Proportionality of Website Blocking: Internet Connectivity Providers as Copyright Enforcers

Article 8(3) of the Infosoc Directive [15] obliges Member States to provide a possibility for copyright injunctions against intermediaries:

Member States shall ensure that rightholders are in a position to apply for an injunction against intermediaries whose services are used by a third party to infringe a copyright or related right.

Per Recital 59, the conditions and modalities are to be determined in national legislation.Essentially identical provisions also exist for other intellectual property rights in fine of Article 9(1)(a) and 11 of the Enforcement Directive. [16] Indeed, the E-Commerce Directive allows national law to provide specific injunctions against connectivity providers in Article 12(3), but on the other hand prohibits general monitoring obligations in 15(1):

[12(3)]. This Article shall not affect the possibility for a court or administrative authority, in accordance with Member States’ legal systems, of requiring the service provider to terminate or prevent an infringement.

[15(1)]. Member States shall not impose a general obligation on providers, when providing the services covered by Articles 12, 13 and 14, to monitor the information which they transmit or store, nor a general obligation actively to seek facts or circumstances indicating illegal activity.

While Internet service providers are not liable for information transmitted or stored, they may be subject to various obligations. [17] As connectivity and hosting providers provide a different kind of service, the conditions and scope of potential obligations also differ. [18]

The liability exemption of connectivity providers is based on neutrality, passivity and technical nature of automatic communication. According to the recital, this implies lack of knowledge and control over the transmitted information; deliberate collaboration in order to undertake illegal acts is also excluded. [19]

On the other hand, the exemption of hosting providers is conditional on awareness or knowledge of illegal activities or facts or circumstances from which illegality is apparent. Upon obtaining awareness of illegal material, hosting providers also need to act expeditiously to remove or disable access to it. The exemption does not apply if the provider has authority or control over the user and the content. [20] The awareness of facts relating to illegality appears to have been the grounds to exclude the operators of sites such as The Pirate Bay from the hosting defence; even if material on the site might not be infringing, its role in overall infringing activities has been apparent. [21]

In the latest case on hosting providers, L‘Oréal v eBay, it was held that the operator must not have an active role allowing it to have knowledge of the data stored. To measure awareness (or “neutrality” [22]), a standard of diligent economic operator was established as to when illegality should have become apparent. [23]

This interpretation is inapplicable to connectivity providers, because their liability exemption is not tied to knowledge or awareness in the same manner as hosting providers. [24] Further, their role is more passive, neutral and automatic, and transmitted data is transitory. They also have no obligation to act upon obtaining awareness of illegality [25] unless explicitly required by national law as provided by Article 15(2) of the E-Commerce Directive. [26]

The prohibition against imposing monitoring obligations of a general nature applies to all kinds of providers. Likewise, a general obligation to seek facts or circumstances indicating illegal activity is forbidden. However, monitoring (in a fairly narrow meaning) “in a specific case” by orders of national authorities is possible. [27] Court or administrative authority may also require termination of present or prevention of specific future infringement. [28] In particular, in UPC Telekabel Wien rather severe requirements regarding judicial review were imposed on generic orders targeting a website. [29] In contrast, the Advocate General had more explicitly suggested that a specific kind of generic blocking order would not amount to a general monitoring obligation. [30]

Only hosting providers may be subjected to “duties of care” in order to detect and prevent certain types of illegal activities. [31] National legislation may also establish procedures governing the removal or disabling of access to information (i.e. notice-and-takedown mechanisms). [32] Examples of court orders to prevent future infringements in the context of hosting providers were suspending infringing users’ accounts or making user identification on the marketplace easier. [33] In national courts, there have also been “stay-down” orders, obliging the hosting provider to ensure that a specific infringement is not repeated. Unless carefully constructed, these might be disproportionate or amount to a general monitoring obligation. [34]

All in all, the differences in the legal basis and the scope when comparing connectivity and hosting providers suggests that appropriateness of injunctions varies. Because for connectivity providers the conditions for liability exemption are broader, similar limitations could very well also apply to all types of injunctions. That is, it is likely more proportionate and technically more feasible to order a hosting provider to implement a certain kind of blocking than a connectivity provider. In contrast, issuing orders grounded on duties of care or knowledge assumed by a diligent economic operator at connectivity providers would not be appropriate. Also, accepting the premise of narrower injunctions, all the conditions in case law relating to hosting providers that restrict the orders (e.g. what constitutes a monitoring obligation and principles for weighing proportionality) would be prima facie valid. On the other hand, those conditions which expand the scope should be critically evaluated.

Table 1 summarizes private and public interests at stake; detailed elaboration is not possible here.

Intellectual property rights are not absolute, and it is impossible to completely protect them through enforcement. Therefore, trying to find the least restrictive means (LRM) to eliminate IPR infringement is misguided. This specifically applies to copyright, given the extensive exclusive rights also governing non-commercial activities, whereas in contrast e.g. trademarks or patents can be used by private individuals more freely. This was emphasized in Scarlet Extended and affirmed in UPC Telekabel Wien very explicitly as follows: [35]

There is, however, nothing whatsoever in the wording of [Article 17(2) of Charter stating that “Intellectual property shall be protected”] or in the Court’s case-law to suggest that that [intellectual property] right is inviolable and must for that reason be absolutely protected.

A better question is to ask which degree of IPR enforcement is deemed appropriate when balanced with other issues at stake, in particular other fundamental rights. [36] This reflects the underlying (and unresolved) policy issue on the level and constraints of enforcement an IPR holder is entitled to.

Rather than trying to find LRM in general, one must survey the means and their trade-offs (e.g. effectiveness, costs and other impacts). Based on this analysis, the appropriate balance between certain degrees of enforcement using a specific means and other rights can be considered. In UPC Telekabel Wien,issuing a generic order incurred requirements to the national procedure in order to ensure proportionality evaluation. [37] This was noted more explicitly in the Advocate General’s Opinion that proportionality cannot be evaluated if the necessary measures could not be reviewed. [38]

The Infosoc Directive does not provide guidance on how to balance injunctions of Article 8(3). [39] On the other hand, Article 3 of the Enforcement Directive applies to (all) measures, procedures and remedies necessary to ensure the enforcement of IPRs covered by the Directive. These shall be “fair and equitable and shall not be unnecessarily complicated or costly, or entail unreasonable time-limits or unwarranted delays”, and also “effective, proportionate and dissuasive and shall be applied in such a manner as to avoid the creation of barriers to legitimate trade and to provide for safeguards against their abuse”. [40]

It is challenging to realize all of these at the same time. In consequence, case-by-case balancing will needed. [41] In Table 2 these are classified as “strong” (pro-rights), “weak” (pro-users) and “neutral” principles. However, in different interpretative contexts–for example, against infringer versus a third party–the principles could have an entirely different meaning. [42] Per Article 2(1), a Member State may provide stronger enforcement only “in accordance with Article 3”, i.e. the balance must not be upset. [43]

Norrgård noted in 2005 that a national judge could reasonably adopt both a weak or strong enforcement ideology based on the discretion granted by the directive. [44] A balanced interpretation was suggested based on principles of Article 3, fundamental rights, context-sensitivity and the comparative method. [45] These have since then materialized in ECJ case law, and this paper also continues on that path. Similarly, Ohly has underlined the need of proportionality to balance effectiveness and dissuasiveness, referring in a similar fashion to taking due account of the specific characteristics of the case as noted in Recital 17. [46]

Both directives have been geared towards infringers. Applying principles of effectiveness and dissuasiveness against a faultless intermediary is questionable. [47] Indeed, in his opinion in L’Oréalv eBay, Advocate General Jääskinen was not convinced that the identical scope of injunctions available against the intermediary and the infringer would be a reasonable interpretation of Article 11 of the Enforcement Directive. [48] Similarly, the requirement for context-sensitivity would suggest caution with regard to third-party injunctions. [49] Further, the notions attached to injunctions are different and comparisons difficult between infringers and intermediaries as well as between legal systems. [50]

National law may specify the grounds for a proportionality evaluation or leave issuing the order at the court’s discretion. While this may give the national court leads on what to evaluate, it is important to note that EU law provides the minimum (and maximum) level of protection to various competing rights. This will perforce affect the court’s discretion and evaluation. [51]

As an example of national law, [52] in Finnish Copyright Act (404/1961, as amended by 679/2006) Section 60c(1): [53]

[...The] court [...] may order [...the] intermediary to discontinue [...] unless this can be regarded as unreasonable in view of the rights of the person making the material available to the public, the intermediary and the author.

On the other hand, many countries seem to have transposed Article 8(3) in a very minimal fashion, almost verbatim, or without substantial additional details or modalities. [54]

The latest blocking requests targeting connectivity providers at the user end in EU/EEA are illustrated in Table 3. In other EU/EEA countries, blocking has not been requested. The table notes the highest court which has made the latest decision. The rulings are final except the Netherlands (pending appeal to Supreme Court) and Austria (pending resolution after ECJ). The type of proceedings (civil and/or criminal) has also been noted.

“(X)” under “ruling” implies that the decision has been reversed in contrast to a lower court instance or earlier proceedings. This illustrates that while the general trend has been to order one or more ISPs to block access to a website, in some cases requests have also been rejected on various grounds. The most prominent reason for rejection (particularly in Ireland and Norway, and arguably in Germany) has been attributed to the lack of legal basis due to theimplementation (or lack thereof) of Article 8(3). [55]

Next, a summary of all cases is provided, with particular focus on issues relating to proportionality. The goal is to review arguments used in ECJ and national decisions as a step of formulating the criteria for balancing.

As for blocking injunctions, [56] in L’Oréal v eBay, it was held that courts of member states must be able to order online marketplaces to take measures to stop current infringements as well as – if justified by the circumstances – prevent future ones. These must be effective, proportionate and dissuasive and must not create barriers to legitimate trade. The measures must also be fair and not excessively costly. A couple of examples were provided. [57] This provides little guidance in the present context.

Scarlet Extended applied the aforementioned proportionality principles to whether a connectivity provider could be ordered to install a comprehensive filtering system. [58] The system would have monitored all the data relating to all customers in order to prevent any future infringement of claimants’ intellectual property rights. ECJ held that it would have constituted a general monitoring obligation in violation of Article 15(1); nonetheless, proportionality was also considered. [59] As introduced in Promusicae, protection of the IPR must be balanced against other fundamental rights. Specifically, the connectivity provider’s freedom to conduct a business would be inappropriately balanced against the interest to protect the IPR, because the order would require installing a complicated, costly and permanent system at the provider’s own expense. The cost and complicated nature of the system were also against Article 3(1) of the Enforcement Directive. [60] The system would also be questionable from the perspective of users’ data protection and freedom of expression, as it would have applied to all the communications and would also have misidentified and blocked lawful communications. [61] This judgment is relevant, but given the intrusiveness of the contested filtering system, the constraints it provides for blocking orders are limited. [62]

UPC Telekabel Wien concerned a more restricted form of blocking, only one specified site.In summary, the ECJ in principle accepted generic orders (where adopted measures are chosen by the provider), but imposed judicial review requirements that might make issuing them unattractive for national courts. This restricted the applicability of generic orders and transformed generic orders into court-approved specific orders. [63]The Advocate General was more straightforward and suggested rejecting them instead of crafting requirements. In the end, the essential result was the same. [64] More detailed analysis follows.

The ECJ held that national law must be transposed in a manner allowing a fair balance to be struck between competing interests. The law must also be interpreted in the manner that takes fundamental rights and proportionality into consideration. [65] The fundamental rights conflict was construed between Charter Articles 17(2) (IPR holders’ right to property), 16 (the provider’s freedom to conduct a business), and 11 (users’ freedom of information). [66]

Business freedom was characterized as inter alia the right to freely use the economic, technical and financial resources available to the provider. An injunction was considered to constrain the free use of resources, as it obliges taking measures which may represent a significant cost, have considerable impact on the organization of activities, or require difficult and complex technical solutions. However, such an injunction was held not to infringe on the very substance of conducting business. [67] With a generic order, the provider can choose measures that are best adapted to the resources, abilities and challenges facing it. The provider can also avoid being in breach of an order by proving at the execution stage that all reasonable measures have been taken. Specifically, no unreasonable sacrifices are required. [68] However, legal certainty was held to require that the evaluation of reasonableness–i.e. which means can be expected–is done in court prior to issuing any sanctions due to the lack of sufficient compliance. [69] Essentially, the ECJ transferred the evaluation of proportionality from the ordering phase to the execution phase, with a mandate to provide for court review. In consequence, the provider can obtain confirmation of compliance in court by first doing a minimal implementation. Very likely this will decrease the use of generic orders. [70]

With generic orders and uncertainty, the A.G. had been concerned with providers opting for intrusive means, endangering users’ freedom of information or facing the risk of disobeying the order. [71]The ECJ’s approach was to impose requirements on the adopted measures and require allowing court review for both providers and users. In particular, measures must comply with Internet users’ right to freedom of information: they must be strictly targeted to end infringement, and must not affect users lawfully accessing information. [72] The A.G. provided an example of collateral damage by referring to the ECtHR ruling Yildirim v Turkey, where the whole Google Sites service was blocked when only one site hosted there should have been targeted. This violated the freedom of expression guaranteed by Article 10 ECHR in various ways. [73]

The ECJ held that national courts must be able to verify that the measures are appropriate. This may not be possible if the provider chooses the measures and their use is not contested. In consequence, in this case users must be provided locus standito assert their rights before the court. [74] It is submitted that this also applies to national courts issuing specific orders, unless proportionality has also been reviewed from the users’ perspective. [75]

As noted supra, the IPR is not inviolable in that it must be absolutely protected. It is possible that a complete cessation of IPR infringement is not achieved, if for example some measures that would achieve the result would not be reasonable for the provider, or because means do not exist or the goal is not in practice achievable. [76] The adopted measures should still be “sufficiently effective”or at least discourage users from accessing the site and make accessing the site difficult to achieve. [77] As such, even if infringement is not completely prevented, the fair balance and proportionality in the light of Article 52(1) is achieved provided that freedom of information to lawful material is not unnecessarily prevented and that blocking prevents access or at least makes it difficult and seriously discouraged. [78]

The ECJ’s emphasis on effectiveness, a high degree of IPR protection, and the means which may place significant obligations on the provider could be read to imply that the balance should be shifted more towards IPR protection. This reading must be rejected. The ECJ appears to have taken no clear stance on the balance of protecting the IPR and business freedom.

The fourth question was whether it was proportional to order a provider to implement specific means even if these incurred significant costs and the blocking could be easily circumvented without technical knowledge. The ECJ’s final answer displays naiveté or evasion, because those issues were also in the background in the third question.The problem is apparent when considering the ineffectiveness of blocking in general, and that there may not be “sufficiently effective” and “seriously discouraging” means that would not also be complicated, costly and/or too intrusive. The ECJ provided no guidance on the provider’s required degree of involvement. The problem of unlawful material on a site which also includes lawful material was also not addressed. In consequence, essentially both balancing exercises (IPR holder vs provider and IPR holder vs user, respectively) were effectively deferred to national evaluation or a future referral. Therefore, the A.G.’s guidelines on the fourth question are valuable and will be used tentatively as a basis.

When assessing suitability, the A.G. noted that protecting the rights of others was a valid grounds for limitation. It was questionable whether ineffectiveness of blocking could imply unsuitability, but he opined that this would not necessarily be the case. While users can circumvent blocking, it does not follow that everyone of these users will do so. If the user learns about a website’s illegality, he might forgo accessing the site. Finally, even if many users could circumvent blocking, all of them cannot. Also, even if the same site is available through another domain name or IP address, it does not prima facie mean that blocking would be unsuitable. The users here could also forgo accessing the site, and they need to use a search engine to find the site. With repeated blocking, it is also more difficult to find the site with search engines. In consequence, the A.G. held that generally speaking even somewhat ineffective means are not unsuitable. [79]

When assessing necessity, appropriateness and proportionality strictosensu, the A.G. noted that the least restrictive means should be adopted. It would be the responsibility of the national court to make the assessment based on guidelines and the non-exhaustive list of considerations provided. First, the estimated efficacy of the order would be one factor to consider. However, the possibility to circumvent blocking would not in general exclude the means. Second, the complexity, costs and duration needs to be assessed. The A.G. expected that this may be a test case, and more blocking injunctions might be pending. If one of them were disproportionate under those three grounds, the court might need to consider whether proportionality would be ensured by making the rightholder liable for costs wholly or in part. Third, the rightholders should have some means against an infringing site, but notably the provider has no contractual relationship with the infringer. This implies that the rightholder must primarily target the infringer or the infringer’s provider. Fourth, an injunction must not undermine the legitimate business of providing Internet connectivity. In general, the A.G. underlined freedom of expression and its necessity in a democratic society as essential and access to the Internet as an essential right. [80]

All of these seem applicable against the backdrop of the ECJ’s judgment, with a potential caveat with regard to the third criterion. The ECJ had rejected contractual relationship as a prerequisite for injunction. [81]The A.G. had done likewise, but had opined that it is “obvious” that an intermediary without contractual link can “in no circumstances” be held unconditionally responsible for stopping infringements, laying out implications for proportionality of the third consideration. [82] Further, “in many cases such intermediaries are best placed” in Recital 59 (and respectively Recital 23 in the Enforcement Directive) implies that in some cases intermediaries are not best placed. In that case, the infringer should be targeted. This also conveys an additional issue of which intermediary is best placed; typically there are three to five connectivity providers in the “chain” between the user and the source. As an example, the A.G. opined that an injunction would be appropriate against a website connected by a non-European provider, because the website and its operators often cannot be prosecuted. [83] E contrario websites operated in the EU or connected by providers in the EU might be evaluated differently.

Orders may target the infringing website’s connectivity provider, a provider in the middle of the Internet, or a provider at the user end. [84]A contractual relationship between the provider and the website operator exists only in the first case. In the second case, some providers may obtain indirect financial gain (through transit agreements between providers), but in the third case there is not even such indirect benefit. [85] This study focuses on blocking at the user end, because actions against infringing customers are in many aspects more proportionate, and in any case different from those against a completely third party.

A few interim observations are in order before moving on to examine national case law. The rights to balance in enforcement have been confirmed. However, the evaluation has been rather thin except for the A.G. opinion on the fourth question (unaddressed by the ECJ) in UPC Telekabel Wien. In all likelihood, in the future courts will mainly issue specific orders. Also, neither the ECJ nor the A.G. addressed the required level of efficiency when constraining a provider’s rights. [86]The ECJ’s position on the provider’s function to discourage users could possibly be seen as approving the suitability of educating the users of the site’s illegality. Finally, the A.G. made a number of points that the ECJ did not get a chance to address. He noted that proportionality would be affected by the right holder’s possibility of primarily targeting the infringer or the infringer’s connectivity provider.Second, effectiveness could be used as a factor both in suitability (with limited impact) and necessity. Third, the assumption seemed to be that the connectivity provider might bear the costs, but an alternative cost model could also be used particularly if the number of blocking requests grew.

A Usenet newsgroup service Newzbin,whose content consisted primarily of infringing files, was ordered to stop copyright infringement with a restricted injunction. [87] The site ceased operations, but essentially the same service reappeared overseas, and the claimants sought an order against the connectivity provider British Telecom to block access to the new Newzbin2 site. Judge Arnold examined the issues in depth and inter alia found the specific injunction order to be compatible with Articles 12 and 15(1) of the E-Commerce Directive and Article 10 ECHR. Against the backdrop of the referral of L’Oréal v eBayand Scarlet Extended, he held that the court had jurisdiction to block the whole site. [88]

Arnold also considered and rejected four reasons why the court should exercise its discretion to decline the request. First, the order targeting the whole site would also affect other rightholders. This benefit would rather support issuing the order, and further, the incidence of non-infringing uses was considered de minimis. [89] Second, it was immaterial that accepting the request might lead to more requests in the future. [90] Third, ineffectiveness of blocking was not decisive. Specifically, users will need to acquire additional expertise to circumvent blocking. Even if they were able to do so, it was not clear that the users wished to expend the time and effort. Arnold also seemed to rely on advocacy research claiming that in Italy, the use of The Pirate Bay had been markedly reduced. All in all, he stated that the order would be justified even if it only prevented access to the site by a minority of users. [91] Fourth, Arnold evaluated proportionality against EHCR in the light of L’Oréal v eBay. He held that it was necessary to protect the rightholders’ property rights, and this clearly outweighed freedom of expression by the users and the connectivity provider, and even more clearly by the site operator. The order was narrow and targeted, it included safeguards against changes of circumstances and the cost of implementation would be modest and proportionate. He also rejected the requirement of notifying infringing files by URLs as being disproportionate and impracticable for the claimants when the site did not have a substantial proportion of non-infringing content. [92]

After the main judgment in Newzbin2, additional issues on details of the order were raised, some of which may be of relevance here. Arnold found it proportional to apply the order to all services which use BT’s Cleanfeed, but in essence excluded access and wholesale business from blocking. Other IP addresses and URLs, whose sole or predominant purpose was to enable or facilitate access to the Newzbin2 website, could also be added later to the block list without judicial determination. The request to allow temporary shutdown of blocking (e.g. due to operational reasons) without court or claimant approval was rejected. The relatively modest implementation costs would be borne by the connectivity provider, at least in this case. The legal fees were divided based on the merits of the arguments. [93] Later the blocking was also extended to other connectivity providers. [94]

Blocking of The Pirate Bay was soon to follow in Dramatico Entertainment, where it was held that the users and website operators infringed copyrights of the claimants in the UK. After Newzbin2, the connectivity providers have elected not to participate in court proceedings. [95] In a subsequent ruling, the injunction was issued on the terms the claimants had agreed with connectivity providers. Again, Arnold considered the discretion to grant the order. He stated that the proportionality of orders must be considered, referring to the principles of the Enforcement Directive and L’Oréal v eBay. While the terms of the order may be proportionate between the rightholders and connectivity providers, it was the duty of the court to ensure the proportionality from the perspective of those who are not before the court (in particular, the users). Given that the IP address of The Pirate Bay was not shared, blocking it was appropriate. [96] Since this judgment, specific terms of orders have been agreed in advance between the claimants and providers, and those have not been described in the ruling or scrutinized by the court. [97]

In Emi v Sky, blocking was extended to cover three more sites (KAT, HEET and Fenopy). In this case, proportionality was more extensively discussed in the light of recent English judgments. Specifically, Arnold had established a balancing test of conflicting fundamental rights in the context of an order to obtain identities of subscribers, and it had been later endorsed by the UK Supreme Court. [98] Also, he considered the referral questions posed in UPC Telekabel Wien. Arnold rather brusquely dismissed the third preliminary question on whether “prohibition of outcome” would be inappropriate, essentially stating that UK courts carefully consider such matters before any blocking order is made. [99] Arnold also seemed to dismiss the Austrian Supreme Court’s desire for European uniformity in assessing the proportionality of specific blocking measures by referring to the context-sensitivity of such evaluations. [100] The cost of compliance was also not relevant in the UK context because no connectivity provider had resisted making the orders on the basis that compliance would be unduly burdensome or costly. He still held that blocking order may be justified (emphasis added) even if it only prevents access by a minority of users, and that the efficacy depends on the precise form of the order. For example, orders in Dramatico Entertainment were likely less easy to circumvent due to the ability of the rightholders to revise the list of blocked resources. Finally, Arnold noted that evidence indicates that orders are reasonably effective, again pointing to the Italian study and the drop of The Pirate Bay in Alexa site popularity rankings. [101] In conclusion, the orders were deemed necessary and appropriate to protect IPR, and these interests outweighed the freedom of expression rights by users, connectivity providers and website operators. [102] Since then, blocking has extended to cover over 40 sites, but no new considerations of proportionality have come up. [103]

Access to FirstRow, a site consisting of user-generated streams of sports events, was also blocked in a similar manner. The claimant, FAPL, contended that the order would be proportionate because 1) the providers do not oppose the order and implementation costs are modest and proportionate; 2) the orders are necessary to protect copyrights infringed on a large scale and identifying and bringing proceedings against the operators of the website would be difficult, leaving no other effective remedy in this jurisdiction; 3) the orders are necessary or at least desirable to protect sporting objectives in general; 4) while also foreign content is available, the vast bulk of content infringes the rights of FAPL; and 5) the orders are narrow and targeted ones, and include safeguards against changes of circumstances. The orders will also likely be reasonably effective, even if not completely efficacious. Arnold did not object to these grounds. [104] FAPL had submitted evidence that FirstRow’s IP address is not shared, but this later seemed to turn out to be incorrect, leading to substantial overblocking. [105] Further, Arnold added an additional provision to the orders introduced earlier by Mann J, so that any website operator claiming to be affected by the order would have permission to apply to vary or discharge it. [106]

The Supreme Court of Denmark has issued two relevant judgments on preliminary injunctive relief. The first judgment in 2006 concerned a file transfer (FTP) server and TDC as the connectivity provider for that subscriber. Because the injunction would have essentially required terminating the Internet connection, it was held that Article 8(3) of the Infosoc Directive requires consideration with balancing of interests. [107] In a case concerning blocking users’ access to a website, it was submitted that injunctive relief should be rejected as ineffective under the balance of interest requirement, but this claim was denied with reference to obligations ofArticle 8(3). [108] In the second judgment, on blocking The Pirate Bay, in the Supreme Court it was mainly argued that the requested interim relief was too imprecise and the balance of interest did not favour injunction. This was rejected because the court felt the order was precise enough to be granted, and there would be no harm to the connectivity provider in a way which would make it disproportionate. [109]

In Sweden there have been no court cases on blocking at the user end. One reason may be the Swedish Copyright Act tying injunctive relief to contributing to infringement. [110]However, a website’s connectivity provider has been ordered to implement blocking. In Black Internet, the Svea Court of Appeal upheld the order to cease providing connectivity to The Pirate Bay. The provider’s argument was that the site also had other connectivity providers, the access prevention would not be effective and the site had been reorganized so that the order would not be effective. These did not prevent injunctive relief, and the balance of interest also favoured granting it. Further, the freedom of expression or other fundamental rights would not be restricted in a manner that would violate the ECHR. [111] Similarly, in Portlane, a connectivity provider was ordered to prevent access to a peer-to-peer tracker service “tracker.openbittorrent.com”. The court noted that the relief was fair and equitable and not unnecessarily complicated or costly as required by the Enforcement Directive. [112] Later, Portlane and other providers werepersuaded to stop servicing The Pirate Bay. [113]

In Norway, the request to block access to The Pirate Bay was rejected because providers had no obligation to remove or block illegal content in the implementation of Article 12 of the E-Commerce Directive. Therefore, there was no legal basis for granting an injunction. [114] Since then the Copyright Act has been amended in this respect, and a blocking request is expected in the near future. [115]

In Danish and (former) Norwegian statutes, injunctive relief shall not be granted if it would cause harm or inconvenience to the defendant in a manner that would be obviously disproportionate to the plaintiff’s interest in the injunction. Balancing thus occurs between the parties. However, because the court has some discretion, in special circumstances third-party interests may also be considered, and this has indirectly happened in case law. In contrast, in Sweden the balance of interests has been developed in case law and legal theory. In the two described cases, third-party interest was not considered. As such, in all three countries extraordinary circumstances seem to be required before a court would take third parties into account when balancing interests. [116] In Finland, the statute also provides for reasonableness balancing only between the IPR holder, intermediary and infringer. [117] These are in stark contrast to the more recent ECJ case law.

In Finland, the three largest providers have been ordered to block access to The Pirate Bay with specific orders to avoid uncertainty for the providers. Also, specifying and updating the list of blocked IP addresses and domain names was deemed to require judicial review. It was acknowledged that all measures differed with regard to their effectiveness, precision, cost and implementation time. By citing earlier reports, it was held that none of the proposed technological measures was so ineffective as to preclude issuing the order. The efficacy of blocking could be measured only after implementation. Costs and harm of the injunction were not unreasonable. [118] The blocking order affected the providers’ whole network, and for example the connectivity provided to other providers or the government could not be excluded. [119] It was considered proportionate that the providers needed to cover their own implementation expenses (ca. 10,000 euro) and legal fees (ca. 100,000 euro) each. [120] While the blocking must not endanger third parties’ communications, the minor amount of legal content on the website did not prohibit blocking. Targeting the infringers in Sweden was also not required by law. [121]

In The Netherlands, summary proceedings to block access to The Pirate Bay failed in July 2010. The judge argued that access could only be blocked from those who directly infringe copyrights rather than non-infringing visitors of the site, and that the individual infringers should have been targeted first. [122]

In new proceedings,Ziggo and XS4ALL were ordered to block access to the site, with the right holders having authority to update the list as needed. On subsidiarity the claimant had already sued The Pirate Bay operators and hosting providers, and proceedings against connectivity providers were appropriate; suing users was not needed. On proportionality, given the amount of illegal content on the site, the interests of the copyright holders outweighed those of ordinary Internet users. Also the necessity in a democratic society was briefly addressed. Claims as to effectiveness of blocking were made, which subsequently turned out to be unfounded. [123] Other providers have also been ordered to block access, but this time the list of sites needed to be updated in court. [124]

The first order was overturned on appeal in January 2014. [125] Based on recent research by the University of Amsterdam, the blocking was found to be ineffective, particularly if it targeted only one site, because most of the affected users would just use another website or circumvent blocking. [126] Only the images of art works of music albums and video covers and such material were found to be infringing, and blocking access to these was not sufficiently important when contrasted to the operators’ freedom to conduct a business. [127] It seemed that the court might have been more amenable if the claimants had requested blocking more sites at once instead of proceeding step-by-step. [128] This would have arguably made blocking more effective.

In Belgium, interlocutory proceedings to block access to The Pirate Bay also failed in July 2010. Immediate blocking was deemed disproportionate, especially since the site had already existed for years. [129] Blocking a fixed list of 11 domain names was established in September 2011 on appeal. The court held that subsidiarity is not required by Article 8(3) of the Infosoc Directive, and the order was also proportionate. [130] In a different set of cases, in criminal investigation against unidentified operators ofThe Pirate Bay, all connectivity providers were ordered to block access to the site in April 2012. This was appealed but upheld in February 2013. In October 2013, the Supreme Court upheld the decision that required providers to stop all current and future Pirate Bay domain names and monitor them. Strangely enough, it was not considered a general monitoring obligation. [131]

In Ireland, a blocking injunction was rejected in 2010 due to lacking implementation of Article 8(3) of the Infosoc Directive. [132] The law was amended, and six providers were ordered to block access to The Pirate Bay in June 2013. [133] A draft order had been agreed in advance between the parties, though concern with overblocking was voiced. The blocking list could be updated without judicial determination as in Newzbin2. Providers bore the cost of implementation and most of their legal expenses. The judge considered the draft order both proportionate and reasonable. [134]

In Italy the access to The Pirate Bay was blocked for a while in 2008–2009 as a criminal seizure, then the blocking was rejected and then again blocked, the latest order being from February 2010. [135] There is no explicit provision to issue connectivity providers a blocking order in civil proceedings, and the easier criminal procedure has been used. [136] There have also been subsequent orders to block other sites. [137] At the end of 2013, the law changed so that the regulatory authority AGCOM could block sites, though the EU Commission had serious doubts about various aspects of the draft regulation. [138]

In Germany there is no explicit statutory provision on Article 8(3) of the Infosoc Directive. The alternative “Störerhaftung” (disturber) liability imposes a high bar on these injunctions. A prevailing opinion has been that the Infosec Directive has been adequately implemented, and Störerhaftung is an acceptable solution as to “the conditions and modalities”. Website blocking has also been considered with strong scepticism. Yet UPC Telekabel Wien may require reconsideration through expanding interpretation of Störerhaftung or legislation. [139]

The lack of an explicit legal basis and evaluation against Störerhaftung has been adopted in a number of cases. [140] For example, in the Hamburg Court of Appeals ruling, the legal basis for injunction was deemed insufficient. It was also stated that Article 8(3) would not require providing for blocking injunctions. [141] Last, in the Cologne District Court ruling, a request to order a connectivity provider to block access to certain URLs in the eDonkey peer-to-peer networking service was likewise rejected. This was due to the lack of explicit legal basis, Störerhaftung liability not being met, and the lack of technical capabilities in preventing infringements. In consequence, the blocking would have been unreasonable and ineffective due to the small effort of circumventing such measures. [142] On the other hand, a preliminary injunction has been issued to an operator providing connectivity to The Pirate Bay. [143]

There has been no blocking in Portugal, but a request to block The Pirate Bay is to be expected. [144] In Spain, peer-to-peer indexing sites such as The Pirate Bay have been held to be legal, but theSinde Law could have changed this. There has not yet been news of website blocking, however. [145] In France, due to the efforts aimed at users through Hadopi, there has been no activity to block access to The Pirate Bay. Courts have issued orders at search engines, though. [146] In Greece, only two music sites have been blocked. [147] There has been no news of any blocking activity in Eastern Europe. [148]

Blocking cases essentially concern four parties,each with its own private interests: the IPR holder, provider, user and website operator. [149] The operator’s interests have not usually been considered explicitly, and they are also not in the main focus here. [150] Public interests in the background may also implicitly affect the weight court affords each private interest.

Proportionality has been considered in case law, though often in a summary fashion. Typically it is difficult to even find the criteria used in the evaluation. The criteria may also have been formulated at so high a level (as with most ECJ judgments) that applying them in concreto is challenging. However,there are also exceptions. The A.G. opinion on the fourth, unanswered question in UPC Telekabel Wien is useful, as well as UK cases Newzbin2 and FAPL v Sky. Also, several recurring themes can be noticed from other national case law. These concern in particular effectiveness (or lack thereof), impact of subsidiarity, effect on the third parties and implementation costs. [151] Further, in the literature, comprehensive lists of criteria have been formulated at least by Lodder and van der Maulen, [152]Husovec, [153]Savola [154] and Pihlajarinne. [155] These also have many recurring issues, but the interest for brevity precludes detailed discussion of each. Various requirements for adequate safeguards have also been suggested, but these are less relevant here. [156]

A generic balancing rule, which as such is not yet very helpful in concreto, could be formulated as follows. The more significant negative economic impact infringements cause to the IPR holder, the stronger enforcement mechanisms should be available, and in contrast the more significant negative impact of ordering such blocking needs to be on the other parties. Respectively, the more costs or constraints blocking causes to the provider, and the more it impacts freedom of information of the users, the more significant the losses must be to the IPR holder.

In consequence, in the crux are effectiveness and costs (or burden in general), respectively. If blocking is ineffective, it cannot mitigate the economic losses of IPR holders; such blocking would only have symbolic value. On the other hand, if blocking is cheap, non-intrusive and precise, it usually does not have a major immediate impact to providers and users. The more intrusive, expensive or vague blocking becomes, the stronger its justification and effectiveness needs to be in order to be proportionate.

In the following, concrete evaluation criteria are submitted. Interpretation of these criteria is based on the enforcement principles and fundamental rights.

The legal basis of blocking, the illegality of the source, has implications on proportionality. If some other court has already established the illegality (as with The Pirate Bay), [157] or the court issuing blocking does so when the operators have been served summons, more extensive blocking may be appropriate. Blocking is always dubious if the operators have not been represented and the illegality has not been subject to rigorous analysis. [158] Proxies that enable circumventing blocking provide a problematic example of this.

Is the intent of blocking to prevent site operators from (facilitating) making available copyrighted material, [159] blocking access from users as such, or to prevent users from copying and sharing the material? If illegality cannot be clearly attributed to the site (e.g. because it at most facilitates infringements by others), the legal basis of blocking the site may be in doubt. Because accessing the site is not as such infringing, the focus should be on preventing the main infringement, i.e. users downloading and sharing material. It might be sufficient to target blocking only to those users–i.e. consumers–who are expected to become infringers, rather than e.g. wholesale business, companies or the government. [160]

The scale of infringement also matters as to how invasive blocking mechanisms may be appropriate. For example, a site which has essentially no legal material at all could more easily be blocked as a whole, but this may not be appropriate for a site which has substantial legal uses or where the illegality is doubtful. Specifically, in some such cases URL blocking (for example) has been required instead of blocking the whole site. [161] Also, if blocking is targeting a site hosted in the EU/EEA, it also seems relevant to assess whether the site is operating legally in the origin state. [162]

The relevance of ineffectiveness has come up in almost every case as well as in the literature. [163] Users’ and operators’ ease to circumvent blocking, respectively, could be distinguished. The accuracy and completeness of measures may also be measured. [164] The possibility of revising the blocking list already allows taking into account the website operator’s actions, although blocking proxies would still be challenging. Overblocking will be discussed infra. Therefore, in practice the effectiveness–or more precisely,cimcumventability–of blocking from the user’s perspectiveis most relevant.

IPR holders have performed unverifiable advocacy research on the effectiveness of blocking. For example, Italian or Belgian studies, where effectiveness ranges between 73 and 80%, have been cited first when requesting blocking to support its effectiveness. Yet afterwards the ineffectiveness has been implicitly or explicitly argued to justify extending blocking. Alexa ratings have also been cited, but those are unreliable as they do not apparently account for those that arrive at the site through non-conventional channels. [165] In contrast, in a University of Amsterdam study, it was noted that blocking could in theory only affect the behaviour of those 27 to 28% of consumers who download or intend to download from illegal sources. Of these infringing consumers, the large majority (70-72%) was found to be non-responsive to blockingby finding other ways to access the same or a different site, for example. There was essentially no lasting effect, and even the awareness effect wore off quickly. In consequence, blocking affected only 4 to 6% of all users. [166] Connectivity providers have also similarly observed no significant impact on traffic levels. This is also supported by an intuitive finding that those who are already aware of the site (say, “The Pirate Bay”) can just enter the name in a search engine and obtain the list of proxy services on the first page of results. [167]

It is obvious that blocking may affect some users, especially the ones who are not already familiar with the sites and arrive at the site by web references or by searching for specific content. [168] Equally obvious is that blocking will not affect those regular users who know sites by name, and will continue using them or the alternatives through other, essentially equally easy means (e.g. proxies). This is exacerbated by the fact that infringing users are more knowledgeable and determined than users on average. [169] As such, blocking seems to – at most – accomplish a slight deterrence against some non-recurring users, provide “education” to those users, [170] and make it seem that IPR holders are doing something. It seems obvious that almost everyone downloading or sharing from clearly notable unauthorized sources is aware of its illegality. In consequence, education as a goal seems difficult to justify. [171] Indeed, the reasoning that users do not bother or will forgo accessing the site when encountering a block seems to apply at most with non-recurring users. [172] Therefore, the A.G.’s justification for ineffective blocking as deterrence is questionable. As will be discussed in the context of legitimacy infra, it is not obvious that subordinating providers to such a task is proportionate, especially if it incurs expenses.

Blocking with inefficient mechanisms usually results from having to resort to a least bad solution. While a more effective mechanism would in principle be desirable, usually one does not exist or must be rejected on other grounds. Therefore, the main alternative would be to reject the request. As noted supra, it seems easier for courts to accept inefficiency if a more effective order could not be granted. [173]

Acceptance of inefficient mechanisms therefore depends on the associated costs, degree of efficiency, and whether using blocking for (somewhat ineffective) educational, symbolic or political purposes is considered legitimate. It is submitted that blocking must be at least reasonably effective by substantially reducing IPR holders’ economic losses. Otherwise, it would be reduced to a symbolic gesture, a useless attempt to educate users, or as a political power play in the field of legal policy. [174]

With blocking, the IPR holder shifts the burden of enforcement to third parties, i.e. providers. [175] For providers, this implies implementation cost and expenses for legal services. In all the reviewed cases, the cost of implementation has been borne by providers, even if this is not required by EU legislation. [176] Given that costs have been modest, at most 10,000 euro for a large provider, these are only important in principle. [177] On the other hand, requiring the installation of a URL-blocking mechanism (for example)would likely be disproportionate. [178] Providers likely fear the expansion of mechanisms and costs. Arnold J and the A.G. anticipated this and noted that there may be reasons to cover some or all costs in the future. [179]

Given that blocking furthers the IPR holder’s private interest, and the IPR holder is also the sole beneficiary, it would not seem unreasonable for it to cover all costs. [180] This would guide it to do a rigorous assessment of which blocking would be economically justified. This might also allow adopting more efficient yet costlier means.

However, to date, implementation costs have been dwarfed by 10 to 30 times larger legal expenses. It would be disproportionate to require the provider also to bear the claimants’ costs if an order is granted. This applies in particular to countries where the provider is not even legally allowed to implement blocking without court approval. [181] Likewise, if the provider makes good arguments in court and the claimant’s request is restricted, the provider’s costs should also be compensated. [182] An appropriate default might be that both parties bear their own costs if the blocking order is granted essentially as requested.

Other burdens may also be relevant because the order will restrict how the provider is able to use its resources. [183] The order may also require reducing the security and reliability of the network. [184] Likewise, there may be limits (usually on the order of hundreds or thousands of entries) of blocked targets until the provider faces a serious performance penalty. In consequence, there may also be an issue of scalability if the number of blocking entries grows significantly. [185]

Usually blocking orders have been ordered without a time limit. This imposes restrictions on the provider because as a matter of business, services (e.g. BT’s Cleanfeed and equivalents) might need to be replaced or retired. Blocking orders should not restrict innovation and maintenance of services. [186] Also, even if the site disappears, blocking will stay in place until the order is discharged. In that case, the blocking might unduly affect the new user of (in particular) IP addresses. It would be appropriate for the order to have a fixed time limit of two to five years, but renewable as need be with a simple substantiated request. [187] Alternatively, the IPR holder could have the obligation to apply for discharging or varying the order if circumstances change or face liability for potential damages. [188]These would hopefully ensure that the list of blocked sites will contain only relevant and up-to-date entries.

The impact of this kind of burden must be accounted for, though admittedly the potential technical and innovation implications may be difficult to quantify.

The issue of addressing the infringement at its source has been a recurring theme in national case law as well as in the literature. While the ECJ confirmed that the connectivity provider at the user end may be issued a blocking order, it is not always necessarily best placed to end or prevent infringement. [189] In contrast to subscriber identification requests that the ECJ referred to, the user end provider is not the only provider that could implement blocking. Further, the A.G. opined that as a consideration of proportionality, if possible, primarily the infringer or infringer’s provider should be targeted. This might be the case in particular if an upstream provider is located in the EU. [190] It must also be observed that Article 8(2) of the Infosoc Directive does not require member states to provide for injunction (against infringers), except when infringement occurs on its territory. [191] While wider protection may be provided, this portrays the general principle that in some cases it is not unreasonable to require that the IPR holder pursue the case abroad. [192]

All of this seems to indicate that subsidiarity does have some role in evaluating proportionality. This is underlined especially if the measures would be burdensome, ineffective, the infringer is operating in EU, there has been no attempt to target him/her or to disconnect the site and/or the illegality is suspect. On the other hand, if there is evidence that the website has repeatedly changed providers (as with The Pirate Bay) or the administrators are anonymous and difficult to identify, [193] addressing the issue at the user end might be more easily proportionate.

The ECJ has emphasized the importance of freedom of information, and the A.G. also underlined its necessity for democracy and supporting the provider’s legitimate business. [194] Legal communications in particular must not be “unnecessarily” disturbed. [195] This implies that a minor disturbance might be acceptable in some circumstances; one example might be a site that has a small fraction of legal content. However, it is difficult to conceive of a scenario where overblocking–i.e. blocking affecting unrelated sites–would be acceptable. This comes up (and has come up) in particular with IP address blocking. For example, Arnold J has noted in two UK blocking cases that IP address blocking would not be appropriate if the address was shared. [196]

While overblocking has not yet caused significant problems with the top infringing sites, the ever-expanding blocking implies that this would become a problem. Specifically, for example, proxy sites and less significant websites will most likely almost always use a shared IP address. Thus IP address-based blocking would be excluded. On the other hand, more detailed forms of blocking (e.g. URL blocking) might be unavailable, and more generic ones such as DNS blocking might suffer from other problems. Given the crudeness of the website blocking as a tool, the result may be that blocking would be disproportionate.

Another difficult-to-qualify issue, as already touched on in the context of burdens, concerns the technical, architectural and security implications of blocking mechanisms, among others. All blocking mechanisms have their problems. DNS blocking undermines the security of the Internet by decreasing trust in domain name lookup services. IP blocking is simple, yet it has a risk of overblocking if not used with caution. There is also a danger of permanent blocking orders hindering innovation as some addresses cannot be recycled to new uses due to lingering blocking orders all over Europe. URL blocking may be expensive and it is incompatible with secure web connections. All of these might cause collateral damage through specific problems or as unreliability in general. [197]

The objective might be construed as IPR protection in such a manner that it does not disproportionately affect other rights. Prima facie this could be accepted as legitimate as is, but closer examination reveals doubts on the ultimate goal of blocking.

The general goal of blocking would seem to be protecting the private interests of IPR holders. [198] The political motives of IPR holders include strengthening the position against other parties, obtaining ever more effective enforcement mechanisms, and being seen to be doing something. The latter is essential to legitimize the existence of central copyright organizations to the actual copyright holders, as well as to affect public opinion that “piracy” cannot be tolerated. On the other hand, the more concrete motive would be to decrease the amount of infringement to regain some “lost” sales.

The crucial question is which and to which degree these underlying motives of blocking can be accepted as legitimate. Specifically, are enforcement mechanisms (especially at intermediaries’ expense) available only for efficient enforcement that produces economically quantifiable results? [199] Or is blocking also acceptable for educational, symbolic or politically motivated gestures? [200] There is likely no single right answer to this because the symbolic value of copyright enforcement is also tied to the public interests, with varied valuations. However, atleast using enforcement to enhance your own political agenda at others’ expense must be rejected.

The Advocate General suggested that, in general, specific blocking is not inappropriate for the purpose of furthering the aim of protecting the IPR holder’s rights. This was essentially based on a belief that many users would forgo accessing the site and would not bother to search for it because finding a way to access the material would become significantly more difficult. These also seemed to rest on an uneasy assumption that users would not be familiar with the illegality of the site, and blocking would have an educational purpose, affecting users’behaviour. This displays naiveté or undue optimism. Almost everybody seems to know that downloading and especially sharing is illegal; clearly illegal sites are typically also identifiable as such. Most infringers are likely recurring users who are not affected either by blocking or such subtle forms of education. Also, as demonstrated, searching is also trivial, and the increase in deterrence is minimal. As such, the A.G. seemed to overestimate the suitability of blocking as a deterring and educational mechanism. [201]

The University of Amsterdam study [202] and other public studies confirmed that the anticipated efficacy of deterring and also educating users is too optimistic. In the survey, 71% of infringing users reported that they did not intend to decrease or stop their infringing behaviour; after blocking, 77% continued. This was one of the grounds for the Appeals Court rejection of blocking in The Netherlands. It raised an issue, however: would an order be more easily proportionate if the IPR holder requested extensive blocking at once? The court rejected proceeding with a step-by-step approach, which perforce is more ineffective than blocking more sites at once. The court seemed to think “yes”, but it is not clear if this was a fair assessment given that The Pirate Bay was only a test case. [203] On the other hand, a step-by-step approach demonstrates IPR holders’ rhetorical twists: first it is claimed that blocking is effective and as such the request should be granted, and later insufficient effectiveness is used to argue for extending blocking. [204]

What is deemed to be the legitimate goal of blocking affects the evaluation of effectiveness and subsidiarity in suitability analysis. If symbolic gestures, education or politics are accepted, even completely ineffective means could be deemed appropriate, assuming that the negative impact on providers and users is minimal. If the goal is to make a significant impact on sharing and reduce the economic loss of IPR holders, a much higher degree of effectiveness may be required. The scale also tilts towards higher requirements if the negative impact on others increases. For example, Arnold J seemed to accept blocking even if it only affected a minority of users. [205] This might be suspect unless a broad notion of the goals of blocking is adopted.

Necessity concerns whether there are better means of achieving the objective. With the objective of preventing infringements, targeting the infringer (if possible) or infringer’s provider (especially if located in the EU, unless the operator has already switched providers multiple times) would be preferable to a patchwork of step-by-step blocking separately in all Member States. Blocking at the user end incurs much higher costs overall in the form of implementation costs and legal expenses than addressing the problem at its source. [206] The necessity to try alternative means first, if at all possible, was also underlined in the A.G. opinion in UPC Telekabel Wien. This is more important in particular if the cost and burden for the provider or the impact on users’ freedom of expression is not minimal.

Again, if the objective of blocking is deemed to (also) be symbolic, educational or political, there may not be much better means. It is in the IPR holder’s interest to increase the duties and responsibilities of intermediaries irrespective of efficacy. Further, blocking appears to be the second- or third-best solution to educating the users, with graduated response probably being preferable; there seems to be a mixed reaction to the relatively expensive and,in PR terms, uncertain trend to sue the users. [207] On the other hand, deploying ineffective mechanisms may also result in ridicule, especially from infringing website operators. It would be a much more powerful message from IPR holders to pursue infringers directly.

Criteria and their impact have already been discussed. Here only a general observation is made. As has already come up repeatedly, the goal of blocking as perceived and accepted by a court doing a proportionality evaluation incurs implications on the level of accepted proportionality. This is in particular the case with ineffective measures that do not cause significant costs or burden to the provider or the users. With a stricter focus on the economically quantifiable results–that is, proof that blocking significantly reduces the losses by IPR holders–ineffective mechanisms might be more easily rejected.

One point is worthy of noting,however. The effectiveness or lack thereof in a particular country or context can be measured only (and even then with difficulty) after blocking has been implemented. Erring on the side of restraint,however,might cause the issues to be addressed at higher courts or referred to the ECJ. This might also allow (or force) the IPR holder to conduct more rigorous studies of effectiveness in other countries. [208]