JOINT DISSENTING OPINION OF JUDGES SAJÓ AND TSOTSORIA

To explain our dissent, we will offer a detailed and traditional analysis of the case, as is common in the Court ’ s practice. There are, however, some broader issues which are more important than our dissatisfaction with this judgment ’ s troubling departure from the prevailing understanding of the case-law. These fundamental concerns will be spelled out first.

I.

Collateral censorship

1 . In the present judgment the Court has approved a liability system that imposes a requirement of constructive knowledge on active Internet intermediaries [8] (that is, hosts who provide their own content and open their intermediary services for third parties to comment on that content). We find the potential consequences of this standard troubling. The consequences are easy to foresee. For the sake of preventing defamation of all kinds, and perhaps all “ illegal ” activities, all comments will have to be monitored from the moment they are posted. As a consequence, active intermediaries and blog operators will have considerable incentives to discontinue offering a comments feature, and the fear of liability may lead to additional self-censorship by operators. This is an invitation to self-censorship at its worst.

2 . We regret that the Court did not rely on the prophetic warnings of Professor Jack Balkin [9] . As Professor Balkin has demonstrated, the technological infrastructure behind digital communication is subject to less visible forms of control by private and public regulators, and the Court has just added another such form to this panoply. Governments may not always be directly censoring expression, but by putting pressure and imposing liability on those who control the technological infrastructure ( Internet service provider s , etc.) , they create an environment in which collateral or private-party censorship is the inevitable result. Collateral censorship “ occurs when the state holds one private party A liable for the speech of another private party B, and A has the power to block, censor, or otherwise control access to B ’ s speech ” [10] . Because A is liable for someone else ’ s speech, A has strong incentives to over-censor, to limit access, and to deny B ’ s ability to communicate using the platform that A controls. In effect, the fear of liability causes A to impose prior restraints on B ’ s speech and to stifle even protected speech. “ What looks like a problem from the standpoint of free expression ... may look like an opportunity from the standpoint of governments that cannot easily locate anonymous speakers and want to ensure that harmful or illegal speech does not propagate. [11] ” These technological tools for reviewing content before it is communicated online lead (among other things) to: deliberate overbreadth; limited procedural protections (the action is taken outside the context of a trial); and shifting of the burden of error costs (the entity in charge of filtering will err on the side of protecting its own liability, rather than protecting freedom of expression).

3 . The imposition of liability on intermediaries was a major obstacle to freedom of expression for centuries. It was the printer Harding and his wife who were arrested for the printing of the Drapier ’ s Letters and not the anonymous author (Jonathan Swift), who continued to preach undisturbed. It was for this reason that exempting intermediaries from liability became a crucial issue in the making of the first lasting document of European constitutionalism, the Belgian Constitution of 1831 [12] . This is the proud human rights tradition of Europe that we are called upon to sustain.

The general context

4 . It is argued in the present judgment that the Court is called upon to decide the case at hand, but this is only part of our duty and such an argument is dangerous in its one-sidedness. As the Court summarised matters in Rantsev v. Cyprus and Russia ( no. 25965/04, § 197, ECHR 2010 ) [13] :

“ [The Court ’ s] judgments serve not only to decide those cases brought before it but, more generally, to elucidate, safeguard and develop the rules instituted by the Convention, thereby contributing to the observance by the States of the engagements undertaken by them as Contracting Parties (see Ireland v. the United Kingdom , 18 January 1978, § 154, Series A no. 25; Guzzardi v. Italy , 6 November 1980, § 86, Series A no. 39; and Karner v. Austria , no. 40016/98, § 26, ECHR 2003-IX). Although the primary purpose of the Convention system is to provide individual relief, its mission is also to determine issues on public-policy grounds in the common interest, thereby raising the general standards of protection of human rights and extending human rights jurisprudence throughout the community of the Convention States (see Karner , cited above, § 26 , and Capital Bank AD v. Bulgaria , no. 49429/99, §§ 78 to 79, ECHR 2005- XII ). ”

5 . Further, as the Court stated in Animal Defenders International v. the United Kingdom ([GC] no. 48876/08, § 108, ECHR 2013 ):

“ It emerges from that case-law that, in order to determine the proportionality of a general measure, the Court must primarily assess the legislative choices underlying it ( see James and Others [ v. the United Kingdom , 21 February 1986] , § 36 [, Series A no. 98] ). ”

6 . The present judgment expressly deals with the general context (see the “ Preliminary remarks ” section starting at paragraph 110 ) but without determining “ issues on public-policy grounds ” . The Internet is described as an “ unprecedented platform ” and while there is reference to benefits, it is described as posing “ certain dangers ” , the advantages being scarcely mentioned. We disagree. The Internet is more than a uniquely dangerous novelty. It is a sphere of robust public discourse with novel opportunities for enhanced democracy. Comments are a crucial part of this new enhanced exchange of ideas among citizens. This has been the Court ’ s understanding so far in its case-law (see Ashby Donald and Others v. France , no. 36769/08, § 34, 10 January 2013, and also Węgrzynowski and Smolczewski v. Poland , no. 33846/07, § 58, 16 July 2013) [14] .

7 . It is noteworthy in this context that the thirteen lines of comparative-law analysis in the present judgment do not refer to specific national practices. While there are novel restrictions on posting on the Internet in the recent legislation of a couple of European countries, the Estonian approach is rather unique. In the overwhelming majority of the member States of the Council of Europe, and also in genuine democracies all over the world, the regulatory system (in conformity with the expectations of the rule of law) is based on th e concept of actual knowledge. A safe harbour is provided by the rule of notice and action (primarily “ notice - and - take - down ” ). This Court has not been known for developing rights restrictions which go against the prevailing standards of the member States, except in a few cases where a narrow majority found that deeply held moral traditions justified such an exception .

Consequences

8 . The Court has endorsed the standard of the Estonian Supreme Court, namely that active intermediaries must remove comments “ without delay ” after publication (see paragraph 153 of the present judgment), and not upon notice or on other grounds linked to actual knowledge. Active intermediaries are therefore invited to exercise prior restraint. Moreover, member States will be forced to introduce a similar approach because otherwise, according to the logic of the present judgment, there is no proper protection for the rights of those who feel defamed by comments. To avoid trouble, for active intermediaries the safe harbour will simply be to disable comments [15] .

9 . The Court is aware of the unhappy consequences of adopting a standard that can be satisfied only by constant monitoring of all comments (and implicitly, all user-generated content). For the Court, “ the case does not concern other fora on the Internet ... or a social media platform ... where the content provider may be a private person running the website or blog as a hobby ” (see paragraph 116 of the present judgment). It is hard to imagine how this “ damage control ” will help. Freedom of expression cannot be a matter of a hobby.

II.

Delfi ’ s role as active intermediary

10 . Turning to the specific case, we find that the Estonian Supreme Court did not provide relevant and sufficient reasons for the very intense interference with the applicant company ’ s rights and did not apply an appropriate balancing exercise. This amounts to a violation of the Convention.

11 . This case is about interference with the freedom of expression of Delfi as an active intermediary . Delfi published an article on the destruction of ice roads by a public service ferry company on its news portal and it enabled comments on the article. It is undisputed that there was nothing illegal in the article. It was accepted by the national courts, and we could not agree more, that Delfi was engaged in journalistic activities and that the opening up of a comment s section formed part of the news portal. However, the news portal was not the author of the unedited comments. Furthermore, at least in the Chamber ’ s view (see paragraph 86 of its judgment), the debate concerned a matter of “ a certain degree ” of public interest. We believe that the article dealt with a matter of public interest and that the comments, even the impugned ones, were part of the debate even though they may have been excessive or impermissible. Delfi was held liable under the Civil Code of Estonia for defamation originating from comments posted in the comment s section of the article. This concerned twenty comments .

The nature of the comments

12 . Throughout the whole judgment the description or characterisation of the comments varies and remains non-specific. The Supreme Court of Estonia has its own interpretation: it refers to “ insult in order to degrade ” and “ degrade human dignity and ridicule a person ” and finds Delfi liable for disrespecting the honour and good name of the person concerned. According to paragraph 117 of the present judgment, “ the impugned comments ... mainly constituted hate speech and speech that directly advocated acts of violence ” [16] (see also paragraph 140). However, according to paragraph 130 ( “ the legitimate aim of protecting the reputation and rights of others ” ), the offence in issue concerned the reputation and unspecified rights of others. It is not clear to which comments the Court is referring. Does the comment “ a good man lives a long time, a shitty man a day or two ” (comment no. 9 – see paragraph 18 of the judgment) amount to advocating violence [17] ?

13 . It is unfortunate that the characterisation of the comments remains murky. What is really troubling here is never spelled out: that some of the comments are racist. Comment no. 2 is a recital of anti-Semitic stereotypes ending with a reference to the annihilation by fire of the addressee as a Jew.

14 . We are not going to discuss here to what extent some of the references satisfy the strict requirements of incitement to violence, given the nature of the Internet. Does a call for violence or a wish to see someone killed have the same effect on the Internet as a similar statement made in a face-to-face encounter in a situation like the present one? This is not a call to arms by an extremist group. The answer has to be established by means of a proper judicial process. No criminal action against the commenters was taken, notwithstanding the reference to lynching [18] . The question of the extent to which such comments amount to a real threat would have deserved proper analysis. However, the judgment simply accepts the findings of the Supreme Court, which says only that the illegality of the comments is manifest (and then, like the judgment, characterises them in different ways).

15 . We will also refrain from an analysis of the impact of the hateful messages regarding their capacity to incite imminent violence or even to build up lasting hatred resulting in harassment or real threats against L. Racism and constraining others to live in an environment full of hatred and real threats cannot find refuge in freedom of expression. This legitimate concern must not, however, blind those who are called upon to take action, and they must be reminded that “ hate-speech regulations put actual feelings, often honourable ones, ahead of abstract rights – which seems like common sense. It takes an active effort to resist the impulse to silence the jerks who have wounded you ” [19] .

Interference and the right of active intermediaries

16 . There is general agreement that the Estonian Supreme Court ’ s judgment interfered with Delfi ’ s freedom of expression, though the nature of the right remains somewhat non-specific. In our view the rights concerned are the rights of the press. User comments may enrich the article. The rights of an active intermediary include the right to enable others to impart and receive information.

Lawfulness of the interference: the problem of foreseeability

17 . According to the prevailing methodology of the Court, the next question to be asked concerns the lawfulness of the measure. This entails a review of the foreseeability of the law. It is accepted by the Court that the applicable law was the Civil Code and not the Information Society Services Act. The Information Society Services Act apparently exempts service providers and provides a “ safe harbour ” in the sense that , once the service provider becomes aware of the illegal content and removes it expeditiously, it cannot be held liable. Neither the domestic authorities nor the Court explain why the provision of binding European law that is part of the national legal system is immaterial, except to say that the present case concerns a matter of publication rather than data storage. It is, of course, not for this Court to interpret European Union law as such. This does not mean that we should not regard it as part of the domestic system, attributing to it its proper constitutional weight. Be that as it may, section 10 (liability for storage) of the Information Society Services Act provides a “ safe harbour ” rule for service providers in the case of storage. In these circumstances, a reasonable justification should be required for the choice of the higher level of liability under the Civil Code. The (highly problematic) choice of a publisher ’ s liability does not address the issue of the supremacy of European Union law or the problem of lex specialis . It is possible that where the information storage provider generates content, the Information Society Services Act is inapplicable, but this must be demonstrated and must also be foreseeable. Moreover, the service provider in the present case did not generate the impugned content: that content was user-generated. To argue that the commercial nature of the data storage brings the activity within the liability regime applicable to publishers is not convincing. Storage is considered to be a commercial activity but that did not change the equation for the Information Society Services Act, which allowed a “ safe harbour ” regime.

18 . One of the requirements flowing from the expression “ prescribed by law ” is foreseeability. Thus, a norm cannot be regarded as a “ law ” within the meaning of Article 10 § 2 unless it is formulated with sufficient precision to enable the citizen to regulate his conduct; he must be able – if need be with appropriate advice – to foresee, to a degree that is reasonable in the circumstances, the consequences which a given action may entail. Those consequences need not be foreseeable with absolute certainty (see Lindon, Otchakovsky-Laurens and July v. France [GC], nos. 21279/02 and 36448/02, § 41, ECHR 2007 ‑ IV) [20] . A legal adviser could not have informed Delfi with sufficient certainty that the Directive on certain legal aspects of information society services did not apply. The applicable law was not obvious, to the extent that even in 2013 a court in Cyprus found it necessary to ask the Court of Justice of the European Union for a preliminary ruling in a related matter, namely the liability of news portal publishers (see Case C ‑ 291/13, Papasavvas , CJEU). If there was uncertainty in 2013 in the European Union on a similar but less complicated matter, which was clarified in 2014, how could learned counsel have been sufficiently certain in 2006?

19 . More importantly, it was not foreseeable that the applicant company ’ s liability under the Civil Code would be that of a publisher. The Supreme Court judgment itself refers to another Supreme Court judgment of 21 December 2005. That judgment, perhaps already available to Delfi on 24 January 2006 (the date of the article), was summarised by the Supreme Court as follows:

“ [F]or the purposes of section 1047 of the Obligations Act disclosure [ avaldamine ] means communication of information to third parties and the discloser is a person who communicates the information to third parties. ... in the case of publication [ avaldamine ] of information in the media, the discloser/publisher [ avaldaja ] can be a media company as well as the person who transmitted the information to the media publication. ”

The Supreme Court applied this consideration in the following way:

“ Publishing of news and comments on an Internet portal is also a journalistic activity. At the same time, because of the nature of Internet media, it cannot reasonably be required of a portal operator to edit comments before publishing them in the same manner as applies for a printed media publication. While the publisher is, through editing, the initiator of the publication of a comment, on the Internet portal the initiator of publication is the author of the comment, who makes it accessible to the general public through the portal. Therefore, the portal operator is not the person to whom information is disclosed. Because of [their] economic interest in the publication of comments, both a publisher [ väljaandja ] of printed media and an Internet portal operator are publishers/disclosers [ avaldajad ] as entrepreneurs. ”

20 . This (too) raises serious concerns as to the foreseeability of the Civil Code as applied in the present case. The Supreme Court clearly states that “ it cannot reasonably be required of a portal operator to edit comments before publishing them in the same manner as applies for a printed media publication ” . The Internet portal operator is called a “ publisher/discloser ” in the English translation. The term used in the original Estonian does not look the same as that used for a publisher ( “ väljaandja ” ) but appears to be a different one ( “ avaldajad ” ). The applicant company argued that other “ disclosers ” or disseminators (libraries, bookstores) were not held to be publishers under existing tort law. Why would one assume that an Internet operator falls under the duty of care applicable to “ väljaandja ” instead of “ avaldajad ” ? There is a contradiction here that hampers foreseeability. As the Chamber rightly acknowledged (see paragraph 75 of its judgment), the text of the relevant provisions of the Constitution, the Civil Code (General Principles) Act and the Law of Obligations Act was “ quite general and lack[ed] detail ” . The provisions of the Law of Obligations Act are all directed at a person or entity that defames – the tortfeasor , in this instance, being the author of the comments in question on the applicant company ’ s website – and do not directly address the novel situation of an intermediary providing a platform for such expressive activity while not being the aut hor or a traditional publisher. Only divine legal counsel could have been sufficiently certain that a portal operator would be liable for a comment it was not aware of, under a kind of strict liability that applied to publishers (editors) who operated in full knowledge of the whole publication. It is noteworthy that the three competent levels of jurisdiction applied three different theories of liability. Vaguely worded, ambiguous and therefore unforeseeable laws have a chilling effect on freedom of expression. A troubling uncertainty persists here [21] .

21 . The Court has previously held that “ the policies governing reproduction of material from the printed media and the Internet may differ. The latter undeniably have to be adjusted according to the techno ­ logy ’ s specific features in order to secure the protection and promotion of the rights and freedoms concerned ” (see Editorial Board of Pravoye Delo and Shtekel , cited above, § 63). This point of principle provides an important benchmark when examining whether the appli ­ cation of domestic law in the present case was reasonably foreseeable for the applicant company, as regards user-generated content on its website. In Recommendation CM/ Rec( 2011)7 on a new notion of media , the Committee of Ministers noted that “ [t]he roles of each actor can easily change or evolve fluidly and seamlessly ” and called for a “ differentiated and graduated approach ” .

Necessary in a democratic society

22 . The next question to answer is to what extent the measure aimed at the avoidance of hate speech [22] (which was the most likely justification for the interference, at least in the Court ’ s view, but not that of the domestic authorities – see paragraph 140 of the present judgment) was necessary in a democratic society [23] . The reference to the conflict between Article 8 and Article 10 rights in paragraph 139 points to the applicability of a balancing exercise with a wider margin of appreciation.

23 . The Court first states, and we agree, that some of the impugned statements are not protected by the Convention. That does not in itself solve the problem, as one cannot, in the circumstances of the case, equate the expressions used by the commenters with the activities of an active intermediary.

Shift to a “ relevant and sufficient reasons ” analysis

24 . The Court considers in paragraph 142 of the present judgment that within the proportionality analysis its task is to examine “ [i]n the light of the Supreme Court ’ s reasoning ... whether the domestic courts ’ finding of liability on the part of the applicant company was based on relevant and sufficient grounds in the particular circumstances of the case (see paragraph 131 above) ” . No reference is made here to the established principle that the Court, in the exercise of its supervisory role, is not satisfied if the respondent State exercised its discretion only reasonably, carefully and in good faith. Sufficient reasons are more than simply reasonable.

25 . More importantly, the “ relevant and sufficient grounds ” test is only part of the proportionality analysis [24] . Once the Court has found that the reasons given are relevant and sufficient, the proportionality analysis begins rather than ends. The “ relevant and sufficient grounds ” test is a threshold question to determine whether and how the margin of appreciation is to be applied; it is relevant in the determination of the existence of a pressing social need (see all the authorities cited in paragraph 131 of the present judgment). Why is there a need to determine that the grounds relied upon by the domestic authorities were relevant and sufficient (which is more than simply reasonable – see above)? Because, as the Court has always said, and as it also reiterates in this case (see paragraph 131), “ the Court has to satisfy itself that the national authorities applied standards which were in conformity with the principles embodied in Article 10 and, moreover, that they relied on an acceptable assessment of the relevant facts ” (The Court did not go into an evaluation of the national authorities ’ assessment of the relevant facts, although this consideration might have been relevant.)

26 . The Court has concluded that the Estonian Supreme Court did provide relevant and sufficient reasons for the level of liability it applied. It reached this conclusion after consider ing the following relevant points: the context of the comments, the measures applied by the applicant company in order to prevent or remove defamatory comments, the liability of the actual authors of the comments as an alternative to the applicant company ’ s liability, and the consequences of the domestic proceedings for the applicant company. These may be relevant, but there may be other relevant considerations as well. We will address only the sufficiency of some of these elements.

Publishers ’ liability extended: base economic interest

27 . The Estonian Supreme Court ’ s judgment is based on the assumption that an active intermediary is a publisher. The case-law of the Court has so far pointed in the opposite direction [25] . The international - law documents cited by the Court emphasise the importance of differentiation, given the specific nature of Internet technology. It has already been mentioned that such differentiation had been recognised a few months previously by the Estonian Supreme Court. However, in the present case, the Estonian Supreme Court equated publishers with active intermediaries: “ [ b ] ecause of [their] economic interest in the publication of comments, both a publisher of printed media and an Internet portal operator are publishers/disclosers as entrepreneurs ” ( see paragraph 112 of the present judgment). The Court sees no reason to call into question the above approach, although it notes there has been “ a certain development in favour of distinguishing between the legal principles regulating the activities of the traditional print and audio - visual media , on the one hand , and Internet-based media operations , on the other. ... Therefore, the Court considers that because of the particular nature of the Internet, the ‘ duties and responsibilities ’ that are to be conferred on an Internet news portal for the purposes of Article 10 may differ to some degree from those of a traditional publisher as regards third-party content ” (see paragraph 113). We could not agree more, but for us it is impossible to see how a recognition of difference may result in eliding publishers and active intermediaries on the sole basis of their commercial nature. The Court seems to accept as relevant and sufficient the position of the Estonian Supreme Court. In this approach, economic interest is sufficient for the identification of the active intermediary with a publisher, although the two were considered to be different just a sentence earlier. No explanation is offered as to how this is compatible with the point of reference of the Court, namely the Committee of Ministers ’ Recommendation CM/Rec(2011)7 (quoted in paragraph 46 of the present judgment), which calls for a “ graduated approach ” to apply to the intermediary. The additional reasons referred to in paragraphs 115 to 117 concern the nature of the expression and the size of the intermediary, which are neither relevant nor sufficiently connected to the liability of a traditional publisher.

28 . To find that responsibility of the press (or of any speaker, for that matter) is enhanced by the presence of an economic interest does not sit comfortably with the case-law. It is true that the margin of appreciation is broader in the commercial sphere (see Mouvement raëlien suisse v. Switzerland [GC], no . 16354/06, § 61, ECHR 2012). “ It is however necessary to reduce the extent of the margin of appreciation when what is at stake is not a given individual ’ s purely ‘ commercial ’ statements, but his participation in a debate affecting the general interest ” (see Hertel v. Switzerland , 25 August 1998, § 47, Reports of Judgments and Decisions 1998-VI). The fact that the original article and the comment s section (offered to the general public for free!) is part of the economic activity of a news portal operator does not change the equation. The article and the related, dedicated comment s section are protected because they facilitate and take part in a debate on a matter of public interest.

29 . Over the last three hundred years, ideas have been generated for money and this has never been held to reduce the level of protection otherwise granted to speech. We do not live in the aristocratic world of the Roman auctor who could afford not to care about the financial return on ideas (though very often being dependent on imperial pleasure). It cannot be held against a newspaper or publisher that they operate an outlet as a commercial enterprise. One cannot expect the production of ideas for free. There will be no generation of ideas without adequate financial means; the material reward and the commercial nature of the press enterprise are not (and cannot be) grounds for diminishing the level of protection afforded to the press. Information is costly; its efficient communication is not a mere hobby. The same platform that has been understood as commercial, and thus subject to increased liability, is also a platform for enhanced, interactive discourse on a matter of public interest. This aspect has not been taken into consideration in the balancing exercise.

30 . However, the Court does provide at least one relevant consideration for extending the liability of an active intermediary. It is certainly true that the active intermediary can exercise control over the comments that appear on its site and it is also true that by creating a comments section , and inviting users to participate, it engages in an expressive activity that entails responsibility. But the nature of the control does not imply identification with a traditional publisher.

31 . There are additional differences between a publisher (understood here as a newspaper editor, someone who controls content) and an active intermediary:

(a) in a newspaper the journalist is typically an employee (although there are good reasons to protect a journalist against his or her editor/employer); and

(b) in principle, the editor is in a position to know in advance the content of an article to be published and has the decision-making power and the means to control the publication in advance.

Contrary to the case of a publisher, these elements are only partially present in the case of active intermediaries who host their own content and actively monitor all data (that is to say, are in a position to read it and remove it after the data are made accessible), as in the case of Delfi. The active intermediary has prior control only to the extent that this is made possible by a filtering mechanism. It also has the power to remove a message or block access to it. However, in normal circumstances the active intermediary has no personal control over the person who posts the message. The commenter is not the employee of the publisher and in most cases is not known to the publisher. The publication occurs without the decision of the editor. Hence the level of knowledge and control differs significantly.

32 . Control presupposes knowledge. In this regard the difference between the editor/publisher and the active intermediary is obvious.

The level of responsibility

33 . While Delfi cannot be defined as a publisher, the company does voluntarily provide an opportunity for people to post comments and, even if this activity is a matter of freedom of expression of a journalistic nature, this does not exempt the activity from liability. The Information Society Services Act does envisage such liability, among other things for storage, as is the case here. The Act bases liability on “ actual knowledge ” and entails a duty of expeditious removal. The Court found this to be insufficient.

34 . The Court finds it to be a relevant and sufficient consideration that the Supreme Court limited the responsibility of the applicant company to post-publication liability. However, as quoted in paragraph 153 of the present judgment, the Supreme Court stated that the applicant company “ should have prevented the publication of comments ” . The fact that it “ also held ” that there was a duty of removal after the disclosure does not change the first statement. Both pre- and post-disclosure liability are being advocated here and this cannot be disregarded when it comes to the evaluation of “ sufficient reasons ” [26] . It was under this standard that Delfi was found to be at fault for the disclosure of the information, which could not have been undone by removal upon request.

35 . The duty to remove offensive comments without actual knowledge of their existence and immediately after they are published means that the active intermediary has to provide supervision 24/7. For all practical purposes, this is absolute and strict liability, which is in no sense different from blanket prior restraint. No reasons are given as to why only this level of liability satisfies the protection of the relevant interests.

36 . Are there sufficient reasons for this strict liability [27] , disguised by the fault rules of the Civil Code? The Court reviewed the precautionary measures applied by Delfi and found them inadequate. These were fairly standard measures: a disclaimer as to illegality, a filtering mechanism, the separation of the comment s section from the article, and immediate removal upon notice. It was decisive for the Court that the filtering mechanism failed. There is no review of the adequacy of the filtering mechanism (was it state-of-the-art; can there be a duty to apply state-of-the-art systems; is there any reason for being held liable with a state-of-the-art filtering system?). The Court itself found that filtering must have been a simple task and that the system failed. No expert opinion, no cross-examination. We are simply assured that setting up a dedicated team of moderators is not “ private censorship ” . There is no consideration of the possibility of less intrusive measures; only removal “ without delay ” , that is, upon posting (see paragraph 159 of the present judgment ), satisfies the goal of eliminating hate speech and its progeny [28] . This insatiable appetite for preventive protection results in circular reasoning: a publisher has a similar liability, therefore an active in termediary is like a publisher.

37 . Neither the domestic courts nor the judgment provide sufficient and relevant reasons for a de facto strict liability rule. The Court was satisfied that it could find relevant and sufficient reasons in the Estonian Supreme Court ’ s judgment in view of the extreme nature of the comments, the nature of the commercial operation, the insufficiency of the measures applied by the applicant company, the interest in ensuring a realistic prospect of the authors of such comments being held liable, and the mildness of the sanction. Apparently these are the reasons that compelled the Court to endorse constructive knowledge. The Court found that the absolute duty of immediate take-down on publication (as applied to the applicant company) was proportionate to the aim of protecting individuals against hate speech.

38 . We would claim, in conformity with all the international documents cited, that an active intermediary which provides a comments section cannot have absolute liability – meaning an absolute duty of knowledge or, in practice, construed (constructive) knowledge. The protection of freedom of expression cannot be turned into an exercise in imposing duties. The “ duties and responsibilities ” clause of Article 10 § 2 is not a stand-alone provision: it is inserted there to explain why the exercise of the freedom in question may be subject to restrictions, which must be necessary in a democratic society. It is only part of the balance that is required by Article 10 § 2.

Balancing (lack of)

39 . If one applies a balancing approach, then the other side of the balance must also be considered. According to the case-law, there must be proper consideration of the following factors, among others:

(a) the interference concerns the press and journalism. Delfi pursued journalistic activities, both in providing a news portal and by attaching a comment s section to an article. Journalism is not exempt from liability, but it triggers stricter scrutiny. “ [ T ] he safeguard afforded by Article 10 to journalists in relation to reporting on issues of general interest is subject to the proviso that they are acting in good faith and on an accurate factual basis and provide ‘ reliable and precise ’ information in accordance with the ethics of journalism ” ( see Stoll v. Switzerland [GC], no. 69698/01, § 103, ECHR 2007 ‑ V ) [29] . There is no consideration of good faith in the judgment. Moreover, when it comes to online journalism and the responsibility of an active intermediary, due consideration must be given to the role of self-regulation of the profession;

(b) the Court has held that the “ punishment of a journalist for assisting in the dissemination of statements made by another person in an interview would seriously hamper the contribution of the press to discussion of matters of public interest and should not be envisaged unless there are particularly strong reasons for doing so ” (see Jersild v. Denmark , 23 September 1994, § 35, Series A no. 298 ). The Court has found this to be a relevant principle and we agree that this is of importance for the press, including news portals and active intermediaries. However, this principle is simply not discussed in the judgment;

(c) the opening of a comment s section provides a forum for the expression of views concerning public matters. As such it contributes to more robust speech, and enables others to receive and impart information that does not depend on centralised media decisions. Any restriction imposed on the means necessarily interferes with the right to receive and impart information (see, for example, Öztürk v. Turkey [GC], no. 22479/93, § 49, ECHR 1999 ‑ VI);

(d) the debate was about a matter of public interest. The comments related to the highly controversial behaviour of a large corporation.

40 . The Court is reluctant to consider the possibility of less intrusive means, but in our view at least some justification is needed to explain why only the equivalent of prior restraint and absolute liability satisfies the non-specific duties and responsibilities of active intermediaries.

41 . Without speculating on the outcome of the balancing analysis, we note that these considerations have been left out. Where part of the required considerations were not included in the balancing exercise carried out by the domestic court, the Court must find a violation.

42 . We do not intend to close our eyes to the problem of racist speech. The fact that the comment s section technically facilitated the dissemination of racism should be part of the proportionality analysis. In fact, the comment s section facilitates the dissemination of all views equally. However, we accept, even without specific evidence, that the more comments there are, the higher the likelihood that racist comments are made. We accept this, though only as a hypothesis, as no evidence to this effect was produced in the proceedings, or referred to by the Court.

43 . Even assuming such increased likelihood of racist comments on comment sites (once again, a matter subject to proof), it remains appropriate to consider what the proper level of care is in the face of such risk. Perhaps the filtering mechanism was inadequate to meet this challenge. This was the position taken by the Court, without defining what the appropriate level of care would have been in 2006 in Estonia. We do not know and cannot know. The Court cannot replace the lack of a domestic analysis with its own analysis. Moreover, it is not for the Court to take on the role of national legislat or . We cannot rule out that the need to fight racist speech (a matter of public order and not simply a personality right) might dictate a duty of care that would impose duties beyond the measures applied by Delfi. But the task of the Court is to determine whether the interference by the domestic authorities was actually based on proper and credible grounds. These are absent here; hence there was a violation of the Convention.

APPENDIX

We trust that this is not the beginning (or the reinforcement and speeding up) of another chapter of silencing and that it will not restrict the democracy-enhancing potential of the new media. New technologies often overcome the most astute and stubborn politically or judicially imposed barriers. But history offers discouraging examples of censorial regulation of intermediaries with lasting effects. As a reminder, here we provide a short summary of a censorial attempt that targeted intermediaries.

In Reformation England, the licensing system of the Catholic Church was taken over by the State and it became a State tool for control of all printed publications. Licensing provided the Crown with “ censorship prior to publication and easy conviction of offenders ” [30] . These laws cut seditious material off at the place of mass production – the printer. Initially involving prosecution by the Star Chamber, the licensing scheme punished any printer who failed to receive a licence for the material he intended to print (a licence conditional on royal approval). With the abolition of the Star Chamber there was a brief end to the licensing laws during the English Civil War. Parliament, however, did not like the spreading of radical religious and political ideas. It decided to replace Crown censorship with its own, also in order to protect the vested business interests of the printers ’ guild. The result was the Licensing Order of 14 June 1643, which reintroduced for Parliament ’ s benefit the previously despised order of the Star Chamber Decree (prepublication licensing; registration of all printed material with the names of author, printer and publisher; search, seizure and destruction of any books offensive to the government; and punishment of printers and publishers). Post-revolution, people tend to reinvent the same tools of oppression that the revolutionaries stood up against. (See also the Alien and Sedition Act enacted in the United States.) The Stationers ’ Company was given the responsibility of acting as censor, in return for a monopoly on the printing trade. The licensing system, with the financial interest of the publishers ’ /printers ’ guild, was a more effective censor than seditious libel laws.

This restored licensing system became John Milton ’ s target in Areopagitica: A Speech of Mr John Milton for the Liberty of Unlicens ’ d Printing in November 1644. It was resistance to the self-censorship imposed on intermediaries (the printers) that produced Areopagitica , the first and most important manifesto of freedom of expression. Areopagitica attempted to persuade Parliament that licensing had no place in the free pursuit of truth . It argued that an unlicensed press would lead to a marketplace of ideas in which truth might prevail. It could not undo the bigotry of Parliament. We hope that it will have more success today.

[1] . Thi s is an allusion to an Estonian saying, “E very ram has its Michaelmas ” , which historically refers to slaughtering wethers (castrated rams) in autumn around Michaelmas day (29 September) but is nowadays used to mean that one cannot escape one ’ s fate.

[2] . The Estonian words avaldama /avaldaja mean both publish/publisher and disclose/discloser.

[3] . Samuel D. Warren and Louis D. Brandeis, “ The Right to Privacy ” , 4(5) Harvard Law Review 193-201 (1890). The article is available in its entirety at www.jstor.org/stable/1321160?origin=JSTOR-pdf&seq=1#page_scan_tab_contents ( u pdated on 23 March 2015) .

[4] . See Gert Brüggemeier , Aurelia Colombi Ciacchi, Patrick O ’ Callaghan (eds.), Personality Rights in European Torts Law , Cambridge University Press, Cambridge 2010, p. 18 and note 51.

[5] . See Jean-Christophe Saint-Paul (ed.), “Droits de la personnalité ” , 2013 at http://boutique.lexisnexis.fr/jcshop3/355401/fiche_produit.htm (Updated on 23 March 2015). Translation: “Person alit y rights are the set of legal prerogatives relating to moral interests (identit y , private life , hon o ur) and the huma n body o r the means of realising them (correspond e nce, home, image); they are exercised by anyone with legal personality ( natural persons or legal entities ) and are enforced by means of actions in the civil courts ( injunctions to desist , claims for damages ) and the criminal courts .

This issue, at the crossroads of civil law (persons, contra c ts, property ), criminal law and human rights law and also civil and criminal procedure , has given rise to an extensive body of case-law at domestic and European level, based on various sources of national law ( C ivil Code, Criminal Code, Data Processing and Civil Liberties Act , Freedom of the Press Act ) and international law ( European Convention on Human Rights , International Covenant on Civil and Political Rights , Universal Declaration of Human Rights , Charte r of Fu ndamenta l Rights ), involving a judicial balanc ing exercise between protection of the perso n and of other valu e s such as freedom of expression o r evidentiary needs.”

[6] . See the Marlene Dietrich Case , BGH 1 ZR 49/97, at Institute for Transnational Law – Foreign Law Translations, Texas University School of Law, at www.utexas.edu/law/academics/centers/transnational/work_new/german/case.php?id=726 ( u pdated on 23 March 2015) .

[7] . See http://hudoc.echr.coe.int/sites/eng/pages/search.aspx?i=001-126629 ( u pdated on 24 March 2015) .

[8] . The term is used in the literature: see Justin Hurwitz, “ Trust and Online Interaction ” , University of Pennsylvania Law Review , Vol. 161: 1579.

[9] . Jack M. Balkin, “ Old-School/New-School Speech Regulation ” , 127 Harvard Law Review 2296 (2014).

[10] . Ibid. , p. 2309.

[11] . Ibid. , p. 2311.

[12] . See E. Chevalier Huyttens (ed.), Discussion du Congrès national de Belgique 1830-1831 (Tome premier, 10 novembre-31 décembre 1830) , Brussels, Société typographique belge Adolphe Wahlen et Cie (1844). See Nothomb ’s speech, pp . 651-52.

The specific language of the 1831 Constitution was a compromise and did not reflect the prin cipled approach of the liberals who stood for Constitutionalism (writ large), but even this compromise, which we find today reproduced in Article 25 of the Belgian Constitution, states that “ When the author is known and resident in Belgium, neither the publisher, the printer nor the distributor can be prosecuted ” . Back to 1830?

[13] . Confirmed most recently by the Grand Chamber in Konstantin Markin v. Russia [GC], no. 30078/06, § 89, ECHR 2012 .

[14] . While the passage from Editorial Board of Pravoye Delo and Shtekel ( v. Ukraine , no. 33014/05, §§ 63-64, ECHR 2011 ) quoted in the judgment (see paragraph 128) seems to take a neutral position on the balance between the good and bad sides of the Internet, it is important to note that in that judgment the negative aspects did not prevail and the “ risk of harm ” argument was followed by a “ nevertheless ” , opting in favour of Internet freedoms.

[15] . Social media operators have already institutionalised over-censorship by allowing a policy of banning sites and posts which have been “ reported ” , without conducting a serious investigation into the matter. The policy adopted by Facebook is another victory for the troll mentality. Note that Facebook requires (all) user-imposed censorship to take place in a legal environment that grants service providers immunity under section 230 (a) of the Communications Decency Act. Imagine what will happen where there is no immunity.

[16] . “ Hate speech ” remains undefined. “ There is no universally accepted definition of ‘ hate speech ’ . The term encompasses a wide array of hateful messages, ranging from offensive, derogatory, abusive and negative stereotyping remarks and comments, to intimidating, inflammatory speech inciting violence against specific individuals and groups. Only the most egregious forms of hate speech, namely those constituting incitement to discrimination, hostility and violence, are generally considered unlawful ” (Report of the Special Rapporteur on minority issues, Rita Izsák (A/HRC/28/64), Human Rights Council, Twenty-eighth session).

See more at: www.ohchr.org/EN/NewsEvents/Pages/DisplayNews.aspx?NewsID=15716&LangID=E#sthash.XYM1WUqO.dpuf

The lack of an identifiable concept in twenty very different comments makes the application of the judgment unforeseeable.

[17] . The Court has rather clear requirements as to what amounts to an impermissible call for violence (see Sürek v. Turkey (no. 1) [GC], no. 26682/95, § 62, ECHR 1999 ‑ IV; DaÄŸtekin v. Turkey , no. 36215/97, 13 January 2005; Erbakan v. Turkey , no. 59405/00, § 56, 6 July 2006; Lindon, Otchakovsky-Laurens and July v. France [GC], nos. 21279/02 and 36448/02, § 56-58, ECHR 2007 ‑ IV; Otegi Mondragon v. Spain , no. 2034/07, § 54, ECHR 2011; and Vejdeland and Others v. Sweden , no. 1813/07 , § 55, 9 February 2012).

[18] . It was of relevance in Stoll v. Switzerland ( [GC], no. 69698/01, § 54-56, ECHR 2007 ‑ V), in the determination of the Government interest at stake, that no criminal action was taken against the applicant; hence the argument relating to the protection of national security was found to be of no relevance.

[19] . George Packer, “ Mute Button ” , The New Yorker , 13 April 2015.

[20] . It is remarkable to note that the rest of this quotation is not taken into consideration in this judgment. The original paragraph contains an important qualification: “ Accordingly, many laws are inevitably couched in terms which, to a greater or lesser extent, are vague and whose interpretation and application are questions of practice. ” In the present case, however, the issue is not the use of vague terms, for example the fact that the Directive employs a vague term when it refers to “ service provider ” . The issue was that there were two laws and the applicant company believed that the Directive was applicable as European Union law and as special law, whereas the Supreme Court took the view that the other law was applicable, because the service provider was a publisher.

[21] . In Editorial Board of Pravoye Delo and Shtekel (cited above), involving an Internet-related dispute under Article 10, although different from the one presented here, the Court found a violation of Article 10 on the sole basis that the interference was not adequately prescribed by law, taking account, amongst other things, of the special problems arising in the Internet era .

[22] . Sometimes incitement to violence is also mentioned.

[23] . Where the balancing exercise between these two rights has been undertaken by the national authorities in conformity with the criteria laid down in the Court ’ s case-law, the Court would require strong reasons to substitute its view for that of the domestic courts (see Axel Springer AG v. Germany [GC], no. 39954/08, § 88, 7 February 2012, and Von Hannover v. Germany (no. 2) [GC], nos. 40660/08 and 60641/08, § 107, ECHR 2012). It is probably for this reason that the Court ’ s analysis in the present case focuses on the sufficiency of the reasons provided by the domestic courts. However, the domestic courts have only selectively considered the criteria laid down in the Court ’ s case-law.

[24] . The principles quoted in the judgment refer to relevant and sufficient reasons as part of the consideration of the margin of appreciation. This makes sense, for example, when the national authorities provide reasons on the appropriateness of the means or the aims; if these are relevant the margin of appreciation may change and the level of scrutiny may diminish. In the present case, however, the requirement that relevant and sufficient reasons be given becomes detached from the margin of appreciation. A restriction of a Convention right, where the reasons for the limitation are not provided, is arbitrary and therefore cannot be held to be necessary in a democratic society. It is important for the rule of law and the exercise of rights that the restrictive measure itself contain reasons and that these are not made up ex post facto . It would be even less acceptable to allow this Court to speculate about possible reasons of its own motion.

[25] . See Ashby Donald and Others , cited above, § 34, and Węgrzynowski and Smolczewski , cited above, § 58.

[26] . We read the evaluation given by the Court as a declaration of lack of clarity with regard to the Supreme Court judgment: “ Therefore, the Supreme Court did not explicitly determine whether the applicant company was under an obligation to prevent the uploading of the comments to the website or whether it would have sufficed under domestic law for the applicant company to have removed the offending comments without delay after publication to escape liability under the Obligations Act ” (see paragraph 153 of the present judgment).

[27] . There is no way to exculpate the active intermediary as it should have known that illegal content had been posted and should have removed it immediately.

[28] . In a standard liability case the contribution of the victim is a matter for consideration. Delfi was blamed for the fact that the illegal content remained online for six weeks. Why did L. and his company not follow an article on a very important news portal concerning their economic activities and report these comments earlier?

[29] . The Court did not include this part of the established case-law in its analysis of journalistic responsibilities in paragraph 132 of the present judgment , where it mentions that the duty of the press “ is nevertheless to impart – in a manner consistent with its obligations and responsibilities – information and ideas on all matters of public interest ” . Here, the Court is dealing with a case involving a debate on a matter of public interest. This is not the place to express our doubts regarding the construction of press rights as duties, but we note that alternative language is also in use in our case-law: “ Not only do the media have the task of imparting such information and ideas; the public also has a right to receive them ” (see News Verlags GmbH & Co.KG v. Austria , no. 31457/96, § 56, ECHR 2000 ‑ I; Dupuis and Others v. France , no. 1914/02 , § 35, 7 June 2007; and Campos Dâmaso v. Portugal , no. 17107/05, § 31, 24 April 2008). See also Axel Springer AG (cited above), §§ 80 and 79.

[30] . Philip Hamburger, “ The Development of the Law of Seditious Libel and the Control of the Press ” , 37 Stanford Law Review 661, 673 (1985). See also John Feather, A History of British Publishing , Routledge, second edition (2002).