Third State Responsibility to Prevent Genocide

Introduction

Genocide occupies a unique place among international crimes. It is often described, including by the International Court of Justice itself, as a crime that shocks the conscience of humankind and that the international community as a whole has an interest in preventing.¹ Yet the Convention on the Prevention and Punishment of the Crime of Genocide (the “Genocide Convention”) does not only criminalise genocide after the fact. Article I begins with a promise that is easy to overlook precisely because it comes before anything about punishment: the contracting parties “confirm that genocide, whether committed in time of peace or in time of war, is a crime under international law which they undertake to prevent and to punish”.² Prevention, in other words, is not simply a policy aspiration attached to the criminal prohibition. It is a separate, freestanding obligation.

The difficulty is that prevention, unlike punishment, does not have an obvious territorial home. Punishment is usually exercised by the State where the crime occurred, or by a State with some recognised jurisdictional link to it. Prevention is different. Genocide is frequently planned, organised, and carried out by the very State that would otherwise be expected to stop it, or by non-State groups operating with a degree of independence from any government. If the duty to prevent applied only to the State on whose territory genocide is occurring, it would in practice mean very little, since that State is usually the one least willing to prevent what it is itself doing or tolerating. This raises the central problem addressed here: can a State other than the one where genocide is occurring, a “third State”, incur international responsibility for failing to prevent genocide committed by another State or by non-State actors, and if so, what standard determines when that duty is triggered?

This question has moved from the realm of doctrine into the realm of live litigation. In 2007, the International Court of Justice decided Bosnia and Herzegovina v. Serbia and Montenegro, the first case to give sustained attention to what the duty to prevent actually requires.³ In 2022, the Court upheld its jurisdiction in The Gambia v. Myanmar, a case brought by a State with no direct connection to the alleged genocide, and confirmed that State’s standing on the basis that the obligations in the Convention are owed by every State party to every other State party.⁴ And in proceedings instituted in 2024, Nicaragua has argued before the Court that Germany’s provision of military equipment to Israel, together with its suspension of funding to the United Nations Relief and Works Agency for Palestine Refugees in the Near East (UNRWA), engages Germany’s own obligations under the Genocide Convention and international humanitarian law, as distinct from the obligations of Israel, which is not a party to those proceedings.⁵ Read together, these three cases sketch the outline of a body of law on third-State responsibility for failure to prevent genocide, but they leave a great deal of that outline unfilled.

The central argument here is that the duty to prevent genocide under Article I is not confined to the territorial State or to States that are themselves implicated in genocidal conduct. It is a positive, extraterritorial obligation of due diligence that can engage the responsibility of a third State where that State possesses the capacity to influence the situation and knows, or should reasonably know, of a serious risk of genocide. At the same time, the current framework remains imprecise about the threshold of risk that triggers the obligation, the degree of influence that is required, and the measures a third State is actually expected to take. The paper proposes a four-factor framework, organised around knowledge, risk, capacity, and available measures, and structured as a graduated rather than a binary obligation, as a way of giving the existing jurisprudence more workable content.

The paper does not attempt to resolve, and deliberately avoids taking a position on, any specific ongoing conflict. Contemporary situations, including the proceedings concerning Gaza, are referred to only as illustrations of a general legal problem that would exist regardless of which States happen to be involved in any particular case.

The legal architecture of the duty to prevent

Article I of the Genocide Convention is often read quickly, as a preamble-like statement leading into the more detailed provisions that follow. The ICJ rejected that reading in Bosnia v. Serbia. It considered, and dismissed, a reading, which it noted would support Serbia’s principal position, that the reference to prevention in Article I is absorbed into the more specific obligations set out later in the Convention, particularly the duty to punish, and generates no distinct obligation of its own. The undertaking in Article I, the Court held, “is not merely hortatory or purposive”: it imposes on contracting parties a direct obligation to prevent genocide, which “is not merged in the duty to punish, nor can it be regarded as simply a component of that duty”, and which has its own scope extending beyond the Convention’s other provisions.⁶

This finding matters because it means that a State can be found in breach of the Convention even where no organ or agent of that State physically participated in genocidal acts and even where no individual has yet been convicted of genocide. Breach of the duty to prevent is conceptually independent of breach of the duty to punish, and both are independent of the question whether the genocide that occurred is attributable to the respondent State’s own conduct. The Court added one qualification: a State can be held responsible for failing to prevent genocide only if genocide is actually committed.⁷

The Court was equally clear, however, that the obligation to prevent is not a guarantee of results. It is what international lawyers describe as an obligation of conduct rather than an obligation of result. A State does not become responsible merely because genocide occurs somewhere within its sphere of influence; it becomes responsible only where it “manifestly failed to take all measures to prevent genocide which were within its power, and which might have contributed to preventing the genocide”, the obligation being to employ all means reasonably available to it.⁸ This is a demanding but not impossible standard to meet, and the distinction between conduct and result runs through everything that follows: no State, however powerful, is expected to guarantee that genocide will never happen; every State is expected to try, using the tools genuinely available to it.

The consequence of framing the obligation this way is that the duty to prevent becomes, in substance, a due diligence obligation. Due diligence standards are already familiar elsewhere in international law, for example in relation to environmental harm or the treatment of foreign nationals, and they share a common structure: a State is not the insurer of a particular outcome, but it is expected to exercise a reasonable standard of care given what it knew or should have known and what it was capable of doing. The Court in Bosnia v. Serbia applied this due diligence approach in the context of genocide, holding that the notion of due diligence, “which calls for an assessment in concreto”, is of critical importance, so that the standard varies with the circumstances rather than operating as a fixed rule.⁹ That flexibility is useful, because it allows the obligation to apply to States with very different levels of power and influence, but it is also the source of much of the uncertainty discussed below.

The extraterritorial nature of the duty

If the duty to prevent were confined to a State’s own territory and its own organs, it would add little to the ordinary prohibition on committing genocide, since a State that is genuinely preventing genocide on its own territory by its own agents is, almost by definition, not committing genocide in the first place. The real significance of Article I lies in situations where the potential perpetrators are not under the direct authority of the State whose responsibility is in question.

The Court addressed this directly in Bosnia v. Serbia. It found that the acts of genocide at Srebrenica had been committed by members of the Bosnian Serb army (the VRS), and that those forces were neither de jure organs of the Federal Republic of Yugoslavia nor acting on its instructions or under its effective control, so that the genocide could not be attributed to the Respondent.¹⁰ That finding did not, however, exhaust the question of Serbia’s responsibility under Article I. The obligation to prevent, the Court explained, requires States to employ the means at their disposal to prevent persons or groups not directly under their authority from committing genocide, and whether a State has discharged it depends first on its “capacity to influence effectively the action of persons likely to commit, or already committing, genocide”.¹¹ Serbia had maintained that the relevant provisions engaged a State’s responsibility only for “failure to prevent or to punish acts of genocide committed by individuals within its territory or . . . its control”; the Court, however, held that the substantive obligations arising from Articles I and III “are not on their face limited by territory”.¹² The United Nations’ own account of the Convention has since described the prevention obligation, as interpreted by the ICJ, as having an extraterritorial scope.¹³

This holding is the doctrinal foundation for the entire idea of “third-State” responsibility for failure to prevent. It means that the relevant question is not whether the State being examined is the one committing genocide, or even the one on whose territory genocide is occurring, but whether that State had the capacity to influence the actors who were committing or about to commit genocide, and whether it used that capacity. On this reading, State A can, in principle, incur responsibility under Article I in respect of genocide being committed by State B, or by a non-State armed group operating within State B, provided State A had the requisite knowledge and capacity to influence the situation.

This is, admittedly, in some tension with the ordinary principle that a State’s obligations are generally understood by reference to its own territory and its own jurisdiction. But the tension is not unique to the Genocide Convention; it mirrors debates in human rights law about the circumstances in which a State’s obligations extend beyond its territory when it exercises effective control or authority over a situation abroad. What is distinctive about Article I, however, is that the Court did not require anything resembling effective control over territory. It instead referred to the State’s “capacity to influence effectively”, a much lower and more diffuse threshold, which in Serbia’s case it found in the strength of the political, military and financial links between the Federal Republic of Yugoslavia and the Republika Srpska and its army.¹⁴ That is precisely what makes the obligation potentially applicable to States that have no physical presence in the territory where genocide is occurring at all, but that have other forms of leverage: military, economic, or diplomatic.

The due diligence standard: breaking down the elements

The Bosnia v. Serbia judgment can be read as setting out, even if not in a fully systematic way, the elements that must be present before a third State’s failure to prevent genocide becomes internationally wrongful. Drawing these elements together, and treating them as the heart of the present analysis, produces roughly the following structure.

First, knowledge or constructive knowledge. The duty to act does not arise merely because a State is party to the Convention; it arises when a State “learns of, or should normally have learned of, the existence of a serious risk that genocide will be committed”.¹⁵ This introduces both an actual-knowledge and a constructive-knowledge limb. A State cannot escape responsibility by deliberately avoiding information that a reasonably attentive government would have obtained, but equally a State is not expected to have anticipated a risk that no reasonable observer, with access to the same information, could have foreseen.

Second, a serious risk of genocide. Knowledge alone is not enough; what must be known, or reasonably knowable, is a serious risk that genocide specifically, rather than some lesser form of violence or persecution, will occur. The Court did not lay down a checklist for what evidence satisfies this threshold, and this is, in the present author’s view, one of the most significant gaps in the existing law. Should the threshold be met by classified intelligence assessments, by formal warnings issued by United Nations bodies, by credible reporting from non-governmental organisations, by public statements of genocidal intent by political or military leaders, or by direct evidence of mass atrocities already underway? Each of these sources of information differs enormously in reliability and in how quickly it becomes available, and a standard that does not specify which of them is sufficient risks being either too easily satisfied or, more likely in practice, too easily denied.

Third, capacity to influence. Even where a State has the requisite knowledge of a serious risk, its obligation to act is calibrated to what it is actually capable of doing about it. The Court’s language, described above, ties the obligation to the State’s capacity to influence effectively the persons or groups likely to commit genocide.¹⁶ This has an intuitive appeal: it would be absurd to hold every State party to the Convention, including States with negligible connection to a given crisis, to an identical standard of expected action. But it also raises a genuinely difficult question, addressed further in Section VIII below, about how influence is to be measured and compared across States with very different kinds of leverage: military power, economic weight, diplomatic standing, or simply regional proximity.

Fourth, reasonably available measures. A State that has knowledge of a serious risk and the capacity to influence the situation is expected to use the means reasonably available to it. What counts as “reasonably available” is again left largely to the circumstances of each case, though the range of possible measures is not hard to imagine: diplomatic pressure, suspension of military cooperation, restrictions on arms transfers, economic sanctions, support for international investigations, referral of the situation to United Nations bodies, or the conditioning of ongoing assistance on changes in the recipient’s conduct.

Fifth, and running through the whole analysis, proportionality and causation. The Court’s insistence that the obligation is one of conduct, not result, means that a third State is not expected to take measures that are wildly disproportionate to its own interests or resources, but it does mean that the measures actually available and reasonably proportionate must be used. This raises a causation question that the Court has answered only in part: if a third State does nothing, and genocide occurs, does the applicant have to show that different conduct by the third State would probably have prevented or reduced the genocide, or is it enough to show that the third State failed to try measures that might have contributed to prevention? For the breach itself, the Court’s formulation, referring to measures that “might have contributed” to preventing genocide, together with its statement that it is irrelevant whether the State proves that all the means reasonably at its disposal would not have sufficed, points to the second answer.¹⁷ At the stage of reparation, however, the Court declined to award compensation because it could not conclude “with a sufficient degree of certainty” that the genocide at Srebrenica would have been averted had the Respondent complied with its obligation.¹⁸ The precise role that causation plays in establishing and remedying responsibility therefore remains open.

Taken together, these five elements suggest a workable, if still loosely specified, test: serious risk, combined with knowledge or constructive knowledge, combined with capacity to influence, combined with a failure to take reasonably available and proportionate measures, produces potential responsibility for breach of Article I. The remainder of the paper asks what additional legal architecture, in particular the doctrine of erga omnes partes obligations and the emerging case law on military assistance, adds to this basic structure, and where the structure itself still needs to be filled in.

Third-State standing and erga omnes partes obligations

A separate but related question is who may actually invoke a third State’s responsibility for failure to prevent. Historically, international law has been cautious about allowing States with no direct injury of their own to bring claims on behalf of the international community. The Genocide Convention has developed differently, in a way that reinforces the extraterritorial reading of Article I described above.

As early as its 1951 advisory opinion on Reservations to the Genocide Convention, the Court described the Convention as “manifestly adopted for a purely humanitarian and civilizing purpose”, one in which “the contracting States do not have any interests of their own; they merely have, one and all, a common interest, namely, the accomplishment of those high purposes which are the raison d’être of the convention”.¹⁹ This idea, that compliance with the Convention is a matter of common interest rather than a bundle of bilateral relationships, laid the groundwork for what the Court would later confirm explicitly.

In its 2020 order on provisional measures in The Gambia v. Myanmar, which found that The Gambia had prima facie standing, and then more fully in its 2022 judgment on preliminary objections, the Court held that the relevant obligations under the Genocide Convention are obligations erga omnes partes: each State party has an interest in compliance with them by every other State party, and may invoke another State party’s responsibility whether or not it can demonstrate a special interest of its own.²⁰ On this basis, the Court accepted that The Gambia, which had no direct connection to Myanmar or to the Rohingya population, could nonetheless invoke Myanmar’s responsibility for alleged breaches of the Convention, including the duty to prevent.

This development has two consequences for the present analysis. The first is procedural: it confirms that a State does not need to be the victim of genocide, or even a neighbouring or otherwise closely connected State, in order to bring a claim concerning another State’s failure to comply with the Convention. The second is more conceptual, and arguably more important for the third-State responsibility question addressed here. If every State party has a legal interest in the compliance of every other State party, it becomes easier to understand why the duty to prevent should not be read narrowly as applying only to the State on whose territory genocide occurs. The erga omnes partes character of the Convention and the extraterritorial reading of Article I reinforce one another: a Convention built around a genuinely collective interest in preventing genocide would be a strange vehicle for an obligation that applied only to the perpetrator State itself.

It is worth being precise about what The Gambia v. Myanmar does and does not establish. It establishes standing, the ability of a State with no direct injury to bring a claim. It does not, by itself, establish that any particular third State breached its own duty to prevent in that case; the merits of Myanmar’s own conduct, as the territorial State, are a different question from the one addressed here, which concerns the responsibility of States other than the one where genocide is alleged to have occurred. The Court heard the Parties on the merits in January 2026 and, at the time of writing, has not delivered its judgment.²¹ The importance of the case for present purposes is structural rather than substantive: it shows that the Convention as a whole is understood by the Court as creating obligations owed to the community of States parties, a reading that supports treating Article I’s duty to prevent as reaching beyond the territorial State.

Prevention, complicity, and aid or assistance: drawing the line

A third State’s relationship to genocide committed elsewhere can, in principle, take several distinct legal forms, and it is worth separating them clearly before turning to the specific problem of military assistance in Section VII.

A State’s own organs might themselves commit genocide, which engages that State’s direct responsibility under the Convention and under the ordinary rules on State responsibility for internationally wrongful acts. Alternatively, a State might not itself commit genocide but might knowingly provide assistance that facilitates another State’s or group’s genocidal conduct; this raises questions of complicity, and more generally of responsibility for aiding or assisting an internationally wrongful act, a concept recognised in the customary rules of State responsibility.²² Finally, and this is the scenario at the centre of this research, a State might neither commit genocide nor assist it, but might have knowledge of a serious risk of genocide, together with real influence over the situation, and simply fail to use the means available to it; this is a potential breach of the separate duty to prevent.

The Court in Bosnia v. Serbia was explicit that these are separate bases of responsibility. It held that Serbia had breached the distinct obligation to prevent even though the Court found that Serbia had not itself committed genocide, conspired to commit or incited genocide, or been complicit in genocide.²³ In other words, a State can be entirely cleared of direct participation and complicity, and still be found to have breached Article I simply by failing to try to stop a genocide it knew, or should have known, was seriously at risk of happening and over which it had real influence.

This distinction is conceptually tidy, but it becomes considerably harder to apply to a single, continuous course of conduct. Consider a State that continues to supply weapons to another State after it becomes aware of credible evidence of genocidal conduct by the recipient. Is this properly analysed as a failure to withdraw a “reasonably available measure” of prevention, namely suspending the arms supply, or is it better analysed as complicity or aid and assistance, on the theory that the continued supply itself materially facilitates the ongoing violence? The two framings are not mutually exclusive as a matter of law; a single course of conduct could in principle satisfy both. But they carry different evidentiary burdens. A complicity claim requires a positive act of aid or assistance given, at the least, with awareness that genocide was about to be committed or was under way, which is often a demanding standard to meet. A failure-to-prevent claim, by contrast, rests on omission and does not require proof that the State’s own conduct facilitated the genocide; it is enough that the State was aware, or should normally have been aware, of the serious danger of genocide and failed to use the available and proportionate means at its disposal.²⁴

The practical significance of this is that framing a case as a failure to prevent, rather than as complicity, may in some circumstances be the more promising route for an applicant State, precisely because it does not require establishing the same tight causal and mental link between the third State’s conduct and the harm. This is speculative, since the Court has not yet had to resolve a case squarely presenting both theories side by side on a comparable set of facts, but it is a genuine and, in this author’s view, underexplored feature of the current law, and one that the Nicaragua v. Germany proceedings, discussed next, may eventually help to clarify.

Contemporary application: military assistance and the Nicaragua v. Germany proceedings

The clearest contemporary illustration of the third-State responsibility problem is found in the proceedings Nicaragua instituted against Germany before the International Court of Justice in 2024. Nicaragua alleged, among other things, that Germany’s political, financial and military support to Israel engaged Germany’s own obligations relating to the prevention of genocide and to international humanitarian law more broadly.²⁵ By fifteen votes to one, the Court found that the circumstances, as they then presented themselves, were not such as to require the exercise of its power to indicate provisional measures, having taken note of Germany’s account of its export-licensing framework and of the significant decrease since November 2023 in the value of the material for which licences had been granted.²⁶ The Court nonetheless reminded all States of their international obligations relating to the transfer of arms to parties to an armed conflict, and stated that these obligations are incumbent upon Germany in its supply of arms to Israel.²⁷

The outcome on provisional measures should not be overstated; a decision not to indicate provisional measures is not a decision on the merits, and it reflects a demanding, largely procedural threshold rather than a considered view on whether Germany’s conduct did or did not breach Article I. The case has since moved to Germany’s preliminary objections, on which the Court held hearings in September 2026.²⁸ What is of more lasting interest for the present analysis is the declaration of Judge Cleveland, who, while joining the Court’s decision, wrote separately to emphasise that Article I of the Genocide Convention, like common Article 1 of the Geneva Conventions, imposes positive obligations on third States to prevent violations by other States and by non-State armed groups, and that in the context of military assistance these obligations require States to exercise due diligence in their arms exports.²⁹ Judge Cleveland’s declaration does not purport to resolve the case; it signals, from within the Court itself, that the theory underlying Nicaragua’s claim, namely that a State supplying weapons to another State can, in principle, incur responsibility under Article I for failing to prevent genocide, is not regarded as legally far-fetched.

Read alongside Bosnia v. Serbia and The Gambia v. Myanmar, the Nicaragua v. Germany proceedings can be understood as the point at which the abstract doctrine of extraterritorial, due-diligence-based prevention meets one of its hardest practical applications: an ongoing relationship of military assistance between a State that is not itself accused of committing genocide and a State that is the respondent in separate proceedings alleging it. This is a considerably harder case than Bosnia v. Serbia, where Serbia’s relationship to the Bosnian Serb forces was close, continuous, and involved a substantial degree of political and military coordination. An arms-supplying State’s relationship to the recipient’s conduct on the battlefield is typically more attenuated, and the argument that such a State has “capacity to influence effectively” the recipient’s conduct is correspondingly harder to establish on the facts of any given case, even if it is not impossible in principle.

A further contemporary illustration, referred to here only for its structural relevance and not as a comment on the merits of that dispute, is the set of proceedings brought by South Africa against Israel beginning in late 2023. By an order of 26 January 2024 the Court indicated that Israel shall, in accordance with its obligations under the Convention, take all measures within its power to prevent the commission of all acts within the scope of Article II of the Convention in relation to Palestinians in Gaza, and it reaffirmed and supplemented those measures by orders of 28 March and 24 May 2024.³⁰ That litigation concerns the obligations of the State on whose conduct the allegations are centred, and is therefore a different question from the one addressed in this analysis, which is concerned specifically with the obligations of States other than the one accused of the underlying conduct. It is mentioned here only because it demonstrates, alongside Nicaragua v. Germany, that Article I continues to generate active litigation, and that the third-State dimension of the prevention obligation is no longer a purely academic question.

The limits of the current framework: five sources of uncertainty

The preceding sections have shown that the ICJ has established the basic architecture of third-State responsibility for failure to prevent genocide: a distinct obligation of conduct, governed by due diligence, extraterritorial in scope, owed erga omnes partes, and separate from questions of complicity. What the case law has not done, largely because the Court has only had occasion to address these questions in a small number of cases with quite specific facts, is give this architecture the kind of precise content that would allow States, or future litigants, to know with confidence when the obligation is triggered and what it requires. Five gaps stand out in particular.

The first concerns the threshold of “serious risk”. The Court has repeatedly used this phrase without specifying what kind or quantity of evidence is sufficient to establish it. This matters enormously in practice, since a State’s political incentives to deny knowledge of a serious risk are often strongest precisely when the evidence is genuinely ambiguous, for example where reports of atrocities are contested, where intelligence assessments differ, or where a government’s public rhetoric is inflammatory but has not yet been matched by physical violence on a genocidal scale.

The second concerns the degree of influence or capacity required. The Court’s language, “capacity to influence effectively”, avoids the extremes of requiring either total control or accepting that trivial influence is enough, but it does not explain how a State’s diplomatic standing, economic weight, membership of international organisations, or military relationship with the relevant actors should be weighed against one another, nor how much influence, in absolute terms, is enough to trigger a legal duty to use it.

The third concerns the content of the measures a third State is actually expected to take. The Court has referred to the use of “all means reasonably available”,³¹ but has not indicated, for example, whether this can extend to measures that are economically or politically costly to the third State itself, such as sanctions against a major trading partner or an ally, or whether the obligation is effectively capped at low-cost measures such as diplomatic protest.

The fourth concerns causation. As noted in Section IV, the Court’s formulation, referring to measures that “might have contributed” to preventing genocide, suggests a relatively relaxed approach to causation at the level of breach, while its treatment of reparation in Bosnia v. Serbia shows that the question returns once compensation is claimed.³² The precise evidentiary burden on an applicant State to show that different conduct by the respondent would probably have made a difference has never been tested in a fully contested merits judgment against a third State that was not itself implicated in the underlying atrocities.

The fifth is more normative than doctrinal: the risk of selective enforcement. Because the obligation to prevent depends so heavily on a State’s own assessment of risk, knowledge, and available measures, there is an obvious danger that States will be quicker to identify a serious risk, and quicker to conclude that action is required, when the alleged perpetrator is a geopolitical rival, and correspondingly slower when the alleged perpetrator is an ally or an important economic partner. The law as it currently stands does not contain any mechanism for correcting this asymmetry, since the assessment of risk and the choice of measures remain, in the first instance, matters for each State’s own judgment, subject only to after-the-fact review by the Court in the relatively rare case that reaches it.

These five gaps are not a reason to abandon the extraterritorial reading of Article I established in Bosnia v. Serbia. They are, however, a strong argument for developing the due diligence standard into something more structured than a general instruction to act reasonably in the circumstances.

Toward a graduated due-diligence framework

This analysis proposes that the duty to prevent genocide is better understood, and would be more workable in practice, if it were treated as an escalating obligation rather than as a single duty that is simply present or absent. A graduated model would tie the intensity of what is expected of a third State to the seriousness of the risk and the extent of that State’s influence, along roughly the following lines.

At a first level, where the available information suggests only a low or speculative risk, a third State’s obligation would be limited to continued monitoring of the situation and ordinary diplomatic engagement, the kind of attentiveness that any government exercising due diligence over its foreign relations would already be expected to maintain.

At a second level, where the risk becomes serious and credible, supported by consistent reporting from multiple independent sources, the obligation would intensify to include more active diplomatic pressure, support for independent investigation, and, where relevant, humanitarian measures directed at the population at risk.

At a third level, where the risk is both high and imminent, and the third State has substantial influence, whether military, economic, or diplomatic, over the actors involved, the obligation would require the third State to consider suspending or conditioning any assistance it provides, and to pursue stronger economic and diplomatic measures, including through international bodies such as the UN Security Council or General Assembly.

At a fourth level, once genocide is actually occurring and the third State retains continuing capacity to influence the situation, the obligation would require it to take the maximum measures reasonably available to it, recognising that the passage from risk to actual atrocity does not extinguish the duty to prevent further harm, but rather intensifies it.

The value of organising the obligation this way is twofold. First, it gives concrete content to the abstract idea of due diligence by tying specific, foreseeable categories of measures to specific, foreseeable levels of risk and influence, rather than leaving States to guess, after the fact, whether whatever they did or did not do will be judged reasonable. Second, it preserves the flexibility that has allowed the obligation to apply to States in very different positions, since the graduated model does not specify a single fixed measure that every State must take, but rather a relationship between the seriousness of the situation, the State’s own capacity, and the intensity of the response expected of it. This is consistent with, and can be seen as an elaboration of, the Court’s own emphasis that due diligence must be assessed in concreto rather than by reference to a fixed rule.

Framed this way, the research hypothesis underlying this analysis can be stated as follows: the Genocide Convention establishes an extraterritorial duty of prevention applicable to States possessing the knowledge and capacity to influence a situation involving a serious risk of genocide, but the existing jurisprudence of the ICJ leaves insufficient guidance regarding the precise content of that obligation. A graduated due-diligence framework, calibrated to the level of risk, knowledge, capacity, and available measures, would provide considerably greater clarity while preserving the flexible, fact-sensitive character that has made the obligation workable across a wide range of situations.

Conclusion

The International Court of Justice has, across three cases spanning nearly two decades, built the basic architecture of a genuinely extraterritorial duty to prevent genocide. Bosnia v. Serbia established that the duty is distinct from the duty to punish, that it is an obligation of conduct governed by due diligence, and that it can require a State to use its influence over actors not under its own authority. The Gambia v. Myanmar confirmed that the Convention creates obligations owed by every State party to every other State party, reinforcing the idea that the duty to prevent is not confined to the State on whose territory genocide occurs. Nicaragua v. Germany, still pending, has brought the doctrine into contact with one of its hardest practical settings, the ongoing supply of military assistance by a third State, and has produced at least one judicial voice, in Judge Cleveland’s declaration, willing to treat the underlying theory as legally serious.

What remains missing is a clear, operational standard for when the duty is triggered and what it demands: how serious a risk must be, how much influence is enough, what measures a third State can be expected to adopt even at a cost to itself, how causation is to be assessed, and how the law might guard against the danger that States will apply the standard selectively, according to their own political interests rather than a consistent legal test. This analysis has argued that these gaps should be filled not by abandoning the flexible, fact-sensitive due diligence standard the Court has developed, but by organising it into a graduated framework built around knowledge, risk, capacity, and available measures, with the intensity of what is required of a third State rising as the risk becomes more serious and its own influence becomes greater. Such a framework would not make the law of genocide prevention mechanical or easy to apply. It would, however, provide a structure that States, applicants, and the Court itself could use with somewhat greater confidence than the present, still largely open-textured standard allows.

*****

Footnotes

1. Reservations to the Convention on the Prevention and Punishment of the Crime of Genocide, Advisory Opinion, 1951 I.C.J. 15, 23 (May 28) (describing genocide as a denial of the right of existence of entire human groups which “shocks the conscience of mankind”, and the Convention as “manifestly adopted for a purely humanitarian and civilizing purpose”).

2. Convention on the Prevention and Punishment of the Crime of Genocide art. I, Dec. 9, 1948, 78 U.N.T.S. 277 (entered into force Jan. 12, 1951).

3. Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Bosn. & Herz. v. Serb. & Montenegro), Judgment, 2007 I.C.J. 43 (Feb. 26) [hereinafter Bosnia v. Serbia].

4. Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Gam. v. Myan.), Preliminary Objections, Judgment, 2022 I.C.J. 477, ¶¶ 107–114 (July 22) [hereinafter Gambia v. Myanmar (Preliminary Objections)].

5. Alleged Breaches of Certain International Obligations in respect of the Occupied Palestinian Territory (Nicar. v. Ger.), Provisional Measures, Order, 2024 I.C.J. 560, ¶ 14 (Apr. 30) [hereinafter Nicaragua v. Germany].

6. Bosnia v. Serbia, supra note 3, ¶¶ 159, 162, 165, 425, 427.

7. Id. ¶¶ 182, 431.

8. Id. ¶¶ 430, 438.

9. Id. ¶ 430.

10. Id. ¶¶ 297, 386, 413, 415.

11. Id. ¶¶ 166, 430.

12. Id. ¶¶ 156, 183.

13. U.N. Office on Genocide Prevention & the Responsibility to Protect, The Genocide Convention (listing among States’ obligations the “[o]bligation to prevent genocide (Article I) which, according to the ICJ, has an extraterritorial scope”), https://web.archive.org/web/20190822222150/https://www.un.org/en/genocideprevention/genocide-convention.shtml (archived Aug. 22, 2019).

14. Bosnia v. Serbia, supra note 3, ¶¶ 430, 434.

15. Id. ¶ 431.

16. Id. ¶ 430.

17. Id. ¶ 430.

18. Id. ¶ 462.

19. Reservations to the Convention on the Prevention and Punishment of the Crime of Genocide, supra note 1, at 23.

20. Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Gam. v. Myan.), Provisional Measures, Order, 2020 I.C.J. 3, ¶¶ 41–42 (Jan. 23); Gambia v. Myanmar (Preliminary Objections), supra note 4, ¶¶ 107–108, 112, 114.

21. International Court of Justice, Press Release No. 2026/1 (Jan. 29, 2026) (announcing the conclusion of the public hearings on the merits in The Gambia v. Myanmar, held from 12 to 29 January 2026).

22. Bosnia v. Serbia, supra note 3, ¶ 420; G.A. Res. 56/83, annex, Responsibility of States for Internationally Wrongful Acts, art. 16 (Dec. 12, 2001).

23. Bosnia v. Serbia, supra note 3, ¶¶ 413–415 (attribution of genocide), 417 (conspiracy and incitement), 418–424 (complicity), 425–438 (duty to prevent), 471(2)–(5) (operative clause).

24. Id. ¶ 432; see also Nicaragua v. Germany, supra note 5, at 591–93, ¶¶ 3, 5 (declaration of Judge Cleveland).

25. Alleged Breaches of Certain International Obligations in respect of the Occupied Palestinian Territory (Nicar. v. Ger.), Application Instituting Proceedings Containing a Request for the Indication of Provisional Measures, ¶¶ 13, 56 (Mar. 1, 2024).

26. Nicaragua v. Germany, supra note 5, ¶¶ 17–18, 20, 25–26.

27. Id. ¶¶ 23–24.

28. International Court of Justice, Press Release No. 2026/23 (Sept. 10, 2026) (announcing the conclusion of the public hearings on the preliminary objections raised by Germany in Nicaragua v. Germany, held from 7 to 10 September 2026).

29. Nicaragua v. Germany, supra note 5, at 591–96, ¶¶ 2–5, 8, 16 (declaration of Judge Cleveland).

30. Application of the Convention on the Prevention and Punishment of the Crime of Genocide in the Gaza Strip (S. Afr. v. Isr.), Provisional Measures, Order, 2024 I.C.J. 3, ¶ 86(1) (Jan. 26); id., Request for the Modification of the Order Indicating Provisional Measures of 26 January 2024, Order, 2024 I.C.J. 513, ¶ 51 (Mar. 28); id., Request for the Modification of the Order of 28 March 2024, Order, 2024 I.C.J. 649, ¶ 57 (May 24).

31. Bosnia v. Serbia, supra note 3, ¶ 430.

32. Id. ¶¶ 430, 462.