The International Law of Intelligence Sharing in Multinational Military Operations: State Fault in Complicity

In my previous post in the series I explained how the fault element of state complicity rules is the single most important determinant of the scope of these rules, in the intelligence sharing context or otherwise. In this post I will elaborate on the different possible modes of fault, starting with the question of how fault can be attributed to abstract entities such as states, and then looking at how state fault can be established.

How Can States Be at Fault?

There is no conceptual barrier in thinking about the mental states – fault, culpability – of states. These are the mental states of concrete human beings, who are engaging in the conduct (the provision of assistance) that is attributable to the assisting state and that falls within the purview of Article 16 ASR or a similar complicity rule. Whatever the mental element required by the rule really is, it is the relevant state organ or agent that must act with such mental element, i.e. the fault lies with an individual or a group of individuals acting on the state’s behalf.

It is of course true that in numerous contexts the relevant culpability standard is objective – for example, whether the state acted reasonably or with due diligence to prevent some kind of harm, which it knew or could/ought/should have known about. But this is not because there is some kind of generally applicable presumption against subjective culpability standards, or because no fault is required at all. Rather, the question of fault or culpability is entirely context-specific and a matter of the proper interpretation to be given to the primary rule or the complicity rule being applied. It is entirely appropriate, for example, for state responsibility for the commission of genocide to depend on the existence of a specific genocidal intent on the part of the relevant state officials, while for the state’s responsibility for failing to prevent genocide to depend on an objective standard of conduct such as due diligence, coupled with an actual or constructive awareness of a risk of genocide occurring. A wholly different matter is the proof of subjective mental states such as knowledge or intent, which will again depend entirely on the context in which such proof is being sought, and will normally require resort to inferences.

State Fault in Article 16 ASR

The culpability element of state complicity is undoubtedly a complex topic, for several reasons. First, even though lawyers all over the world use concepts such as intention and knowledge all the time, the legal usage of such terms is frequently undertheorized and does not correspond to either how these terms are used in common parlance or in disciplines that otherwise employ them, such as philosophy. Second, there is a significant amount of variance in how these concepts are used in different legal cultures and even within such cultures (e.g. in civil as opposed to criminal contexts). Third, the central rule in the interstate context – Article 16 ASR – is frankly something of a mess, which I think even the greatest proponents of the ILC’s work on state responsibility would have to concede. The text of the article speaks only of the ‘knowledge of the circumstances of the internationally wrongful act’ in describing the state’s fault, whereas the commentary in several instances appears to impose a requirement of intention, without explaining its relationship to the knowledge required and while using inconsistent terminology in doing so.

I therefore spend much time in the article (which I will not do here) in unpacking different conceptions of knowledge and intent. With regard to the former, a key point is that in most instances references to knowledge are really about probabilistic foresight – a state (i.e. its officials) predicting the behaviour of the partner. And that foresight can relate to two distinct sets of circumstances, between which the ILC does not properly differentiate – the likelihood that the partner will commit a wrongful act (e.g. torture), and the likelihood that it will use the aid provided (e.g. intelligence) to do so. When it comes to intent, the term is substantially vague and can encompass several modes of liability.

The debate surrounding Article 16 has generally tended to revolve around a dilemma between intent and knowledge. Proponents of knowledge as the relevant fault standard argue that intent to facilitate the wrongful act would be very difficult to prove, that the text of an ILC codification project should be given priority over the commentary, and that knowledge is also the customary fault standard in international criminal law. An intent standard would be underinclusive – it would be unacceptable, so the argument goes, for a state which is (say) selling weapons to another state which it knows the latter would use to commit war crimes to escape responsibility under Article 16, simply on the basis that the assisting state’s officials did not provide the assistance with the purpose of facilitating or committing war crimes while knowing full well that this is what the assistance would be used for.

Proponents of the intent standard, however, emphasize the need to interpret the ILC text and the commentary together and argue that, realistically, a variant of an intent standard is the maximum that all states could agree to. As a matter of policy, a knowledge-based test would be overinclusive and would inhibit state cooperation that can help advance important public interests, e.g. the prevention of terrorism. The proponents of this view argue that state practice is replete with examples of assistance provided to partners with unclean hands, which militates against a customary rule prohibiting such assistance.  

The intent v. knowledge dilemma is in my view a false one, even if the tension between under- and overinclusiveness (apology and utopia) is very real. We can and should try to escape it. In the paper I thus try to explain the relationship between knowledge and intent and argue that the culpability element of Article 16 should encompass three distinct modes of fault: purpose/direct intent; oblique/indirect intent; and wilful blindness, defined as purpose to avoid acquiring information but for which the assisting state would have acted with indirect intent (see also Moynihan). If the assisting states provides assistance with the purpose/direct intent of facilitating that act, i.e. has the facilitation of the wrongful act as a conscious object of its action, it will be responsible under Article 16 even if prior to providing the assistance it only thought that the commission of the wrongful act and its facilitation were merely possible, rather than certain. The purpose/direct intent to facilitate will compensate for deficits in knowledge. If, on the other hand, the assisting state acts without such purpose, it will still be responsible under Article 16 if it was practically certain that the wrongdoing state will commit the wrongful act, practically certain that its own action of providing the assistance would facilitate the commission of the wrongful act, and chose to provide the assistance despite this practical certainty. In such cases the assisting state is acting with indirect or oblique intent.

To be clear, the argument presented here does not reduce the culpability requirement of Article 16 to knowledge, actual or constructive, of the circumstances of the wrongful act. I am not arguing that a state will be responsible for providing assistance whenever it knows, i.e. is practically certain, of the existence or future existence of these circumstances. The state must also have intent to facilitate the wrongful act, as the ILC provides in the commentary. But this intent is construed either as the purpose to facilitate in the sense of a conscious object to do so, or as foresight to the level of practical certainty that the assistance provided will facilitate the wrongful act coupled with a decision to proceed with the assistance nonetheless, consciously accepting that the assistance provided will facilitate the wrongful act, i.e. indirect or oblique intent.

This tripartite conception of the Article 16 ASR fault requirement is coherent, workable, and does not excessively inhibit interstate cooperation. In particular, there is nothing wrong with the culpability requirement having multiple, alternative forms, nor does the inclusion of indirect or oblique intent lead to overly burdensome, unreasonable outcomes. The culpability requirement as defined above does not capture situations of assistance in which there is only a risk that the assisted state will commit a wrongful act, or use the assistance given to it to commit the act (on which more momentarily). Nor does it capture situations in which the assisting state was negligent in acquiring and assessing information that could have, and should have, led it to practical certainty. The imposition of such a broad culpability requirement, which is essentially motivated by concerns that proving any intent element would be unworkable, can only be appropriate in context-specific settings, not for all possible violations of international law. Moreover, the difficulty of establishing intent, direct or indirect, should not be exaggerated; domestic and international courts do so routinely. Witness, for example the European Court recent finding in a series of cases (e.g. that of Alexei Navalny) that states acted with an ulterior purpose of suppressing political pluralism and dissent. Or consider the aggressive approach to the drawing of adverse inferences and reversing the burden of proof in the Court recent Carter judgment, in which it found that Russia assassinated Alexander Litvinenko.

Complicity under IHL and IHRL

The regime-specific complicity rules can be substantially different from Article 16 ASR in their scope and constituent elements – otherwise there would be little point to them. Their added value when compared to Article 16 is at least in covering state assistance to organized armed groups and other non-state actors. Article 16 does not apply to non-state actors by its own terms, and there appears to be little evidence of such a secondary rule of general applicability that would be divorced from the relevant primary rules. We can also say that the CA1 complicity rule is broader than Article 16 in that it covers not only assistance, but also ‘mere’ encouragement.

But the culpability elements of the IHL and IHRL complicity rules remain substantially uncertain. That said, at a minimum, the culpability element of IHL complicity should include all of the modes of fault in the general customary rule in Article 16 ASR. If states were prepared to accept those culpability standards for complicity in all existing rules of international law, regardless of the importance of the interests they protect, then surely the same standards would constitute the high-water mark above which any regime-specific complicity rule could not go. For instance, because its primary purpose is to prevent particularly grave harms against individuals (and states), it can be justified for IHL to impose lower, less stringent culpability requirements, inhibiting more potentially harmful interactions between and among states and non-state armed groups. But it could certainly not impose higher, more stringent culpability requirements than Article 16 ASR.

I thus argue that such less demanding requirements are appropriate for complicity in serious violations of IHL and IHRL, for example those that constitute grave breaches of the Geneva Conventions, deliberate attacks on civilians, or their subjection to torture or inhuman treatment. In such cases state fault should be based on conscious risk-taking – the state providing assistance to its partner while consciously disregarding a risk that the partner would commit a serious violation of IHL or IHRL would be liable for complicity if the partner does, in fact, commit such an act (e.g. torture) and the aid provided (e.g. intelligence) does facilitate it. If the risk ultimately materializes – the facilitated harm occurs – the state will be complicit in it.

This mode of liability (similar to that for inchoate violations of the Arms Trade Treaty and the non-refoulement rule) is functionally equivalent to recklessness or dolus eventualis. It is a subjective standard – not one of so-called constructive knowledge, based on some kind of objective duty of diligence. In the intelligence sharing context, in most situations even the most diligent of states could not have (probabilistically) known to near-certainty what the recipient of the aid would do. These are, in short, ‘pure’ situations of risk. Even after making all possible inquiries the assisting state perceives some degree of possibility that the assisted party would commit a serious violation of IHL or IHRL and would use the aid provided to do so, but it cannot be certain that this will or will not be the case. Holding states responsible for conscious risk-taking in facilitating harm to the most compelling interests protected by international law strikes the right balance between over- and under-inclusivity, whereas a negligence, due diligence standard is more appropriate in the context of positive obligations of prevention and protection.

Risk, Mitigation, and Balancing

Liability for complicity based on conscious risk-taking raises two important questions. First, the role played by any risk-mitigation measures that the assisting state might implement when providing assistance, such as intelligence (e.g. diplomatic or technical assurances). Second, the question whether risk-taking can ever be justified even when the threatened harm is a serious violation of IHL or IHRL, i.e. whether that risk can be balanced against the harm of suspending assistance.

As for the former, if they are effective mitigation measures can prevent the aid provided from facilitating the partner’s wrong – for example, the torture might not happen, or even if it does happen the aid provided would not facilitate it. But even if they don’t work mitigation measures are relevant for assessing the assisting state’s fault. They can show that the assisting state is not acting with the purpose to facilitate, or they can reduce its subjective appreciation of certainty or risk that its partner would commit the wrong or would use the aid provided to do so. Notice, for example, how in the example with which I started this series – that of the United States sharing intelligence with the Taliban – the US clearly took such mitigation measures by limiting the nature of the information being shared (‘We cut down the information we give the Taliban, they don’t get the full range of information we have’) because they knew that the Taliban could use that information to commit serious wrongs, e.g. kill or mistreat individuals who worked for the US in Afghanistan.

The latter question – whether risks of facilitating harm can ever be balanced against risks of suspending assistance – remains largely underdetermined. For example, suspending intelligence sharing with a ‘moderate’ armed group, due to fears that the group would use it to mistreat detainees, could carry with it the cost of potentially prolonging a conflict and causing more death/harm in the long run. I argue that it is difficult for complicity rules to incorporate such balancing in their internal logic, but that this can exceptionally be done under the justificatory rules of specific regimes (e.g. IHRL) or by relying on circumstances precluding wrongfulness such as necessity or distress, whose requirements are quite restrictive. It is very unlikely for such balancing arguments to succeed, and in any event it would not be permissible to balance causally proximate risks of harms such as torture with causally distant, speculative and incommensurate benefits (e.g. advancing geostrategic goals or earning money from weapons sales).

While policy-makers will inevitably engage in a ‘lesser evil’ type of balancing all the time, a state consciously taking a risk of facilitating harm will normally bear the consequences of judging matters wrongly – if the facilitated harm occurs, it will be legally liable for complicity in it, whatever good motives it might have had. Especially when serious violations of IHL and IHRL are at stake, the law cannot simply accept utilitarian balancing across the board. The circumstances in which such an argument could succeed are exceptional; one such case that can reasonably be argued to exist is that of imminent countervailing threats to human life. Consider, again, the example of the US sharing intelligence with the Taliban. Despite the mitigation measures it took, the risk of the shared intelligence facilitating reprisals by the Taliban against certain individuals was likely not reduced to zero. Yet the US nonetheless shared its intelligence and otherwise cooperated with the Taliban, despite the irreducible risks of doing so, in order to save the lives of the people waiting for evacuation from the Kabul airport. This may be an example of justifiable risk-taking.

A Question of Perspective: Ex Ante and Ex Post, Internal or External

As we have seen, state responsibility for complicity will largely depend on how the fault element of general or specific complicity rules is construed and applied. However, the assessment of what the assisting state knew or intended can happen at different points in time, for different purposes and by different actors – and this matters greatly.

First, it is the assisting state itself that must assess what it exactly ‘knows’ at the time the decision to provide or continue providing assistance, such as the sharing of intelligence, is being made. The abstraction of the state cannot, of course, know anything. What this means is that the assisting state’s relevant organs and agents, that is those individuals who are in fact tasked with making this decision, need to make an assessment of the information at their disposal from an ex ante perspective to determine whether it is practically certain that their partner would commit a wrongful act and would use the aid provided to do so, or whether there is some level of risk of this occurring. Such judgments will benefit from direct access to all of the information that the assisting state has at its disposal, including its confidential intelligence.

In this ex ante perspective, purpose-based forms of culpability, including willful blindness, are essentially irrelevant. No state official will openly admit to acting with an improper purpose, certainly not while a decision is being made. When it comes to knowledge to the level of practical certainty or some level of possibility or risk, from an ex ante perspective this will always be a probabilistic assessment. In particular, there is little practical difference (and no categorical one) between a state’s officials thinking that a future event, or the future behavior of a third party, are near-certain or are only highly likely.

Second, the assessment of what the relevant officials of the assisting state knew or intended can be done ex post, after the provision of assistance occurred, and after it did facilitate the wrongful act which was in fact committed. This can be done by some actor other than those organs of the assisting state who decided to provide assistance – e.g. a domestic or international court, a UN expert, a political body, a parliamentary committee, an inspector-general, a public inquiry, the authorities of some other state. Such external assessors or reviewers will have the benefit of hindsight – they will be able to know for a fact whether the violation occurred, to what extent the assistance contributed to it, and to what extent the mitigation efforts were effective or not. They will also have to make inferences about what the officials of the assisting state knew or intended. But even though they would have the benefit of hindsight, some ex post external assessors may be hampered in their ability to fully establish the facts, say because they will not have access to relevant evidence. It is unlikely, for instance, that the assisting state would provide its internal intelligence products to an international human rights body. In short, an ex post reviewing authority can, perhaps somewhat paradoxically, simultaneously know both more and less than the persons who made the decision to assist did ex ante.

Contrary to arguments sometimes made in the literature against intent or purpose-based culpability tests, an ex post external assessor will reasonably practicably be able to make findings that an assisting state acted with the purpose/direct intent to facilitate a wrongful act. The key point here is that in the vast majority of situations of state complicity an assisting state’s purpose to facilitate does not have to be established to anything approaching a criminal law, beyond a reasonable doubt standard of proof. Similarly, for findings of purpose, as with any finding of fact, a court or any similarly situation external assessor can rely on circumstantial evidence, resort to reasonable inferences or even reverse the burden of proof. Thus, while purpose-based culpability standards are of little practical relevance from the ex ante perspective, the same simply cannot be said for the ex post one.

In making their assessments both internal and external decision-makers will need to rely on two contextual factors. First, whether the assistance being provided, such as the sharing of intelligence, is a one-off event or part of a continuing cooperative relationship. Second, whether the future violation of international law that the partner might commit is itself a one-off, isolated event, or part of a pattern of similar violations. If there is a durable cooperative relationship the assisting state is more likely to be able to reliably assess what its partner will do. And a pattern of prior behavior in similar circumstances is likely the best possible evidence of what the partner will do in the future (cf. the judgment of the Court of Appeal of England and Wales in Campaign Against the Arms Trade).

In my next (and final) post in this series I will return to the two basic scenarios of complicity and intelligence sharing.

This post is part of a series; the other essays in the series are available here: Part I, Part II, Part IV.