reasons for concern about Strasbourg jurisprudence on intermediaries
The Court's approach in Sanchez creates several reasons for concern. 114 In this section, we highlight the main seven.
Reason 1: Why Sanchez cannot be equated with Delfi
The Court assessed the proportionality of Sanchez's impugned penalty by considering the Delfi test. 115 However, in Delfi, this test was applied to a situation concerning three types of actors, namely: (i) an Internet news portal being the facilitator; (ii) anonymous authors of unlawful comments; and (iii) victims of aforesaid comments. 116 This formed a triangle, with specific capabilities for each actor. 117
Accordingly, the Court adjudicated based on the asymmetric capabilities of the Internet news portal to delete third-party content, compared with the absent abilities of those anonymous third parties to do the same. 118 This asymmetry, coupled with the inability to identify the authors of the unlawful comments were major factors, inter alia, in the Court's finding that there had been no Article 10 violation. 119 Moreover, the Court couched the Delfi test within those idiosyncratic features by excluding ‘other fora on the Internet where third-party comments can be disseminated’, in particular, ‘a social media platform where the platform provider does not offer any content and where the content provider may be a private person running the website or blog as a hobby’. 120
In Sanchez, the Court similarly applied the established Delfi test, and ‘[saw] no reason to depart from that approach’. 121 However, the situation in Sanchez differed markedly from that in Delfi: (i) Facebook, the company, and Mr Sanchez, as its user, were two independent facilitators of online discussions; (ii) authors of unlawful comments were identifiable and even acted to remove their comments; (iii) Mr Sanchez, Facebook and the authors of comments were all able to delete unlawful comments.
By not discussing these facts, the Court failed to adequately compare the dissimilar parties and their respective triangulation of capabilities between the two cases, which may require amending the original criteria found in Delfi. Our third-party intervention argued that the Court should take these capabilities into account, and there was room for such considerations in the original Delfi test, but the Court did not address this. Nonetheless, by implementing the Delfi test without further explanation nor ‘translation’ to fit the Sanchez dynamics, the Court superficially closed the matter by stating that, ‘there should be a sharing of liability between all the actors involved’ and treated Mr Sanchez as a platform like Delfi. We suggested similar language, but to undertake the comparison of capabilities. 122 However, the Court mechanically applied Delfi without explicitly examining the differences between the two cases to adjust the legal tests.
The Court's usage of Delfi in the question of lawfulness of the interference in Sanchez is even more problematic. In paragraph 140, the Court reiterates Delfi dictum by stating: (…) Cases where third-party user comments take the form of hate speech (…) may entitle Contracting States to impose liability on the relevant Internet news portals (…) if they fail to take measures to remove clearly unlawful comments without delay, even without notice from the alleged victim or from third parties [emphasis added].
123
Lastly, the Court in Delfi adjudicated on the basis that the applicant was a ‘large professionally managed Internet news portal run on a commercial basis’. 129 It excluded from its assessment ‘other fora on the Internet where third-party comments can be disseminated’, in particular, a ‘social media platform where the platform provider does not offer any content and where the content provider may be a private person running the website or blog as a hobby’. 130 In Sanchez, the Court attempted to address this carve-out by stating that while there can be little doubt that Sanchez's ‘wall’ falls within the category of ‘other fora on the Internet where third-party comments can be disseminated’, it chooses to focus on Sanchez's ‘duties and responsibilities’ within the meaning of Article 10, 131 and when doing so, it simply applied the Delfi test, despite stating that Sanchez was different. 132
In short, the Court failed to appreciate how attributing liability for others differs in criminal law from civil law, how the resources of a large corporation differ from a political campaigning on social media by a candidate, how being a platform differs from being a user on a platform (even if both facilitate discussion) and how using a service does not equal knowing everything that happens within it.
Reason 2: Why do facilitators’ capabilities matter?
The capabilities of the various actors in Delfi and Sanchez were markedly different. By overlooking how the digital ecosystem functions, not only did the Court fail to adequately ‘translate’ the Delfi test, but it also missed the opportunity for reflection on how different actors possess different capabilities and liabilities. The Court placed the onus solely on the user of the account and not the other facilitators involved, namely the platform, in this case Facebook, or authors of the comments. Even the Delfi test considered these two options. But the Court failed to do so in Sanchez.
The Court entirely overlooked that the technical capabilities of Sanchez were limited to the tools provided by Facebook. Although Facebook itself now boasts a comprehensive notice-and-take-down system, 133 deploying both people and technology, 134 Mr Sanchez, at the relevant time, did not have comparable tools, such as an automatic filtering system, to manage the comments. 135 In fact, the Court itself recognized this and stated that monitoring of all these comments would ‘require availability or recourse to significant, if not considerable [economic] resources’. 136 Yet it insisted on exactly that.
The Court justified this by stating that ‘to exempt producers from all liability might facilitate or encourage abuse and misuse, including hate speech and calls to violence, manipulation, disinformation and lies [emphasis added]’. 137 It continued to state that there has to be a ‘sharing of liability’ between the actors involved, ‘allowing if necessary for the degree of liability and the manner of its attribution to be graduated according to the objective situation of each one’. 138
The recognition of shared liability could have also helped the Court draw outer limits of the ‘measures taken’ under the Delfi test. The jurisprudence on Article 10 recognizes that notice and takedown is an ‘appropriate balancing tool’ for the rights of the platform, the content creator and the victim. 139 The Court in Sanchez, however, did not delve into whether the notice and takedown choreography was available to Mr Sanchez, but instead, it focused on Mr Sanchez's act of ‘making access to the wall of his account public’ as one of the ‘measures taken’. 140
The Court could have remained within the confines of the notice and takedown choreography as an effective measure and subsequently assessed whether the measures taken were sufficient by acknowledging the disparate tools available to the different actors in the social media ecosystem. Moreover, by drawing such outer limits based on the respective actors’ capabilities, the Court could have then more comprehensively assessed Mr Sanchez's measure of posting a warning message as an adequate step taken under the Delfi test.
Reason 3: Can people be attributed expressions they are unaware of?
The Court's ruling in Sanchez failed to grapple with the most fundamental issue. The Court was not ruling on the expressions of Mr Sanchez or his aides, but on those of unrelated third parties, and that Mr Sanchez only failed to remove them after they were posted regardless of his subjective knowledge of them. The ruling suggests that politicians operating on social media might be attributed expressions they do know about, even if they did not instruct or incite them. The reason for this is that the knowledge standard, which usually acts as the attribution device in such cases, has completely disappeared from the Court's analysis, replaced by a peculiar analysis of an alleged ‘ongoing discussion’.
On one hand, it was settled that French law requires a producer to have knowledge of the impugned comments to be held liable for a failure to prevent their publication or to delete them promptly. 141 It is not obvious whether this necessarily entails knowledge or whether, in cases where certain criteria are met, an irrebuttable presumption of knowledge would be permissible. As noted earlier, the latter seems to be the French courts’ interpretation, as they almost exclusively referred to objective factors rather than referring to whether Mr Sanchez actually knew of the specific comments. The Court itself claimed to deem ‘whether he had actually known about [the comments] at that time’ as at the ‘heart of the matter’. 142 Nonetheless, the Court addressed this key element by affording deference and leaving it to be ‘decided by the relevant domestic courts on a case-by-case basis’. 143 How can such a key question be left entirely to the national court whose decision is being reviewed?
There are numerous ways to determine knowledge, one of them being notification. 144 According to the decision, Mr Sanchez learned about both comments after they were already deleted by their respective authors. S.B.'s comment was even deleted within 24 hours. What factors were then relevant for the attribution of the other people's expressions to Mr Sanchez? The Court's judgment can be read as emphasizing the ‘objective factors’ considered by the French courts. 145 But attributing knowledge of the comments to Sanchez because of his warning message or skills with social media are equally problematic. This is where the Court raises an even more problematic justification. It states that the two comments presented a dialogue ‘forming a coherent whole’, and Mr Sanchez was ultimately convicted for ‘failing to proceed with the prompt deletion of all the unlawful comments [emphasis added]’. 146
The Court states that S.B.'s and L.R.'s comments were deemed to be ‘responding to and complementing each other (…) as shown in particular by the systematic references to [the political opponent]’ rather than forming a mere discussion thread. 147 On this view, Mr Sanchez's public wall and his initial post serve as a stepping stone to a wider dissemination of hate speech; thus, he is deemed responsible for allowing all posts to publicly subsist on his Facebook account and for any interactions therein. Consequently, Mr Sanchez was responsible for S.B.'s comment notwithstanding the fact that S.B. himself deleted his post less than 24 hours after publication because the comments are not deemed to subsist individually as a ‘system of interactive monologues’, and therefore, deletion ‘does not suffice to negate the applicant's liability’. 148 In other words, despite deletion in a reasonably short timeframe by authors, applicants like Mr Sanchez could still remain liable for problematic comments if the Court finds them to form part of a ‘coherent whole’. 149
It is suggested that this interpretation of the facts falls amounts to ‘intellectual acrobatics’ as noted by Judge Ravarani in his dissent, because the Court did not demonstrate how exactly the comments at hand were forming one coherent conversation. 150 Apart from the general statement that L.R.'s comments ‘echoed’ S.B.'s remarks, the Court did not demonstrate this with reference to the text of the comments. 151 As a result, the line between a mere ‘system of interactive monologues’ and an ‘ongoing dialogue representing a coherent whole’ is never clearly established. 152
Attributing S.B.'s comment to Mr Sanchez despite its deletion sets a dangerous precedent. It would mean that in a situation where a user deletes 99% of all hateful comments in due time but misses 1%, he or she could be still held liable for 100% of the comments because they form a ‘coherent whole’. In this sense, the Court adopted a new, potentially far-reaching theory of liability for several comments that are related content-wise. There is no doubt that this principle is capable of having chilling effects on freedom of expression and posing challenges for public debate.
Reason 4: Who is a politician anyway?
The ruling is likely to create many practical problems for politicians.
The need to balance the ‘respect for the equal dignity of all human beings’ with the right to freedom of expression under Article 10 in combating hate speech is not novel. 153 Its engagement with the issue in the context of social media platforms is particularly vital because the risk of harm occurring online is ‘certainly higher than that posed by the press’ due to the speed of online dissemination as well as its permanence. 154 This is particularly pertinent in an electoral context, where, as noted by the Court, the risk and impact of hate speech dissemination is further heightened. 155
However, the main problem with the Court's ruling concerns its discussion of politicians’ duties and responsibilities. 156 According to the Court, these responsibilities include avoiding comments that might foster intolerance, 157 defending democracy with the ultimate aim being to govern, 158 and exercising particular caution in discussing matters that foster the exclusion of foreigners. 159 However, the Court did not explain how these duties directly link and translate to political actors needing to promptly delete expressions made by unrelated other people on their public accounts. The Court merely acknowledged that ‘in general’ politicians do have duties and responsibilities, 160 and that ‘owing to [their] particular status and position in society, politicians can be expected to be ‘all the more vigilant’. 161 In the absence of knowledge of specific expressions or an intention to incite such expressions, it is unclear how politicians can be attributed what other people say or do.
The Court also stated that Mr Sanchez has a special obligation incumbent on him, 162 based on his degree of notoriety and representativeness. 163 While the Court notes that notoriety and representativeness ‘[lends] a certain resonance and authority to [politicians’] words, deeds or omissions’, 164 the degree remains opaque. What degree of notoriety and representativeness is sufficient for politicians to have special duties incumbent upon them? Is it even a matter of degree, or is it a blanket rule that all politicians are subjected to? Does being a popular politician on its own justify the duty to not only watch one's own language but also that of anyone else? What are the limits of such duties?
These are all key questions that the decision raises. The Court failed to discuss the resources required to meet such responsibilities, and incentives for abuse of such duties. Although the Court admitted that effective comment monitoring requires ‘significant, if not considerable, resources’, the Court felt it could not exempt producers like Sanchez from liability, for fear of ‘[facilitating] or [encouraging] abuse and misuse’. 165
Regrettably, the Court's analysis of a politician's duties and responsibilities not only risks confusion for future cases but also undermines the application of the Delfi test. The judgment can be read as holding politicians to a particularly strict standard of diligence, without defining where the line of special obligations begins and where it ends, therefore rendering it exceedingly difficult for such actors to arrange their affairs. It also makes it difficult for politicians to decide how to operate their social media accounts, how to respond to trolling, and generally the deletion of any comments.
Reason 5: Who else has a special responsibility?
Another problem is that it remains unclear to whom these heightened responsibilities apply. Many non-politicians participate in the public discourse that might be of a political nature. Public figures come in different shapes and sizes. They all wield some power over public discourse. These actors include celebrities, activists, academics and influencers, but sometimes also random people whose expressions strike a chord with the public sentiment of the hour.
The Court emphasized that a ‘degree of notoriety and representativeness [emphasis added]’ of a person would be important to conduct a proportionality analysis on the degree of liability that may be imposed on a person.
166
It subsequently carved out different categories of persons, namely that: (…) A private individual of limited notoriety and representativeness will have fewer duties than a local politician and a candidate standing for election to local office, who in turn will have a lesser burden than a national figure for whom the requirements will necessarily be even heavier, on account of the weight and scope accorded to his or her words and the resources to which he or she will enjoy greater access in order to intervene efficiently on social media platforms [emphasis added].
167
Reason 6: Is closing public debate a solution to hate speech?
The Court recognizes the importance of free speech as a foundational block for a democratic society. 169 For this reason, the States are afforded a ‘particularly narrow’ margin of appreciation for interference under Article 10(2) with political debate. 170 The comments in Sanchez related to ‘present circumstances where the political and social climate [in France] was troubled’. 171 Concurrently, however, the Court recognizes that the nature of the Internet results in wider and more instantaneous dissemination of unlawful speech which brings along a higher likelihood of permanence, and that in an electoral context, ‘the impact of racist and xenophobic discourse becomes greater and more harmful’. 172
It is not unreasonable for the Court to attempt to curb hate speech in such a context. In previous cases dealing with similar types of comments, the Court showed a tendency to accept interferences with Article 10 as necessary when the relevant ‘statements were made against a tense political or social background’. 173 It considered challenges associated with the integration of Muslim immigrants in France as such a situation. 174 Yet in other cases the Court equally acknowledged that in a pre-electoral context, ‘it is particularly important in the period preceding an election that opinions and information of all kinds are permitted to circulate freely’. 175 Moreover, the Court acknowledged the special importance of Article 10 for elected politicians. 176 This shows that the Court is no stranger to situations which require a certain sensitivity towards the societal and political context of the relevant acts of speech. In Sanchez, the Court failed to tread a very thin line. It failed to balance the need for protection against expressions of hate speech by political actors with the need to afford political speech with ‘an elevated level of protection’. 177 Mr Sanchez did not express hateful sentiments in the incriminating posts. They were expressed by others, possibly his supporters, who responded to his lawful speech.
The Sanchez ruling may lead to harmful consequences for democracy and public debate. Faced with the threat of liability for third-party comments, political actors with a social media presence may be inclined to err on the side of caution and engage in the over-removal of content. Such an outcome would be regrettable, given the importance of the Internet and social media platforms as a means of ‘enhancing the public's access to news and in generally facilitating the dissemination of information’ but also as a forum for public debate and engagement with such actors. 178 This decision gives politicians a cover to remove the comments on their profiles that they do not like because the criticism obviously makes them look bad. Thus politicians are even more likely to curate the political narrative found on their profiles. Such behaviour inevitably excludes certain voters from the public forum, thereby also encouraging users to self-censor their speech before publication. This European approach contradicts the latest US case law that closely scrutinizes social media activity of officials to assess whether they speak on the state's behalf when blocking expressions of others. 179
Even worse, the liability might attract bad actors who will target their opponents with anonymous unlawful speech to increase the costs of operating their social media profiles to connect with potential voters. Sanchez offers a strategy on how to prevent opponents from using the megaphone of social media. It is enough to coordinate campaigns that bombard particular actors with many comments, some of which are unlawful. Such actions flood the space with nonsense, raise the operating costs to connect with potential voters, and ultimately might force political actors to entirely shut down the debate – an outcome that is clearly not favourable to freedom of expression.
As a result, public figures with sizable social media platforms would be even more exposed to the potential of liability, as a result of the excessive traffic on their accounts and the practical inability to monitor all comments. 180 But, because Mr Sanchez himself had about only 15 comments in response to his post, 181 the Court quickly dismissed ‘the question of the difficulties caused by the potentially excessive traffic on a politician's account and the resources required to ensure its effective monitoring’. 182 Yet the risk of liability remains high, and when faced with such a threat, politicians with large followings may turn to restricting access to their posts, thereby limiting public debate and the exchange of information and opinions. If, on the other hand, future cases potentially assign less responsibility to politicians with large followings to address the problems of scale, paradoxically, they would start favouring political incumbents, or treat more harshly local politicians than national politicians. Thus, whatever the judges do to apply Sanchez, they are faced with bad choices.
Another ‘major factual element’ the Court focused on is Mr Sanchez's public Facebook wall, 183 which it construed as ‘directly linked to the deliberate choice of the applicant’. 184 By actively choosing to open up the visibility of his Facebook wall, according to the Court, Mr Sanchez ‘must [have been] aware of the greater risk of excessive and immoderate remarks that might appear’ and the risk of dissemination to a wider audience. 185 In doing so, he had opened himself up to the possibility of liability. Access to information and public participation would be practically impossible if politicians’ social media platforms were limited to certain people. It is not clear that the Court considered these trade-offs.
The Court's ruling could have severe repercussions of closing off public viewership and engagement since the possibility of being held liable would be a considerable incentive for politicians, and perhaps other relevant public figures, to discontinue offering openly accessible social media profiles.
Reason 7: Does Sanchez contradict EU law?
If the Court were to suggest that the Convention requires that liability be imposed for failing to anticipate third-party comments by those who store them, it would be contradicting the expectations under EU law.
In Delfi, the Court only said that in some areas of sensitive illegal content, such as hate speech, States can experiment with different liability models. However, this leeway does not exist for 40 members of the Council of Europe that are in the orbit of the EU law imposing a set of liability exemptions via the ECD and/or the DSA. 186
Did the Court suggest that they must do so in Sanchez? As indicated earlier, the judgment is vague enough to allow scope for many interpretations, including that the Court intended to suggest exactly that.
With various parts of the Grand Chamber judgment tilting towards a more deferential or prescriptive approach, it is difficult to conclude whether Sanchez on the whole points in either direction. On the whole, the Court recognized that ‘national authorities are better placed than itself to understand and appreciate the specific societal problems faced in particular communities and contexts, or the likely impact of certain acts that they are called upon to adjudicate’. 187
The resulting confusion is best illustrated by a recent judgment, Zöchling, 188 decided by a Committee of three judges that interpreted Sanchez as requiring at least an assessment of whether pre-knowledge duties of care to avoid liability would be proportionate. The three judges, two of whom were in the Sanchez Grand Chamber majority, indeed appear to have utilized the more ‘prescriptive’ Sanchez dicta. 189
The case involved a small Internet news portal that published around six to ten articles per day, and was subject to a civil lawsuit concerning comments under these articles. 190 The company behind the news site deleted the comments within a few hours after receipt of the request and disclosed the users’ data upon request. By the time they were removed, the comments had been visible on the portal for 12 days. The users in question were blocked. The Austrian Court of Appeals faithfully applied the EU liability exemptions and CJEU case law. However, it did not apply Delfi criteria, as the law was clear. According to the Court of Appeals, the news site did not lose liability exemption until it obtained notification, hence it has no pre-notification knowledge of specific unlawful comments. As a result, the company behind the website could not bear any liability. While the Regional Court considered the application of the Austrian media law and its duties of care, the Court of Appeal rightly considered this irrelevant. The Austrian media law cannot impose additional liability on news sites acting as hosts of third-party comments because this would have been against EU law. It is irrelevant whether the article which attracted the comments ‘intentionally stirred up antipathies’ or not, unless it can be argued that the article unlawfully incited such comments. 191
The Court did not carry out an exercise to examine whether the platform could be a ‘large professionally managed Internet news portal’ or ‘other fora on the Internet where third-party comments can be disseminated’. 192 Instead, it relied upon Sanchez as a seemingly broader test. It held that the rejection of the Austrian courts to consider whether it is proportionate to impose liability for the comments of others for the period before the newspaper hosting them acquired knowledge about specific unlawful comments violated Article 8. 193 Thus, while adopting a stance of respect and deference to national factual findings, and perhaps not fully cognizant of the EU law framework, the Court prompted Austrian courts to scrutinize the balance and proportionality of the existing EU law. This is in stark contrast with what Delfi's deference to legislatures stood for. Moreover, Zöchling gives the impression that courts must seriously engage with the question of whether any platforms can be reasonably expected to employ humans and technology to prevent publication or to ensure immediate removal of unlawful comments of others. In other words, this suggests an obligation to engage in general monitoring that is explicitly prohibited by Article 8 DSA (and formerly Article 15 of the EU E-Commerce Directive). 194
To be clear, one should not draw such big conclusions from a single Committee judgment that was meant to apply ECtHR case law in a clear and settled area of law. However, inevitably, this is what some of the courts in the Member States might do when trying to interpret Sanchez. Hence the title of our comment. Expect Grand Confusion.
Thus, while Sanchez might be still compatible with EU law because of its ‘may’ reading, some ECtHR judges themselves are already encouraging reading it as a ‘must’. Zöchling is asking courts to consider options that are clearly against EU law, as interpreted by the CJEU in YouTube. 195 The Luxembourg Court requires awareness of the specific unlawful speech of others to trigger knowledge, whether actual or constructive, to strip content hosts from the liability exemption. 196 General awareness is never sufficient to lose the hosting liability exemption. 197 Monitoring and fact-finding duties cannot be imposed unless they concern specific cases. 198 Thus, as a matter of EU law, the Austrian court cannot question the reasonableness of pre-knowledge efforts. But that is exactly what the ECtHR asking it to do.
Conclusion
The Strasbourg case law on liability for the speech of others online officially descended into chaos without a proper sense of direction. Intermediaries as digital facilitators were supposed to be, according to the Court's own words, subject to a ‘differentiated and graduated approach’ that treats them differently from editorial media. 199 As of late, there is no differentiation of actors and little sense of expectations being graduated by anything predictable. 200 Any argument about the duties and responsibilities that are ‘differentiated and graduated’ has become purely rhetorical.
The only reason why not-so-clear case law led to less visible consequences in the Council of Europe's 46 Member States in the last decade is that national law has been harmonized since 2000 through liability exemptions in 40 of them. However, confusion about Sanchez now has the potential to threaten legal certainty, as illustrated by its application in Zöchling.
Despite the Court's proclaimed deference to national law and increased use of the subsidiarity principle, it is striking that democratically adopted legislation about digital services has been ignored for so long in Strasbourg. It is the law common to the absolute majority of states that are members of the Council of Europe and has been on the books for two decades.
While laws like the EU Digital Services Act offer slower, complex and less radical solutions, they are a product of democratic deliberations. They offer a truly graduated and differentiated approach by distinguishing liability for wrongs of others from accountability for operating the systems, and dosing obligations to operate such systems according to the size, function, and user base of services. As one of us has argued elsewhere, there are many principles underlying the DSA that the Court could borrow from. 201
Footnotes
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
