Safe Third Country (STC) agreements involve the transfer of protection-seekers from one State to another. They take different forms, including bilateral treaties (the UK-Rwanda Treaty), political agreements (the Italy-Albania MoU) and regional mechanisms (Dublin III). They are usually justified as a means of sharing responsibility for protecting refugees and 'managing' access to asylum determination processes, as is the case for the Canada-US Safe Third Country Agreement. In other contexts their explicit purpose is to deter and to punish.
At a bare minimum, the legality of such arrangements depends on refugee safety and the means by which this is guaranteed. The receiving State has to be safe in general and for the individual concerned. Even if safety is ensured, STC agreements offer States the means to offshore and offload their legal and ethical responsibilities to those in need of protection. Within the UK context, for instance, attempts to externalise refugee protection form part of what UNHCR has described as an 'asylum ban'.
Legal challenges, in both 'sending' and 'receiving' States, provides a means of scrutinising STC agreements. A court in Albania is reported to be hearing a case challenging that county's agreement with Italy on constitutional grounds. In the Canadian context, litigation is currently centring on whether or not the US is a safe country for women. In 2016, the Supreme Court of Papua New Guinea ruled that the detention of protection seekers on Manus Island was unconstitutional. Subsequent litigation in Australia failed, however, to challenge the legality of that country's agreement with Papua New Guinea. The scrutiny that some legal processes have been able to provide seem only, therefore, to confirm the 'near impossibility of using courts to challenge the legality of externalisation practices.' Recent developments in the UK seem to offer further support for this conclusion.
In R (on the application of AAA (Syria) and others) (2023) the UK's Supreme Court held that the Secretary of State's policy to remove protection seekers to Rwanda to have their claims determined there was unlawful. There are, the Supreme Court found, 'substantial grounds for believing that there is a real risk that asylum claims will not be determined properly, and that asylum seekers will in consequence be at risk of being returned directly or indirectly to their country of origin' (para 105). The British government responded to this decision with a Treaty and Bill that seek to legislate the fiction, or indeed, the falsehood, of Rwanda's safety. This move demonstrates the fragility of the rule of law, both domestically and internationally. Addressing the latter, this essay shifts the focus from domestic challenges to international ones, exploring whether STCs could be contested as 'forbidden treaties'.
Forbidden Treaties
Writing in 1937, Alfred von Verdross asserted that States' ability to make treaties was more limited than was generally assumed in either international law or diplomatic practice. The 'coming into existence' of a treaty, Verdross explained, depended on a number of norms or principles that existed independently of the will of the contracting parties. International law could not, relatedly, 'admit treaties between juridical subjects which are obviously in contradiction to the ethics of a certain community'. Such treaties were forbidden.
Verdross made a significant contribution to the development of international law, legal theory, and international institutions, his work on jus cogens was 'instrumental' to that concept's development. Notably, as a member of the International Law Commission, Verdross was able to give legal expression to his thinking on treaties and their relationship with jus cogens norms. Article 53 of the Vienna Convention on the Law of Treaties (adopted May 1969, the VCLT) thus states:
A treaty is void if, at the time of its conclusion, it conflicts with a peremptory norm of general international law. For the purposes of the present Convention, a peremptory norm of general international law is a norm accepted and recognized by the international community of States as a whole as a norm from which no derogation is permitted and which can be modified only by a subsequent norm of general international law having the same character.
Verdross' thinking and this provision of the VCLT raise a number of interesting questions that are of more than historical and theoretical interest. Within the context of refugee law and protection, one such set of questions concern the prohibition on refoulement and its status in international law. The next concern the potential for and impact of norm conflict posed by STCs that expose protection seekers to serious harm.
Non-refoulement and Jus Cogens
Verdross' own thinking on the kinds of norms that were jus cogens changed over the course of his career. The International Law Commission (ILC) has, in its work on the concept, adopted a series of draft conclusions on the identification of jus cogens norms and their consequences. In November 2023, the UN General Assembly 'took note' of these conclusions, observing that 'the subject of identification and legal consequences of peremptory norms of general international law (jus cogens) is of major importance in international relations' (UN Doc A/Res/78/109).
On identification, the ILC's non-exhaustive list of jus cogens norms includes prohibitions of aggression, genocide, crimes against humanity, slavery, torture, racial discrimination and apartheid. The right to self-determination and basic rules of humanitarian law are also listed (see Conclusion 23 and its Annex).
Refoulement is the return of someone, in any manner whatsoever, to a situation where their life or freedom may be threatened. It is prohibited by a range of multi-lateral treaties, including the Refugee Convention, and as a matter of customary international law. In his Fourth Report to the ILC, Professor Dire Tladi, then its Special Rapporteur on Peremptory Norms of General International Law (Jus Cogens) (and now judge of the International Court of Justice (ICJ)) concluded that:
The principle of non-refoulement is another principle of international law whose candidacy for peremptory status has ample support (para 131).
Tladi drew this conclusion from a wide-ranging examination of the literature and State practice
In its report on the Seventy-Third Session, the ILC thanked Tladi for his 'outstanding contribution' to the elaboration of the draft conclusions. Significantly, that report also reflects and endorses Tladi's thinking on non-refoulement, making a series of corroborating references to the scholarship, jurisprudence, and a UNHCR advisory opinion that informed his work.
Taking as our starting point the proposition that refoulement is prohibited as a jus cogens norm, what is the legal status of practices which may conflict with it? How might any such conflict be identified?
The Rwanda Agreement and its Aftermath as a Case Study
In R (on the application of AAA (Syria) and others) (2023) the UK's Supreme Court applied clear and uncontested legal principles to the undisputed, expert evidence of UNHCR. It then made a series of factual findings concerning status determination in Rwanda. It affirmed that safety was to be determined by Courts, not the executive, on the basis of the evidence before them.
The UK Government responded to this decision by signing a new treaty with Rwanda and proposing domestic legislation, the Safety of Rwanda (Asylum and Immigration) Bill, to implement it. The Treaty has the same general structure as the MoU that proceeded it, creating a 'mechanism' for the 'relocation' to Rwanda of protection-seekers whose claims are not being considered in the UK (article 2). The Treaty also contains some additions to the MoU that relate to status and monitoring.
The Bill requires decision makers to 'conclusively' treat Rwanda as a safe country (clause 2). It also seeks to dis-apply sections 2, 3, 6-9 of the Human Rights Act 1998 (clause 3) which relate to, amongst other things, the ability of Courts to interpret legislation compatibly with the European Convention on Human Rights (ECHR). It also precludes courts and tribunals having regard to interim measures indicated by European Court of Human Rights (ECtHR) and enables Ministers to decide whether or not to comply with them (clause 5). Clause 9 links the coming into force of the Bill with the entry into force of the Treaty. Interestingly, the Bill provides that those who are Rwandan will not be returned there (clause 7(2)).
That the Government has sought to 'use law to change fact' prompts concern for the rule of law in the UK, the separation of powers, and the values that underpin its 'collaborative constitution'.
A Two-Level Norm Conflict Analysis
Here, I take the UK's attempts to externalise as a case study to offer a two-level norm conflict analysis of an STC agreement.
The first structural level at which a norm conflict may occur is through the acceptance of international legal obligations, in combination with the repudiation of others, to circumvent the Supreme Court's decision. The Treaty is peppered with references to Rwanda's ability to meet its obligations 'in practice' and 'in fact' (see article 3(2)). Acknowledgement of the potential for there to be a gap between the assumption of an obligation and its performance is not, however, the same as performance. The replacement of the MoU with a Treaty, while potentially illustrative of a change in the contracting parties' approach, is not demonstrative of change in practice. What is changed, however, is the creation of a new legal mechanism through which removal may occur notwithstanding a lack of capacity, protection, and safety. A mechanism that, when the Treaty and Bill are evaluated together, also prevents individuals from challenging their deportation to and as we shall see, from Rwanda.
The second level at which norm-conflict may occur is through the absence or presence of particular provisions in the Treaty itself.
While the Agreement makes general references to 'international human rights law' and the Refugee Convention, it does not contain a 'saving clause' to ensure the primacy of the parties' obligations under these (see, for example, article 14 of the Palermo Protocol).
The Agreement also leaves undetermined the legal basis for deciding claims for complementary or other protection. Article 10(2) refers to 'humanitarian protection need', but does not appear to correlate with the UK's broader non-refoulement obligations. These include those that protect from 'flagrant breaches' of the ECHR or from violence against women pursuant to CEDAW.
Finally, space precludes a detailed examination of the Treaty's new, purported safeguards (for which see here). Two examples must be sufficient to reveal their inadequacy. First, article 10(3) prohibits 'relocated individuals' being removed from Rwanda (except to the UK) and requires parties to 'agree an effective system for ensuring that removal contrary to this obligation does not occur…'. The Agreement is, however, silent on what this system involves or when it would be implemented. The Government, when pushed, has also been unable to provide any detail on this point.
Second, article 15(9) refers to the creation of a process to enable 'relocated individuals' to lodge complaints to a Monitoring Committee. Again, however, there is no information on what this may entail. It is also difficult to see how any such process could resolve deficiencies within the parties' legal systems when it is subordinate to them (article 15(10). Centrally and as foreshadowed by above discussion, these so-called protections do not enable someone relocated to Rwanda to challenge any actual or constructive attempt to remove them from Rwanda. The protections provided by the Treaty against refoulement and other human rights violations are not, as required by the ECtHR, 'concrete and effective'. They are not even 'theoretical and illusory'.
The Consequences of Conflict
Echoing article 53 of the VCLT, the ILC's conclusion 10 states:
A treaty is void if, at the time of its conclusion, it conflicts with a peremptory norm of general international law (jus cogens). The provisions of such a treaty have no legal force.
The parties to a treaty that is void because it conflicts with a jus cogens norm have specific obligations, these include bringing '…their mutual relations into conformity with the peremptory norm of general international law…' (VCLT art 71(1)(b) and ILC conclusion 12).
As peremptory norms of international law create obligations that are owed to the international community as a whole (obligations erga omnes):
Any State is entitled to invoke the responsibility of another State for a breach of a peremptory norm of general international law (jus cogens), in accordance with the rules on the responsibility of States for internationally wrongful acts. (Conclusion 17(2)).
A party to a forbidden treaty, or another member of the international community may, therefore, invoke jus cogens as a ground for that treaty's invalidity. Such assertions may, however, have a destabilising effect on States' relationships and the international legal system more broadly. Consequently, before legal obligations may be terminated on the grounds of invalidity, certain procedures have to be followed. These 'guard rails' are set out in the ILC's draft conclusions, which themselves draw on the VCLT. At the risk of (over) simplification, they enable interpretation, require notification, and provide for the possibility of arbitration or litigation before the ICJ (conclusions 20 and 21).
Who Represents Refugees?
The invaliding effect of jus cogens on forbidden treaties has not been considered in the context of refugee protection. This essay has revealed the timeliness of such a consideration, in the UK and beyond. A norm-conflict analysis, like that started here, may offer an additional means of challenging STC practices which place refugees and others at risk of refoulement. They may also require the interpretation of STC Agreements in ways that ensure refugee protection (see ILC conclusion 20).
This essay also prompts reflection on the limits of international legal processes which, like those discussed here, require challenges to be brought by a member of the international community.
Who, in this community of States, is prepared to represent refugees? It is telling that no State has used the dispute mechanism provided by the Refugee Convention, which enables parties to refer disagreements on interpretation or application to the ICJ (article 38).
We may assume that the parties to any impugned STC agreement are unlikely to question its validity. Given the role STC practices play in maintaining racial borders and racialised inequalities within the refugee regime, challenges from (other) global north states are similarly improbable.
Returning to the UK, the International Agreements Committee's scrutiny of the UK-Rwanda Agreement prompted the House of Lords to vote to delay that Agreement's ratification. Arguments like those advanced here may, therefore, lend support to the political scrutiny of STC practices, in addition to offering new avenues for legal challenge. If States are to bring their relations into conformity with international law, they may need the assistance of their courts, their democratic institutions, and the international community as a whole.
Catherine would like to thank Anja Bossow for commissioning this essay and for her thoughtful editorial comments and suggestions.