PARTLY DISSENTING OPINION OF JUDGE PAVLI

1. I have voted with the majority in all respects but one. I regret that I cannot agree with their finding that there has been no violation of Article 6 in respect of the alleged denial of access to investigative data. My disagreement concerns both the general approach followed by the majority under this heading and its specific findings on the merits.

2. The relevant principles regarding the rights of the accused, under Article 6 §§ 1 and 3(b) of the Convention, to obtain access to the investigative materials of the case against them, including any exculpatory evidence contained therein, were summarised in the Van Wesenbeeck v. Belgium (nos. 67496/10 and 52936/12, 23 May 2017) and Natunen v. Finland (no. 21022/04, 31 March 2009) cases, as described in paragraphs 85 and 86 of the current judgment. The defence rights of access to and disclosure of prosecution evidence are not absolute; however, any restrictions must be “strictly necessary” in view of the central role of equality of arms in the Article 6 architecture governing criminal due process (see Van Wesenbeeck , § 68).

3. The majority opinion concedes that denying the applicants access with respect to at least one of the evidentiary data sets (the “tagged” documents) “raises an issue under Article 6 § 3(b)” (see paragraph 91). However, at no time is a proper review carried out as to whether the restrictions meet the test of strict necessity. Instead, the relevant part of the judgment moves to the conclusion that, in view of certain supposed procedural failures by the defence and the availability of judicial review at national level, the applicants were not “denied a fair trial overall.”

4. I find the majority’s approach on this issue, which presents significant novelties for our Article 6 jurisprudence, problematic for a number of reasons. First, it fails to grant sufficient weight, in my view, to the serious disclosure issues raised by the applicants in the context of a highly complex criminal trial. By choosing to bypass the Van Wesenbeeck test in favour of a global finding that the overall fairness of the trial was not affected, the judgment does a disservice to the clarity and consistency of our case law, and misses an opportunity to weigh in on the complicated questions at the intersection of new technologies and high-volume evidentiary issues. On the merits, I find the reasons relied on by the majority to dismiss the applicants’ claims as partly inadmissible and partly insufficient.

5. The current case involves an unusually complex financial crimes prosecution, with multiple defendants and a very high volume of electronic material seized or otherwise obtained by the prosecution. This included at least four data sets of decreasing size: the “full collection of data” resulting from the investigation; the “tagged” data set of potentially relevant evidence, which the prosecution selected from the full data using advanced search technologies; the “investigation documents” resulting from further filtering of the tagged data set; and the actual evidence submitted in court, which included a subset of the investigation documents.

6. To recap, the applicants claim that they were denied adequate access to the first and second data sets. In particular, they were not allowed to conduct their own searches of the first data set, using the same software that had been extensively used by the prosecution. In addition, the prosecution - and the courts on at least three occasions - refused to grant them access to the list of documents “tagged” (as relevant or potentially relevant evidence) as a result of the prosecution’s searches, to summary lists of seized data, and to other investigative materials. The prosecution also denied the defence teams’ specific requests for access to “collections of emails and tapped telephone calls of individuals other than the applicants which had not been put in the investigation file” (see paragraph 75 of the judgment). This had given the prosecution a de facto monopoly on the use of advanced digital search facilities, without proper judicial review in the early stages of the process.

7. The majority’s choice to steer straight into a global fairness review means that there are no clear conclusions as to whether the access rights of the defence were restricted and, if so, to what extent. But some indirect conclusions can be drawn.

8. The judgment draws a distinction between the first and second data sets. With respect to the former, it finds in effect that electronic searches by the defence would amount to a “fishing expedition” in the absence of any specification as to “what could have been clarified by further searches” (see paragraph 90). In this context, involving millions of documents, this argument seems akin to a requirement to specify the precise location of a needle within a haystack. It also tends to turn on its head the prosecutorial duty to disclose any exculpatory evidence (more on this point below). The defence searches would have been no more of a fishing expedition than that carried out by the prosecution.

9. With the respect to the “tagged” data, the majority concedes that, under our case-law, the prosecution cannot exclude the defence from the process of assessing the importance of non-disclosed information; and that further searches of the tagged data set would have been technically “rather straightforward” (see paragraph 91). It is satisfied, however, with the Government’s argument that there was no prosecutorial duty under national law to create documents (specifically, lists of seized data) that did not already exist. There is no discussion of the implications of this national legislative choice for the purposes of Article 6 of the Convention.

10. It is worth recalling at this point that what is at stake in this case is a fundamental tenet of criminal due process, namely equality of arms. In the light of this cardinal principle, the majority’s overall approach seems insufficiently attuned to the complexities of electronic disclosure in criminal (or for that matter, civil) proceedings involving high-volume data; to the use of modern technological tools in this context; and to their combined implications for equality of arms. The assumption that standard rules of disclosure ought to apply unchanged in this context is one that, at the very least, needs to be tested.

11. A basic review of comparative law in jurisdictions with relevant experience in this field – a more extensive version of which would have been helpful in the novel context of this case – suggests that the prosecution is required to provide the defence with the fullest possible access to electronic investigative materials, including the ability to conduct their own searches, in terms and with capacities comparable to those of the prosecution. This is considered the minimum or baseline safeguard required in complex cases by the principle of equality of arms. [2]

12. Furthermore, in order to meet its duty of disclosing any exculpatory material in its possession, as both our case-law and Icelandic law require, the prosecution may also be required to show its good faith through additional proactive steps, for example by indexing the documents, providing files in a searchable format, and specifying any known exculpatory evidence. A finding, on the other hand, that the prosecution has engaged in a malicious “data dump,” in order to make it harder for the defence to analyse the data may lead to a conclusion of suppression of evidence. [3]

13. The above approach recognises that, even where the defence benefits from substantial access, the prosecution still holds distinct advantages: it will normally have had a longer period of time to analyse the evidence, generally greater analytical resources, and more intimate knowledge of the material, including in relation to any exculpatory elements. In view of these considerations, the current majority’s conclusion that the prosecution “did not hold any advantage over the defence” in the circumstances of the current case (see paragraph 90) seems rather far-fetched.

14. Secondly, it is a logical consequence of the preceding argument that full electronic disclosure in high-volume criminal investigations must be provided by default, that is, as a matter of standard prosecutorial practice and without the need for the defence to initiate and litigate a litany of procedural requests. Judicial oversight should, in principle, be exercised at this preliminary stage, when the terms of disclosure and searchability of data ought to be agreed and approved, whenever possible, by a judicial officer.

15. Emerging practice in the Council of Europe area is in line with this general approach. Thus, courts in at least two jurisdictions (the United Kingdom and Ireland) have approved in recent years the use of technology-assisted review, employing a form of artificial intelligence known as predictive coding, for the purposes of electronic disclosure in high-stakes civil litigation. [4] The rationale would apply with equal force in criminal cases of comparable complexity. [5] Again, the underlying premise for the use of such advanced technology is, of course, that both sides are granted the fullest possible access to begin with. And, secondly, that criminal-law frameworks and investigative practices are organised in such a way as to facilitate adequate access for the defence at the appropriate (that is, early) stage of proceedings.

16. On the facts of the current case, I would conclude that the applicants’ access rights were significantly restricted by virtue of the national authorities’ refusal to grant their defence teams meaningful and equitable access to the data sets at issue, and in particular the tagged data.

17. In addition, as a structural matter – and even allowing some margin for the specificities of each national system – the relevant Icelandic laws and prosecutorial practices do not appear to be organised in a way capable of ensuring adequate compliance with the disclosure rights of defendants in high-volume criminal cases. This is obvious from the Government’s explanations, whether of a legal or technical nature, as to why access to most of the additional data requested by the defence teams was not possible or warranted (see below for specifics). Equally, it is fair to say that the complex nature of the criminal investigation in the current case ought to have been apparent to the national authorities virtually from the outset.

18. This structural flaw would be a sufficient basis, in my view, for finding a violation of the overall fairness of proceedings under Article 6 § 1 of the Convention.

19. For the sake of argument, let me nevertheless address the specific justifications offered for the restrictions on the applicants’ access rights and whether they were “strictly necessary” under our case law.

20. While the Government puts forward a long list of arguments as to why the restrictions were justified, the majority appears to single out two main lines of justification, in paragraph 92. First, it places great emphasis on the fact that the applicants did not seek a formal court order “for access to the ‘full collection of data’ or for further [database] searches to be carried out” (see paragraph 92). This line of reasoning is objectionable on at least three grounds. To begin with, the Government’s argument under this heading is in the nature of a non-exhaustion claim; not having raised such an argument at the admissibility stage, the Government is normally estopped from relying on it on the merits. In second place, it does not appear to be entirely correct factually: the applicants did file, for example, requests for lists of documents that would be found by using a certain keyword, which were rejected by the domestic courts on unspecified grounds (see paragraph 83). But finally and most importantly, this line of reasoning is inconsistent with the duty of the prosecution to provide extensive disclosure by default in cases involving large volumes of electronic investigative data, as the only way to begin to ensure genuine equality of arms.

21. The same would apply to the second justification put forward by the majority, namely that the defence failed to specify what kind of additional evidence they were seeking, relying for example on “overviews of the seized items and their rough content” (see paragraph 92 in fine). With respect, this argument severely underestimates, in my view, the complexities of analysing large and interconnected amounts of investigative data, whether one is equipped with “merely” human intelligence or aided by artificial intelligence.

22. When one parses through the multiple lines of justification offered by the Government under this heading, it seems clear that concerns about a supposedly excessive effort played a significant role: the high volume of material that would have to be produced to meet the defence requests (see paragraph 79) or the need to re-import data into a certain software (see paragraph 83). One would hope that this is not what is meant by “strictly necessary” restrictions: it would be a sad day for Article 6 if mere convenience were to trump fundamental fair-trial rights.

23. This is not to say that extensive electronic disclosure of the kind envisaged in this opinion does not raise other thorny issues, including matters related to the protection of the basic rights and interests of third parties. The majority notes these concerns, but finds that they did not present “insurmountable obstacles”, a conclusion that I share on the grounds, inter alia, that the Government failed to show that it had properly and genuinely considered ways to reconcile disclosure to the defence with minimisation of the potential risks to third-party rights. [6] In any event, one would hope that these will be the kinds of issues that will preoccupy our future case-law in this area.

24. As a general matter, the reluctance to engage directly with the applicants’ claims under Article 6 § 3 of the Convention, with its various fundamental guarantees of criminal due process, has consequences – and not only of a theoretical nature. In the first place, it tends to water down the protections guaranteed by the five sub-headings of that provision, as if they were not capable of being violated on their own terms. Secondly, it constitutes a missed opportunity to provide clear and coherent guidance to national courts and other authorities on how these intricate Convention guarantees are to be interpreted and applied at the national level. This, in my view, is one of the central functions of this Court, which tends to be undermined if all competing considerations are reflexively force-fed into a less-than-transparent meat grinder labelled “the overall fairness of proceedings.”

25. To be clear, an ultimate ruling on overall fairness under Article 6 § 1 in no way precludes or replaces the need to carry out a proper review of claims made under the various limbs of Article 6 § 3. A cursory look at recent Grand Chamber judgments would confirm this: e.g. in Ibrahim and Others v. the United Kingdom ([GC], nos. 50541/08 and 3 others, 13 September 2016), a case involving denial of access to counsel and related rights, the Court proceeded, in a clear and systematic fashion, to review, first, whether the applicants’ rights under Article 6 § 3(c) had been restricted, and, secondly, whether the restrictions were justified under the relevant tests. Having addressed those two questions, the Grand Chamber went on to consider the impact of the restrictions on the overall fairness of proceedings for each of the four applicants.

26. The majority relies on Murtazaliyeva v. Russia ([GC], no. 36658/05, 18 December 2018) in opting to review the applicants’ claims from the point of view of Article 6 §§ 1 and 3(d) “taken together” (see paragraph 87). However, even in Murtazaliyeva , a case involving a failure to hear witnesses proposed by the defence, the Grand Chamber identified an (updated) three-step test, with the impact on the overall fairness of proceedings constituting only the final prong. The majority follows this same approach in assessing the current applicants’ claims under Article 6 § 3(d) (starting at paragraph 94), in contrast to the method chosen under paragraph 3(b) of the same article.

27. To put it another way, “taken together” is not the same as mashed together. A rigorous analysis of the complaints made under the various limbs of Article 6 § 3 has the added, and not insignificant, benefit of attenuating the inherently subjective nature of the global fairness review, thus enhancing the legitimacy of the final outcome.

28. Finally, had the majority chosen to engage more meaningfully with the challenges of complex criminal investigations in our high-tech age, it could have provided an incentive to the Icelandic authorities, and perhaps others, to do the same. That will have to wait for another day.

[1] . The Supreme Court has established in its case-law that the term “document” in this provision only applies to documents in their traditional form, which is on paper. Data in electronic form is not considered a document within the meaning of the provision but falls under the term “other materials” or “other data”.

[2] See, for example, the leading U.S. case of United States v. Skilling , 554 F.3d 529 (5 th Circuit 2009) (involving several hundred million electronic records).

[3] Ibid . See also U.S. Department of Justice’s electronic discovery protocol (2012), which recommends that, to meet the challenges of the digital age, prosecutors should provide to the defence a searchable file, as well as a table of contents with a high-level description of the “general categories of information” available within the material. See Recommendations for Electronically Stored Information (ESI) Discovery Production in Federal Criminal Cases , at Strategies, 5(b), available at: https://www.justice.gov/archives/dag/page/file/913236/download .

[4] See Pyrrho Investments Ltd v. MWB Property Ltd [2016] EWHC 256 (UK High Court); and Irish Bank Resolution Corporation Ltd & ors v. Quinn & ors [2015] IEHC 175 (High Court of Ireland).

[5] The predictive coding technology, already in wide use in some jurisdictions, allows parties to save a significant amount of time and resources in analysing large data sets. See, among other sources in the criminal context: Elle Byram, The Collision of the Courts and Predictive Coding: Defining Best Practices and Guidelines in Predictive Coding for Electronic Discovery , 29 SANTA CLARA COMPUTER & HIGH TECH. L.J. 675 (2013); and Brandon L. Garrett, Big Data and Due Process , 99 CORNELL L. REV. ONLINE 207 (2014).

[6] For an example of a considered approach to these questions, see item 10 (Security: Protecting Sensitive ESI Discovery from Unauthorized Access or Disclosure) of the U.S. Department of Justice’s electronic discovery protocol, note 2 above.