Fault, knowledge and risk within the framework of positive obligations under the European Convention on Human Rights

Abstract The European Court of Human Rights has consistently reiterated that positive obligations under the European Convention on Human Rights arise when state authorities know or ought to have known about the risk of harm. This article attempts to describe and assess the role of state knowledge in the framework of positive obligations, and to situate the Court’s approach to knowledge about risk within an intelligible framework of analysis. The main argument is that the assessment of state knowledge is imbued with normative considerations. The assessment of whether the state ‘ought to have known’ is intertwined with, first, concerns that positive obligations should not impose unreasonable burden on the state and, second, the establishment of causal links between state omissions and harm.


Introduction
The European Convention on Human Rights (ECHR) imposes positive obligations upon states to ensure the rights enshrined therein. Failure to fulfil these obligations has led to the establishment of state responsibility multiple times before the European Court of Human Rights (ECtHR or 'the Court'). 1 Due to their prominence in the Court's case law, authors have examined the importance of positive obligations in various subject areas 2 and the role of different principles, such as margin of appreciation and proportionality, that influence the determination of a violation. 3 Yet, some broader analytical questions underlying state responsibility for failure to fulfil positive obligations have remained understudied. One such question that underpins all positive obligations irrespective of the specific subject area concerns state knowledge, as it relates to the issues of risk of harm and foreseeability of harm. The Court has consistently reiterated that positive obligations arise when the state authorities know or ought to have known about the risk of harm. 4 This article seeks to analyse the role of state knowledge in the framework of positive obligations and to situate the Court's approach to knowledge about risk within an intelligible framework of analysis.
To do so, it is important to first provide an adequate frame of reference. This is initially done by clarifying in Section 2 the role of fault in the law of state responsibility more generally. Section 3 then clarifies whether fault is a necessary factor for triggering positive obligations under the ECHR or for determining a breach, and what distinctions have been introduced in the case law in this respect. Since fault in the context of these obligations has been framed as actual or putative knowledge by the state of the risk of harm, Section 4 examines how state knowledge is established in the case law and what principles are used for establishing the knowledge of an abstract organizational entity such as a state. Since the triggering of positive obligations and the determination of a breach are dependent on state knowledge about risk of harm, Section 5 enquires whether any requirements have been imposed as to the nature of this risk. Section 6 examines the role of victim's contributory fault and how it relates to state fault. Finally, Section 7 reflects upon the intertwinements between the elements of knowledge, causation, and reasonableness in the case law on positive obligations under the ECHR.
A brief note regarding methodology is also due. The case law selection is limited to judgments delivered under Articles 2 and 3 of the ECHR (the right to life, and the right not to be subjected to torture, inhuman or degrading treatment) since, in light of the gravity of the harm involved, protection by the state can be readily expected. In addition, these provisions have spawned a rich judicial output on substantive positive obligations, which are of sole concern in this article to the exclusion of any procedural obligations. 5 Priority is given to judgments delivered by the Grand Chamber, but Chamber judgments are also covered, including those that are given prominence in the existing literature as heralding important developments concerning positive obligations. The selected case law can adequately serve the purpose of exploring the role of state knowledge in the context of positive obligations. The method employed here is to explain the approach taken by the Court.

The role of fault in state responsibility
The starting point for the study of state knowledge is the work of the International Law Commission (ILC) on state responsibility, and in particular, on the elements of a wrongful act. The Articles on the Responsibility of States for Internationally Wrongful Acts (the Articles on State Responsibility) 6 define state responsibility as the attribution to the state of conduct (in the form of an act or omission) that breaches that state's international obligations. Every breach entails responsibility without any additional element such as 'fault'. 7 The Articles thus take an agnostic approach to the question of fault, since they are based on the principle of 'objective' responsibility. 8 This principle implies that to conclude whether a state is in breach, a comparison needs to be made between the conduct actually performed by the state and the conduct legally prescribed by the relevant primary obligation. 9 This approach was seen as desirable since not only might it be difficult to identify any subjective element of fault (whether in the form of intent, knowledge or negligence) of an organizational entity such as a state, 10 but equally difficult to prove it. 11 Although no requirement for fault is imposed ab extra, the primary obligations might incorporate such a requirement. 12 This is particularly the case where this primary obligation demands of the state to do something (i.e., it is a positive obligation) and the state fails to do it (i.e., it commits an omission). 13 The ILC Commentaries note that 'it may be difficult to isolate an "omission" from the surrounding circumstances which are relevant for the determination of responsibility'. 14 One such surrounding circumstance can be fault. 15 By referring to the Corfu Channel case, the ILC Commentary gives an example of how knowledge as a circumstance combined with omission gave rise to responsibility. 16 In the Corfu Channel case, the International Court of Justice (ICJ) held that it was a basis for Albanian responsibility that it knew, or must have known, of the presence of mines in its territorial waters and did nothing to warn third states of their presence. 17 Positive obligations raise particularly challenging questions because an omission is at their core. As a consequence, it might be open to question whether there is an obligation upon the state to do something in the first place. 18 Even if there is, there might be no clearly prescribed legal standard against which any omission can be compared so as to determine whether the state breached its positive obligation due to this omission. 19 These challenges can be approached in various ways ranging from so-called 'strict/absolute liability' to failure to exercise 'due diligence', which can be perceived as two ends of a spectrum. 'Strict/absolute' liability implies that once harm materializes, the state is responsible irrespective of any element of fault. In contrast, failure to exercise 'due diligence' leads to state responsibility only if the state was at fault since it knew (or should have known) about the risk of harm, but failed to take diligent measures to prevent it. 20 Positive obligations under the ECHR are of the latter type since they require fault in the form of knowledge or negligence.
The notion of 'fault' describes a blameworthy psychological attitude of the author of an act or omission. Such an attitude can be one of intent (i.e., the actor means to cause the harm), knowledge (i.e., the actor is aware that an omission might cause harm, but behaves differently from the way that could avoid the harm), or negligence (i.e., the actor might not know about the possible harm or risk, but it should have known and did not act in a diligent manner to avoid the harm). 21 In the context of positive obligations, no issue of intent arises. The Court has explicitly rejected the standard of intentional and wilful disregard of the risk of harm for purposes of assessing breach of positive obligations. 22 As to knowledge, it is necessary first to underscore that the state as an organizational entity cannot actually have this psychological and cognitive attitude. Rather, responsibility for omission can be established by comparing the actual state conduct with a conduct that one can legitimately expect from a normally directed and diligent state. 23 This suggests that the standard of fault is negligence, 24 and the type of negligence applied is objective. 25 However, what conduct can be expected from a diligent state can be dependent on the actual availability of relevant information about the risk of harm, which ought to not only be objectively assessed, 26 but also subjectively appreciated. It follows that actual knowledge and subjective appreciation of information by specific individuals who are part of the institutional structures of the state might be of relevance. 27 We shall see how the Court has approached these difficult issues in its abundant case law on positive obligations. A prior clarification is due to the effect that although the Court has referred to the term 'negligently', 28 the consistently used standard is 'knew or ought to have known' about the risk of harm. This standard (i.e., 'knew') reflects actual knowledge by the state. As an alternative, it also reflects negligence by the state (i.e., 'ought to have known'), which implies putative knowledge. As will be shown below, the distinction between the two is blurred in the case law. In light of the terminology deployed in the Court's judgments and the blurring of this distinction, in the 21 G. Palmisano, 'Fault', Max Planck Encyclopedia of Public International Law (2007). 22 Osman v. The United Kingdom, Judgment of 28 October 1998, [1998] ECHR Reports 1998-VIII, at para. 116. In Osman, the respondent government tried to argue that a failure to take preventive operational measures is present only when there is 'gross dereliction or willful disregard' of the authorities' duty to protect life. Pursuant to this argument, a state can be in breach of its positive obligation to take protective operational measures only if the authorities have manifested gross negligence in handling the situation. Alternatively, the respondent government argued in Osman that a state can be in breach of its positive obligation only if its authorities intentionally disregarded the risk to the victim. If these arguments were to be accepted, then the circle of situations when states are under the obligation to act to prevent harm to individuals would be considerably circumscribed. In Osman, the ECtHR explicitly rejected the arguments submitted by the UK: 'The Court does not accept the Government's view that the failure to perceive the risk to life in the circumstances known at the time or to take preventive measures to avoid that risk must be tantamount to gross negligence or willful disregard of the duty to protect life.' 23 A. Favre, 'Fault as an Element of the Illicit Act', (1964) 52 Georgetown Law Journal 555, at 561-2. See Palmisano, supra note 21, at para 17. 24 See Palmisano, ibid., where it is explained that the concept of fault is frequently presented as '"objective failure" to fulfill the content of an international obligation of conduct, imposing a certain degree, or standard, of due diligence (or vigilance, or care), rather than as an additional subjective condition of responsibility'. See also S. Somers, The European Convention on Human Rights as an Instrument of Tort Law (2018), at 185, where it is also explained that the positive obligations under the ECHR are 'very akin to negligence'. 25 Here, one can draw a comparative parallel with criminal law. Criminal law scholarship has shown that there is a subjective and an objective negligence. The first type implies that negligence is examined not only objectively, but also with reference to  One can draw a parallel with the approach to the appreciation of risk in the ILC Draft Articles on Prevention of Transboundary Harm from Hazardous Activities, in whose commentary it is stated that '[t]he notion of risk is thus to be taken objectively, as denoting an appreciation of possible harm resulting from an activity, which a properly informed observer had or ought to have had', 2001 YILC, supra note 8, para 14. following, the term 'knowledge' will be used generally to refer to the fault element required in the context of ECHR positive obligations.
3. Existence and breach of positive obligations under the ECHR Two initial questions concerning the precise role of state knowledge need to be addressed. First, is state knowledge a necessary precondition for the existence of a positive obligation? Second, is state knowledge an element relevant for the determination of whether the obligation has been breached?
In the context of Articles 2 and 3, the starting assumption is that the state is permanently under the positive obligation to ensure that individuals within its jurisdiction are not subjected to ill-treatment. Very similarly to French law on administrative liability, 29 the existence of an obligation is not under question; rather the state is assumed to be under a general obligation to administer competently, which flows from the very nature of state sovereignty. This is, however, an obligation framed at a very general level of abstraction that is detached from the concrete facts of the case. At a more concrete level, the Court has distinguished two types of substantive positive obligations: the obligation of taking protective operational measures, and the obligation of adopting effective regulatory framework to provide general protection to the society at large. 30

Operational measures for individualized protection
In relation to the first obligation, the Court has held that: : : : not every claimed risk to life can entail for the authorities a Convention requirement to take operational measures to prevent that risk from materializing. A positive obligation will arise, the Court has held, where it has been established that the authorities knew or ought to have known at the time of the existence of a real and immediate risk to the life of an identified individual or individuals from the criminal acts of a third party and they failed to take measures within the scope of their powers, which judged reasonable, might have been expected to avoid that risk. 31 It follows that the triggering of the positive obligation of taking protective operational measures to provide 'personal protection of one or more individuals identifiable in advance' requires actual or putative state knowledge. In this sense, knowledge about particular individuals at risk sets into motion the obligation. In the case law, this has been framed as the Osman test since Osman v. the United Kingdom was the Grand Chamber judgment where the Court framed the obligation. 32 Once the obligation is triggered, the determination of a breach is made by reference to the standard of reasonableness. 33 What can reasonably be expected from the state may be contingent on the actual or putative knowledge about the risk of harm the state had, and on the preciseness of this knowledge. It follows that state knowledge also plays a role in the determination of a breach. This determination implies asking the question whether the state took reasonable measures to provide individualized protection to the specific individual.

Effective regulatory framework for general protection
In contrast to the obligations of taking protective operational measures, the obligation upon the state to provide 'general protection to society' is assumed to be applicable all the time. 34 As the Court has framed it, the positive obligation upon the state to 'put in place a legislative and administrative framework designed to provide effective deterrence against threats to the right to life' 35 'must be construed as applying in the context of any activity, whether public or not, in which the right to life may be at risk'. 36 Knowledge that a particular individual identifiable in advance could be harmed is not required. Rather, it is required that the state is aware, or should have been aware, of the existence of a general problem. 37 A particular applicant in a particular case just happens to be a representative victim in relation to this general problem. State knowledge here is relevant for determining whether the state should have acted differently and, accordingly, whether it is in breach of its positive obligation. 38 It is important to initially distinguish between the positive obligation of taking protective operational measures, and the positive obligation of providing general protection, because they imply state knowledge in relation to different things. The first implies knowledge about a particular individual at a specific type of risk framed as 'real and immediate'. The standard of 'real and immediate' narrows the circumstances when this obligation can be breached. This standard will be examined in Section 5.1 below. The second implies state knowledge of a more general risk. Notably, an omission by the state can be scrutinized in relation to both positive obligations.

Actual knowledge versus putative knowledge
The determination of a breach of both positive obligations is contingent on actual or putative knowledge. This means that even if the state in fact had no knowledge of the risk of harm, the Court can also ask whether the state should have known or should have foreseen the harm. 39 Actual and putative knowledge are thus provided as alternatives. 34 Cevrioğlu v. Turkey In Mastromatteo, the Court first distinguished the issue of whether the state had to provide personal protection. It concluded that the obligation of taking protective operational measures was not triggered since the authorities could not have known that the particular victim could be an object of an attack. It then determined 'whether the system of alternative measures to imprisonment engages in itself the responsibility of the State under Article 2 of the Convention for the death of a passer-by inflicted by prisoners serving sentences for violent crimes who had been granted prison leave in accordance with that system'. The Court concluded that ' : : : this system in Italy provides sufficient protective measures for society. It is confirmed in this view by the statistics supplied by the respondent State, which show that the percentage of crimes committed by prisoners subject to a semi-custodial regime is very low'.

Different possible ways of assessing putative knowledge
The standard of 'ought to have known' has remained unclear. To better appreciate it, it is useful to make the following analytical distinctions. In particular, the question whether the state authorities 'ought to have known' of the existence of a risk of harm could be answered by reference to the following considerations. First, was the harm objectively or scientifically foreseeable at the relevant point in time so that the state authorities should have known about it? Second, should the state authorities have correctly assessed the risk of harm based on the information they would have had if they had carried out their obligations? Carrying out their obligations might imply consulting scientific studies and taking decisions accordingly. Third, should the state authorities have known of the risk, based on the information that was actually before them at the particular point in time? 40 The Court has not appreciated these three distinctions in its case law. The first alternative might be the most onerous for the state authorities since it implies, for example, post factum reference to scientific studies about risks of harm that were generally available at the time when the events were unfolding. 41 A possible problem that might emerge here is that the scientific evidence might have been inconclusive at the time when the state might have had to take protective measures. 42 In light of this uncertainty, it might be unreasonable to expect from the state to know about a risk of harm, when there was no objective standard against which any knowledge can be measured. 43 The second alternative (i.e., that state authorities should have correctly assessed the risk of harm based on the information they would have had if they had carried out their obligations) implies that the national authorities were, in fact, under an obligation that they did not fulfil. 44 This might be a premature conclusion since it might also be contingent upon the reasonableness of imposing such an obligation. 45 This obligation might be that of taking measures to predict possible risks of harm by drawing on scientific research or investigating and studying certain phenomenon or events to acquire knowledge. 46  The problem of inconclusive scientific evidence has led to the introduction of the principle of precaution in international law. As a principle for managing risk, precaution is based on the idea that scientific uncertainty should not be used as a justification for not taking protective measures. The precautionary principle can be contrasted with the preventive principle. The latter implies avoidance of known risks or risks that should have been known in light of objectively available evidence.  45 An initial assumption that the authorities had duties and therefore ought to have known about risks of harm, might be warranted or even taken as self-evident in some specific circumstances. See Premininy v. Russia, Judgment of 10 February 2011, [2011] ECHR, at para. 85, a case about a prisoner who was beaten by other prisoners. 46 This approach was applied in Talpis, supra note 30, at para. 118, where the majority could not conclusively determine that the victim was at an imminent risk, but added that the national authorities should have assessed the risk. See Partly Dissenting Opinion of Judge Spano in Talpis, who was skeptical of the majority's approach that implied that investigative passivity by the national authorities gave rise to putative knowledge. their obligations, 47 this might allow the state to use its own faulty omission to excuse itself for the resulting harm. 48 The third alternative (i.e., that the state authorities should have known of the risk based on the information that was in fact before them at the particular point in time) is the most favourable and the least onerous from the perspective of the state. The reason is that the appreciation of state knowledge is done with reference to the information that was actually before the state authorities, with no regard as to what information could have been available or should have been actively pursued by the state authorities.
Although the third alternative might be the least demanding and might imply a lesser likelihood of finding a breach in favour of the victim, it needs to be borne in mind that the state is limited in its capacity to augur potential harms. The existence of relevant knowledge about harms and risks of harm and the accuracy of this knowledge might be contingent upon the availability of state resources. Investment of resources might thus be necessary for the state to acquire knowledge and predict harm. Constant vigilance and 'active anticipation' 49 of harm by the state can be costly. 50 At this juncture, it is clear that the approach to state knowledge for the assessment of positive obligations is also intertwined with other considerations, such as reasonableness. The ECtHR has consistently reiterated that the scope of positive obligations to protect has to be reasonable and 'to be interpreted in such a way as not to impose an excessive burden on the authorities'. 51 This intertwinement will be further explored below in Section 7. The important point here is that for state authorities to play a proactive role in taking initiatives to gain knowledge about risks might be an arduous task. This is acknowledged by the Court with reference to the reasonableness standards. On the other hand, the Court has also held that the requirement for practical and effective protection of the rights and freedoms in the Convention 52 might necessitate that the authorities act proactively. It follows that inherent in every determination as to whether there is a breach of a positive obligation is the tension between effective protection of individual interests as embodied in the ECHR rights on the one hand, and practical considerations on the other. Examples of such considerations that trigger assessment of reasonableness are availability of resources, budgetary constraints or operational choices that need to be made by the national authorities. 53 47 On many occasions, the Court has found the respondent state to be under a procedural obligation to conduct studies so that relevant information about possible risks of harm is obtained or to consult with such studies. See E. Brems, 'Procedural Protection. An Examination of Procedural Safeguards Read into Substantive Convention Rights', in E. Brems

State knowledge necessary implies normative assessment
Further lack of clarity in the 'ought to have known' standard is added by the fact that often the Court does not conclusively establish in its judgments whether a state actually knew about the risk or whether it should have known. It is not clear which of these two standards is actually found fulfilled in the specific case. Consequently, although as a general principle, a distinction is made between actual versus putative knowledge, these two standards are merged when the specific case is analysed by the Court. For example, in Öneryildiz v. Turkey, the Court first said that it was impossible for the authorities not to have known of the risk that the rubbish tip posed to the people living nearby, 54 a determination that implied that the authorities actually knew. But then the Court proceeded to say '[i]t follows that the Turkish authorities at several levels knew or ought to have known that there was a real and immediate risk to the number of persons living near the Űmraniye municipal rubbish tip'. 55 The addition of the expression 'ought to have known' implies that it was not certain whether the national authorities knew, but in any case, they should have known.
This lack of clarity surrounding the distinction between actual knowledge and the 'ought to have known' standard can be related to the fact that the state as an organizational entity does not have awareness in the first place. As mentioned in Section 2 above, the state cannot know about things and, in this sense, the element of fault can only be inferred. The establishment of this element necessarily implies some normative judgments. These find expressions in the standards of 'ought to have known' and reasonableness.

Assessment of knowledge
Despite the inevitability of normative assessments, an attempt can be made to understand how the Court justifies a finding that a state had knowledge. In particular, how does the Court demonstrate in its judgments that an organizational entity such as a state knew or ought to have known?
The adoption of national legislation, sub-laws, and rules to address certain harms might be sufficient for presuming that the particular state knew about these harms. 56 Other standards, however, have also been applied. For example, references to 'objective scientific research' 57 might also be used for concluding that the state had knowledge about harms. The Court has also referred to different national reports that might have been prepared. 58 Communications, for example, in the form of letters or other documents, between various state institutions might also be used as a reference. 59 The state has wide regulatory functions, which makes it involved in many activities, such as issuing permits. This involvement can also enable the conclusion that the state knew about harms and risks of harm. For example, in Cevrioğlu v. Turkey, a case involving a child that drowned in a water pit at a construction site, the respondent state argued that the accident could not have been foreseeable since the construction in question had only recently started. The Court responded by holding that since a permit for this construction has been issued, it can be assumed that the state knew about it. 60 54 Öneryildiz, supra note 36, at para. 101. 55 Ibid., (emphasis added). 56 In O'Keeffe, supra note 49, at para. 168, the Grand Chamber established that the respondent state was aware in the 1970s of risks associated with sexual abuses of children by adults through, inter alia, 'its prosecution of such crimes at a significant rate'. Five judges from the Grand Chamber dissented in O'Keeffe and questioned this approach. See Joint Partly Dissenting Opinion of Judges Zupancic, Gyulumyan, Kalaydjieva, De Gaetano, and Wojtyczek, at para. 13. A similar approach was applied in Brincat, supra note 41, at para. 105, where the Court accepted that as early as 1987 laws were adopted to protect employees from asbestos and therefore since that date the state had known about the dangers associated with this substance. See also Öneryildiz, supra note 36, at paras. 98, 101. 57 Brincat, supra note 41, at para. 106. 58 Öneryildiz, supra note 36, at para. 98. 59 Nencheva, supra note 27, at paras. 121-2. 60 Cevrioğlu, supra note 34, at para. 68.
The nature of the activity within which harm materializes is also relevant for the assessment of state knowledge. If the activity is inherently dangerous in nature, then there is a normative expectation that the state continuously monitors that activity's operation, and thus knows, or should know, about risks. 61 Other circumstances, such as protection of children from a family member already convicted of sexual offences, can also imply an expectation from the state to monitor the situation. 62 It is not clear whether the existence of specific national rules regulating certain activities that might pose risks are sufficient and necessary for the establishment of state knowledge. Do such regulations need to be complemented with, for example, expert reports, so that the Court can conclude that the state knew or ought to have known? For example, in Öneryildiz v. Turkey, the Court placed emphasis on an expert report. 63 However, it also added that it was impossible for the authorities not to have known of the risks 'particularly as there were specific regulations on the matter'. 64 To what extent do contextual circumstances and general patterns suffice for the purpose of fulfilling the knowledge requirement? Are expert opinions and studies that identify patterns of problems in specific areas enough? For example, in Opuz v. Turkey, a domestic violence case, the Court explicitly took note of the existence of domestic violence as a general problem in the country and the measures undertaken in relation to this problem. This was necessary to set out the context within which the particular applicant had suffered harm. 65 However, for the purposes of establishing whether the authorities could have foreseen the abuse, the Court depicted in detail all the circumstances in which the abusive husband harmed the victims. There was a long history of assaults by the husband, and the victims had informed the authorities of the situation on many occasions, which gave grounds for the Court to conclude that the first limb of the Osman test (i.e., the state 'knew or ought to have known') was fulfilled. 66 The ECtHR's case law thus leans towards the conclusion that for the purpose of applying protective operational measures, constructive knowledge in the form of general awareness about the existence of general problematic patterns will not suffice. 67 Protective operational measures are activated when the authorities are aware that a specific individual could be at risk. 68

No benefit of hindsight
Positive obligations are assessed ex post facto by the Court. The problem that arises then concerns the question of which point in time should serve as a reference for determining whether the state knew or ought to have known about the risk of harm. Should this be the point in the past at which 61 it might have been expected from the state to fulfil its positive obligations? Should this be the point in the present when the Court makes its own assessment about events that happened in the past? The Court has emphasized that state knowledge should be assessed without the benefit of hindsight, 69 which means that the first of the two above-mentioned questions can be answered affirmatively. 70 For example, in Vilnes and Others v. Norway, the Court held that 'regard ought to be had to the knowledge possessed at the material timean assessment of liability ought not to be based on hindsight'. 71

Burden of proof
An important question for determining state knowledge concerns the burden of proof: Which party has to prove that the state knew or ought to have known? Does the Court put the onus upon the victim to prove the foreseeability of harm or is the burden on the respondent state to plead that the harm was not foreseeable?
Engagement with these questions has to start with the acknowledgement that, first, the Convention system is subsidiary to the domestic legal systems where the case has been litigated and evidence submitted, 72 and second, the Court has generally been very flexible in its approach to the burden and the standard of proof. 73 It would be beyond the scope of this article to engage with these issues. It would suffice to add that in principle the applicant has the burden of proof and that the Court is reluctant to second-guess the findings of fact made at national level. 74 In relation to the burden of proving knowledge, it is important to highlight that the Court has allowed flexibility. Although the state is not perceived to be an omniscient entity and thus to know about all activities that take place under its jurisdiction by the mere fact of the exercise of exclusive sovereignty, 75 the flexibility implies that important inferences are made from the mere fact that the state has control. In Öneryildiz v. Turkey, the Court held that: One can make a parallel with the Court's approach to the obligation not to refoule individuals upon risk of ill-treatment as implied under Art. 3 of the ECHR. This obligation can be perceived as a positive one (see V. Stoyanova, 'How Exceptional must "Very Exceptional" be? Non-refoulement, Socio-Economic Deprivation, and Paposhvili v Belgium', (2017) 29(4) International Journal of Refugee Law 580, at 584). The Court has clarified that 'the existence of the risk must be assessed primarily with reference to those facts which were known or ought to have been known to the Contracting State at the time of the expulsion; the Court is not precluded, however, from having regard to information which comes to light subsequent to the expulsion. This may be of value in confirming or refuting the appreciation that has been made by the Contracting Party or the well-foundedness or otherwise of an applicant's fears' (emphasis added  Corfu Channel case, supra note 17, at 18: 'the fact of : : : exclusive territorial control exercised by a State within its frontiers has a bearing upon the methods of proof available to establish the knowledge of that State as to such events'. authorities, which are often the only entities to have sufficient relevant knowledge to identify and establish the complex phenomena that might have caused such incidents. 76 This means that the Court is sensitive as to who might be in a better position to discharge the burden of proof. When the applicant is in a weaker position than the respondent state as regards the obtaining of evidence, there is a case to be made for transferring the burden of proof. This implies that it might be more realistic to ask the state to prove that it was not negligent than to ask the victim to prove negligence in how the state had managed the situation. An alternative approach that is possibly more favourable to the state is placing the evidential burden upon the state so that it is expected to initially provide an explanation for the omission.

The nature and the level of risk
It is clear from the case law that for a state to be in breach of its positive obligations it is enough if it knew or ought to have known about the risk of harm. State responsibility therefore centres on the concept of risk and how the state anticipates and deals with risks. This section asks whether the Court has imposed any standards as to the nature and level of this risk.

The 'real and immediate risk' standard
In relation to the positive obligation of adopting effective regulatory framework, no qualifiers have been added as to the nature of the risk of harm that the state knew or ought to have known about. 77 In contrast, in the context of the positive obligation of taking protective operational measures, the standard repeatedly invoked by the Court is one of 'real and immediate risk'. 78 It follows that the triggering and the finding of breach of the positive obligation of taking protective operational measures depends on whether the state knew or should have known about the 'real and immediate risk' of harm.
The Court has never specifically elaborated on the meaning of 'real and immediate risk' and has never engaged in any in-depth elucidation of the stringency of this standard. 79 'Real' risk could be understood as risk that is objectively given. 80 The adjective 'real' could also refer to the probability that the risk will actualize. 'Immediate risk' could be understood as risk that is 'present and continuing'. 81 Immediacy could also be more narrowly interpreted to refer to harm that was expected to 'materialise at any time'. 82 It follows that while 'real' can be linked with the probability or likelihood of the harm occurring, 'immediate' can be linked with its closeness, in terms of 76 Öneryildiz, supra note 36, at para. 93 (emphasis added); see also Stoyanovi, supra note 30, at para. 63. 77 See, for example, Cevrioğlu, supra note 34, at para. 51, where the Court referred to 'potential risk to human lives involved'. 78 It also needs to be added that often in the 'General principles' portion of the judgment, the Court refers to the 'real and immediate risk' standard, while never mentioning it or explaining whether it is fulfilled in the portion on the 'Application of those principles to the present case'. Often, the reason is that the case was such that the positive obligation of taking protective operational measures was not relevant, since the issue was rather of possible failure by the state to afford general protection to the society at large. See timing, to a relevant point in time. 83 'Immediate' can thus express a temporality, i.e., a specific timeframe within which harm could materialize. 'Real and immediate' has also been interpreted as implying a risk that is 'substantial or significant', 'not a remote or fanciful one', and 'real and ever-present'. 84 In light of this ambiguity, it is difficult to assess the stringency of the 'real and immediate risk' standard in the Court's case law. In some cases where the Court found it fulfilled, it is clear that the risk was specific, but of questionable imminence. 85 In other cases, the risk might be assessed as imminent, but its source was difficult to perceive. 86 At the same time, the Court has tended to expand the meaning of the term 'immediacy' and to invoke it in cases where one can hardly identify an immediate risk. 87 A question that has also remained open concerns the timeframe within which a risk can be considered as imminent. For example, in Öneryildiz v. Turkey, a case about a methane gas explosion at a garbage collection point that led to loss of life and destruction of property, the Court observed that: neither the reality nor the immediacy of the danger in question is in dispute, seeing that the risk of an explosion had clearly come into being long before it was highlighted in the report of 7 May 1991 and that, as the site continued to operate in the same conditions, that risk could only have increased during the period until it materialised on 28 April 1993. 88 It follows that the Court considered the risk of explosion to have been imminent years before the explosion actually happened. 89 This implies a very long timeframe of imminent risk.
In other circumstances, the Court has applied a much more restrictive timeframe. For example, in Fernandes de Oliveira v. Portugal, 90 a case about a patient voluntarily hospitalized in a psychiatric hospital who subsequently committed suicide, the Grand Chamber found no violation of Article 2. The reason for this finding was that 'it has not been established that the authorities knew or ought to have known that there was an immediate risk to A.J.'s life in the days' preceding the day when he committed suicide. The absence of immediacy of the risk was key in this case. In rejecting the approach of the Chamber, 91 the Grand Chamber accepted that 'there were no worrying signs in A.J.'s behaviour in the days immediately preceding his suicide : : : '. 92 In his dissent attached to the Grand Chamber judgment in Fernandes de Oliveira v. Portugal, Judge Pinto de Albuquerque observed that the gap of 26 days, with an episode of serious self-harm, between a failed suicide and a successful one, should have been enough to assess the risk of harm 83 The concept of imminence has also been linked to the probability of the risk of harm occurring rather than to its temporal closeness to the present. ECHR a case about the death of a boy after the collapse of a roof. 88 Öneryildiz, supra note 36, at para. 100. 89 Ibid.

90
Fernandes de Oliveira, supra note 30, at para. 131. 91 The Chamber found that Portugal was under a positive obligation to protect the applicant's son since the state was aware of an immediate risk to his life. Fernandes de Oliveir, ibid., at para. 75. 92 Fernandes de Oliveira, supra note 30, at paras. 131-2 (emphasis added). as immediate. 93 In contrast, the majority of the Grand Chamber preferred to assess immediacy with reference to a shorter timeframe, i.e., 'the days immediately preceding' the successful suicide.
The Grand Chamber's approach in Fernandes de Oliveira v. Portugal is consistent with the earlier medical negligence judgment of Lopes de Sousa Fernandes v. Portugal, where the Grand Chamber also invoked the immediacy of the risk to restrict the circumstances leading to responsibility for failure to fulfil positive obligations. In the latter case, the applicant complained under Article 2 of ECHR about the death of her husband after a hospital-acquired infection and a series of alleged medical failures. The Grand Chamber accepted that the responsibility of the state for failure to fulfil substantive positive obligations under Article 2 'may be engaged in respect of the acts and omissions of health-care providers', but only in 'very exceptional circumstances'. 94 To frame these 'exceptional circumstances' the Court invoked the immediacy of the harm as a criterion. 95 Two types of exceptional circumstance were framed: (i) 'where an individual patient's life is knowingly put in danger by denial of access to life-saving emergency treatment', 96 and (ii) 'where a systemic or structural dysfunction in hospital services results in a patient being deprived of access to life-saving emergency treatment and the authorities knew about or ought to have known about that risk : : : '. 97 The threshold of immediacy was thus framed as one of emergency, which, if reached, can allow the triggering of a positive obligation upon the states to protect the life of a particular patient. 98 As a response to the Grand Chamber's approach in Lopes de Sousa Fernandes, Judge Pinto de Albuquerque has argued that in situations revealing structural and systemic deficiencies, no requirement for imminent risk should be imposed. His argument is that: in situations of systemic or structural dysfunction which are known or ought to have been known to the authorities, the Osman test must be qualified, in so far as the requirement of "immediate risk" must be scaled down to one of "present risk". 99 He has also suggested reformulating the Osman test by scaling it down to 'present risk' in the context of domestic violence. 100 93 Partly Concurring, Party Dissenting Opinion of Judge Pinto de Albuquerque joined by Judge Harutyuyan in Fernandes de Oliveira, ibid., at para 22. 94 Lopes de Sousa Fernandes, supra note 74, at para. 190. Circumstances manifesting 'acts and omissions of health-care providers' were distinguished in the judgment from circumstances of 'alleged medical negligence'. It appears from the Grand Chamber's reasons that in the latter type of circumstances, the substantive positive obligation upon the state is less demanding. The Grand Chamber found that the specific case is one of 'medical negligence' and, therefore, 'Portugal's substantive positive obligations are limited to the setting-up of an adequate regulatory framework compelling hospitals, whether private or public, to adopt appropriate measures for the protection of patient's lives', para. 203 (emphasis added). Since the regulatory framework in Portugal did not disclose any shortcomings, the respondent state was not found in violation of Art. 2 of ECHR. See also para. 182 where the Grand Chamber referred to 'cases which concern allegations of mere medical negligence' (emphasis added). 95 Ibid., at para. 182 where the Grand Chamber referred to 'denial of immediate emergency care'. 96 Ibid., at para. 191 (emphasis added). 97 Ibid., at para. 192 (emphasis added). 98 My understanding of the Grand Chamber's reasoning is that if the above-mentioned exceptional circumstances are triggered, the positive obligation upon the state will be more demanding since it will include an obligation to protect the specific applicant. In contrast, when the exceptional circumstances are not applicable, but the case is only one of 'mere medical negligence' (see para. 182), the scope of the positive obligation under Art. 2 is narrower in that it does not include an individualized protection. The circumstances of Lopes de Sousa Fernandes' husband were found by the Court to be one of 'mere medical negligence'. [2013] ECHR, where the following reformulation of the Osman test was suggested: 'If a State knows or ought to know that a segment of its population, such as women, is subject to repeated violence and fails to prevent harm from befalling the members of that group of people when they face a present (but not yet imminent) risk, the State can be found responsible by omission for the resulting human rights violations.' Judge Pinto de Albuquerque's stance in support of rejecting the immediate risk test might be easy to understand, given the haphazard approach of the Court to the 'real and immediate risk' standard, as mentioned above. In particular, the standard has been invoked by the Court to conveniently limit the scope of the positive obligations in areas such as medical negligence. 101 In other cases, it is mentioned, but then it is left unexplained whether and how it is of any relevance. 102 At the same time, however, Judge Pinto de Albuquerque's position might be hard to understand given that, as clarified in Section 3 above, in situations of systemic and structural dysfunction, the positive obligation of affording general protection might be relevant. In the context of this obligation, no requirement for 'immediate risk' has been raised in principle in the Court's case law. In fact, the Court has not introduced clarifications as to the nature and level of the required risk that the state should know about. This implies a margin of flexibility. 103 However, one of the difficulties with finding a breach of the positive obligation of affording general protection to society is that the applicant has to demonstrate the causal link between the specific harm that he or she sustained and some general systemic or structural deficiencies posing risks the state knew, or ought to have known, about. 104 In contrast, when the victim is identifiable in advance as being at 'real and immediate risk', a situation that might call for protective operational measures of an ad hoc nature, the causal link between the harm sustained by the victim and the failure to take these measures might be easier to discern. In this sense, the immediacy of the risk makes it easier to find a causal connection between harm and failures by the state. 105 In addition, it might be unreasonable to expect the state to take protective operational measures of an ad hoc nature when a person is not exposed to an immediate risk of harm. This might be the case not only due to practical and financial considerations; concerns as to whether the state might have assumed too intrusive of a role might also arise. It should be added that protective operational measures by the state might be directed against other individuals (i.e., the alleged abusers). 106 This might create situations where the state's efforts to protect some individuals limit other individuals' rights. 107 One can also imagine situations where the state, by protecting an individual, infringes upon his/her own personal autonomy, which can also be controversial. 108 These possibilities add further strength to the argument that the circumstances when protective operational measures are called for should be an object of some constraint. The requirement for 'real and immediate risk', despite its ambiguous contours in the case law, provides such a restraining function. 101 Lopes de Sousa Fernandes, supra note 74, at paras 182, 184, 191-2. 102 Georgel and Georgeta Stoicescu, supra note 78, at paras 51, 56. 103 With these two positive obligations the Court seems to address two different types of risks. The obligation of adopting effective regulatory framework is arguably intended to address risks that are 'centrally and mass produced' and 'broadly distributed'. The obligation of taking protective operational measures, however, is arguably intended to address risks that are 'in relatively discrete units'. The Court has held that ' : : : the mere fact that the regulatory framework way be deficient in some respects is not sufficient in itself to raise an issue under Article 2 of the Convention. It must be shown to have operated to the patient's detriment'. This means that the applicant has to demonstrate that any deficiencies have concretely affected him/her. See

Man-made versus natural harms
Besides 'real and immediate risk', another distinction can be discerned in the case law based on the predictability of the risk of harm, namely the one between risks posed by human activities and those posed by natural hazards. In the sphere of 'dangerous activities of a man-made nature', 109 the case law suggests that the risk of harm is assumed to be more predictable and, accordingly, more demanding positive obligations are imposed upon the state. 110 In contrast, natural phenomena that are 'beyond human control' are assumed to imply less predictable risks. The reduced predictability might imply less demanding positive obligations to prevent the harm from materializing. 111 In Özel and Others v. Turkey, a case involving a natural disaster, namely an earthquake that caused the collapse of buildings leading to loss of life, the Court pointed out: : : : in connection with natural hazards, that the scope of the positive obligations imputable to the State in the particular circumstances would depend on the origin of the threat and the extent to which one or the other risk is susceptible to mitigation, and clearly affirmed that those obligations applied in so far as the circumstances of a particular case pointed to the imminence of a natural hazard that had been clearly identifiable, and especially where it concerned a recurring calamity affecting a distinct area developed for human habitation or use. 112 This quotation suggests that in the context of natural hazards, breach of positive obligations will be found only if the risk is imminent and clearly identifiable. These requirements have a limitative function that makes the finding of a breach less likely. At the same time, pursuant to the above quotation, the recurrence of the harm is perceived as an indication that the 'the natural hazard was clearly identifiable'.
The Court has thus far not elaborated on the meaning and stringency of the criteria of imminence and identifiability of the natural hazard nor has it resorted exclusively to these criteria in finding no violation. The Court's analysis in Özel and Others v. Turkey was restricted to the procedural aspect of Article 2. It thus remains to be seen how the Court will approach the criteria of imminence and identifiability in future cases involving natural hazards. The Court might take an approach similar to the one in Lopes de Sousa Fernandes where, as explained in the previous subsection, the immediacy of the harm was rendered of paramount importance for finding a breach in the area of medical negligence.

Contributory fault of the victim 113
The negligent conduct of the victim can be an important factor in the Court's assessment of state responsibility, especially when the victim faced a risk that he/she could appreciate and avoid. 114 109  ' : : : Article 2 of the Convention cannot be interpreted as guaranteeing to every individual an absolute level of security in any activity in which the right to life may be at stake, in particular when the person concerned bears a degree of responsibility for the accident having exposed himself to unjustified danger'. Gökdemir v. Turkey, Decision of 19 May The victim might have assumed risks by voluntarily exposing himself/herself to a known and appreciated risk. 115 In this sense, the victim had an understanding of the dangerous situation and voluntarily encountered it. 116 However, the level of appreciation by the victim might be unclear, and this also needs to be taken into account.
It has been a standard assertion in the case law that the conduct of the victim is a relevant factor in the assessment of a breach of positive obligation: Bearing in mind the difficulties involved in policing modern societies, the unpredictability of human conduct and the operational choices which must be made in terms of priorities and resources, the scope of the positive obligation must be interpreted in a way which does not impose an impossible and disproportionate burden on the authorities. 117 It follows that 'the unpredictability of human conduct', and accordingly, the possibility that victims themselves undertake risks, affects the determination whether the finding of a breach would be unreasonable, since it might lead to the imposition of a disproportionate burden on the state. At the same time, the Court's approach is also clear to the effect that the victim's faulty conduct cannot be an excuse for omissions by the state. The victim's fault cannot negate the very fact of the state's omissions, which might constitute the foundation of the finding that the state has failed to fulfil its positive obligations. This is the case even when the victim's contributory fault is engaged in circumstances when he/she has participated in unlawful activities leading to harm in this context. For example, in Öneryildiz v. Turkey the Court held that: In those circumstances [the state encouraged the integration of the rubbish tip, did not react to breaches of town-planning regulations and legitimized the existence of the slump by even taxing its inhabitants], it would be hard for the Government to maintain legitimately that any negligence or lack of foresight should be attributed to the victims of the accident of 28 April 1993 : : : 118 An important nuance, however, is that Turkey itself, the respondent state in Öneryildiz, endorsed and did not sanction the unlawful conduct of the victims. In contrast, when a state reacts to unlawful conduct that might lead to harm and actively tries to prevent it, then a different approach seems warranted.
Another important nuance emerging from Öneryildiz v. Turkey for assessment of contributory fault concerns the state's efforts to disseminate information so that individuals can take precautionary measures. If the state has disseminated relevant information enabling individuals to assess the risks that they might run because of the choices that they make, then it is less likely that the state will be found in breach of its positive obligations. 119 A victim's contributory fault can act as an intervening cause of his/her harm. Due to the contributory fault of the victim, it might not be possible to prove that the state's omission caused the harm. However, in light of the Court's flexible approach to causation in general, 120 this has not been an obstacle for finding states responsible for a failure to fulfil positive obligations. For example, in Cevrioğlu v. Turkey, [t]he Court acknowledges that the primary responsibility for the accident in the instant case lay with H.C. However, the failure of the State to enforce an effective inspection system may also be regarded as a relevant factor in these circumstances : : : 121 The standard of causation between the harm and any omission by Turkey in this case was framed at a very low level: the omission was simply viewed as 'a relevant factor', which sufficed for finding a breach. In comparison, in the similar case of Iliya Petrov v. Bulgaria, 122 where a boy was seriously harmed after entering a transformer and receiving an electric shock, the Court acknowledged that the boy was very reckless to enter such a dangerous place. At the same time, the Court highlighted that the 'decisive factor' leading to the incident was the inadequate control by the authorities regarding the safety of electric transformers. Thus, the 'decisive factor' appears to be more of an exacting standard for causation than a 'relevant factor'.
Finally, it needs to be highlighted that there might be cases where a submission by the respondent state that the victims knew about the risk of harm can backfire. This happened in Brincat and Others v. Malta, where the applicants complained about their exposure to asbestos. The respondent state argued that 'anyone in such a work environment would in any case be fully aware of the hazards involved'. The Court responded that this statement is 'in stark contrast to the Government's repeated argument that they (despite being employers and therefore well acquainted with such an environment) were for long unaware of the dangers'. 123 7. Intertwinement with causation and the standard of reasonableness As already suggested, states are not omniscient. It needs to be added that they are not almighty either. The positive obligations therefore need to be interpreted 'in such a way as not to impose an excessive burden on the authorities'. 124 The Court has referred to the requirement of not imposing 'impossible and disproportionate burden' 125 upon states. The Court has also referred to the standard of reasonableness: states are only expected to undertake 'reasonable steps to prevent illtreatment of which the authorities had, or ought to have had, knowledge'. 126 What constitutes reasonable steps in the particular circumstances of the case can be intertwined with considerations about the level and nature of state knowledge about risks of harm and the causal links between the harm and any omissions by the state. It might therefore be difficult to separate the test of reasonableness and the elements of knowledge and causation. Harm about which the state has comprehensive knowledge might call for more intervention and, accordingly, 119 the test of reasonableness might be applied in a more relaxed way. Harm which is more predictable and more immediate might imply a stronger protection claim on behalf of the victim. Harm that is more difficult to causally link to state omission and knowledge about risk might imply less-demanding obligations upon the state since it may not be reasonable to expect it to act.
Similar to the conclusion in Section 4.2 that the establishment of actual or putative knowledge about risk is underpinned by normative considerations, the determination of what can be reasonably expected from the state and the establishment of causal links between state omissions and harm also implies normative judgments. 127 These might be pulled in different directions by practical considerations on the one hand, and the consideration of effective protection of individual interests on the other.
Despite this intertwinement, analytical clarity demands a separation of the different elements and an inquiry as to their different roles. 128 In some circumstances it would make little sense to enquire what measures could have reasonably been taken to prevent harm, if the state authorities did not know about the risk of such harm in the first place. 129 In other circumstances, if the state could foresee concrete risk of harm with greater precision, then it is more reasonable to expect it to take protective measures. If the nature of the risk of harm is vaguer and its precise origins more difficult to foresee, then it might be less reasonable to impose positive obligations. There might also be circumstances where even if the state had taken measures, it might be questionable whether these could have prevented the harm the specific victim complained of. This might point to an absence of causality between harm and any omissions by the state. 130

Conclusion
Fault is an important element in the assessment of state responsibility for breach of positive obligations under the ECHR. More specifically, the ECtHR has consistently referred to the standard of 'knew or ought to have known' in its analysis, reflecting actual or putative knowledge by the state about risk of harm. This standard is applied to establish a breach of the positive obligation to take operational measures to protect a concrete individual who might have been at 'real and immediate risk' of harm. The standard of 'knew or ought to have known' is also applied for the establishment of a breach of the positive obligation of ensuring an effective regulatory framework aimed at providing general protection. Any deficiencies in this regulatory framework have to be causally linked to the harm sustained by the specific applicant.
Without making a clear and conclusive determination whether the state actually knew or should have known about the risk of harm in the particular case, the Court has referred to various factors to demonstrate actual or putative knowledge such as existence of national regulations, or scientific reports. The Court has also insisted that state knowledge is assessed with reference to the information possessed at the time when protective measures should have been forthcoming, and applicants cannot benefit from information that might have emerged subsequently. The Court has 127 This corresponds to the view that '[t]here is no risk which can even be described without reference to a value'. ECHR, at para. 96, where the Court determined that the harm was not foreseeable in the first place. See also Hiller v. Austria, supra note 84, at para. 53 (the psychiatric hospital staff could not have foreseen that a patient will escape and commit suicide). 130 See, for example, L.C.B. v. United Kingdom, Judgment of 9 June 1998, [1998] ECHR, at para. 40. In this case, the applicant sought to attribute her leukemia to her father's exposure to radiation. The Court observed that ' : : : it is clearly uncertain whether monitoring of the applicant's health in utero and from birth would have led to earlier diagnosis and medical intervention such as to diminish the severity of her disease'. also clarified that, although the burden of proof is on the applicant to demonstrate state knowledge, in some circumstances the state is in a better position to carry this burden.
Despite these principles that can be observed in the case law, the assessment of state knowledge is imbued with normative considerations. Their initial premise is that the state as an organizational entity cannot have awareness, and in this sense, it cannot know about anything. 'Ought to have known' is an inherently normative standard. The points of reference in the assessment as to whether the state 'ought to have known' have remained unclear. Inevitably, this assessment is intertwined with calculations of any causality between the harm sustained by the victim and any state omissions. The assessment of whether the state 'ought to have known' is also intertwined with concerns that positive obligations should not impose an unreasonable burden on the state.