One In, One Out

Twenty-page communiqués from United Nations Special Rapporteurs loom over the United Kingdom (“UK”) and France, respectively, as the two countries advance negotiations over their reciprocal “one-in, one-out” migrant transfer agreement. Under this treaty, individuals arriving in the UK through ‘irregular’ means from the territory of France, i.e., via small-boat crossings, may be returned to France. At the same time, the UK agrees to admit an equivalent number of migrants through authorised pathways.

The communiqués document human rights abuses, including torture, arduous travel, and dilapidated living conditions. In a particularly striking case, an individual fled Gaza, believing his family to be dead in the shelling, and subsequently made his way through France to the UK. Despite having received threats from smugglers that he would be killed if seen back in France, British authorities moved to return him, owing to his lack of legal representation or understanding of his rights. Only after he expressed an intention to end his life did the authorities rescind the readmission to France. The Rapporteurs raised concerns that the way the mechanism is applied undermines refugee protection, including guarantees of access to asylum, the prohibition on penalising refugees for irregular entry, and the principle of non-refoulement.

Such practice reflects a broader emerging pattern known as ‘migration externalisation’, wherein states enter into cooperative arrangements with neighbouring or transit states to manage asylum flows beyond their territorial borders. Contemporary arrangements such as the EU-Turkey Statement (which has a similar one-in, one-out policy), Italy’s cooperation with Libya involving Coast Guard pullbacks, and the Italy-Albania offshore processing protocol differ in design but retain migration control while shifting the locus of protection and the procedural burden beyond territorial borders. These treaties give rise to the “externalisation of customary obligations”, i.e., the transfer of the onus of compliance with customary international law to another state through a treaty mechanism, as highlighted by Gehad Madi, the Special Rapporteur on the Human Rights of Migrants, in his August 2025 Report.

Academic debate has largely approached externalisation through the lens of compliance, overlooking how states simultaneously affirm the validity of non-refoulement even as they reshape how those obligations operate. Using the UK–France arrangement as a focal point, this post examines whether such mechanisms constitute attempts to modify the customary principles of international refugee law or are simply breaches of existing obligations.

Non-Refoulement Related Rights

Prima facie, externalisation agreements do not violate the customary principle of non-refoulement as codified in Article 33(1) of the Refugee Convention. It prohibits expelling refugees to territories where their “life or freedom would be threatened on account of [their] race, religion, nationality, membership of a particular social group, or political opinion.” Arguably, the UK and France have, by virtue of state sovereignty, adopted a route to manage asylum influx, deeming each other safe harbours. While Article 33(1) establishes a substantive obligation, it does not textually specify whether the state of arrival must conduct the protection assessment itself, whether migrants must access an asylum procedure in the territory of arrival, or whether responsibility may be transferred to another state deemed safe under domestic law. This discussion is still evolving under international human rights law. The customary rule is also reflected in Article 3 of the European Convention on Human Rights (“ECHR”), which extends protection to all persons regardless of their migration status and admits no exception of national security, unlike Article 33(2) of the Refugee Convention. Therefore, the UK-France’s invocation of the ECHR imports substantive obligations that exceed those under the Refugee Convention itself.

States are manoeuvring these lacunae to recalibrate the scope of customary international law. Across instruments, the language is strikingly consistent: the EU-Turkey Statement commits to returns “in full accordance with EU and international law… in respect of the principle of non-refoulement” with applications “processed individually”; the Italy-Albania Protocol to management “in conformity with international law and European law” based on “full and inclusive application of the Geneva Convention”; and the UK-France Agreement to action “without prejudice to their rights and obligations under international law, including the Refugee Convention and the ECHR.” State practice here has been consistent in procedurally remodelling Convention obligations. Italy externalises processing and migration control at sea while exercising effective control over the very migrants it claims fall outside its jurisdiction, while destination states fund and train third-state actors like the Libyan Coast Guard to perform “pullbacks” before migrants reach waters where non-refoulement is enforceable. The Dublin III framework allocates responsibility based on entry criteria rather than protection needs, discharging non-refoulement obligations through procedural allocation rather than substantive protection.

The issue, therefore, is not the absence of a governing framework but that states are increasingly employing formal legal instruments to displace protection responsibilities. While reaffirming compliance in their treaty preambles, the states are employing mechanisms that are de facto incompatible with the principle of non-refoulement.

De Jure Modification or De Facto Violation

The central question is whether these treaties are modifying existing customary law. The North Sea Continental Shelf judgment posits that a treaty generates a new custom if it is of a “fundamentally norm-creating character,” receives “widespread and representative participation,” and is backed by “virtually uniform” state practice. If state practice, accompanied by legal justification, reshapes the content of a rule, what appears to be a deviation may instead reflect an emerging consensus about what the law permits. Similarly, if a representative group of states systematically adopts responsibility-shifting arrangements like the UK-France or EU-Turkey deals, it could be argued that the custom of individualised asylum processing is being de jure modified by a new norm of ‘safe third country’ automaticity.

A custom-based modification is particularly attractive to states because it bypasses the procedural requirements of Article 41 of the Vienna Convention on the Law of Treaties (‘VCLT’), which governs treaty modification. Such shifts are not conceptually implausible, as some of these expansions have later been codified as modified custom. E.g., Article 51 of the UN Charter has been argued to accommodate pre-emptive self-defence and the “unable or unwilling” test, thereby lowering the threshold for use of force while simultaneously seeking to de jure modify the Article 2(4) prohibition. This inquiry, therefore, cannot be dismissed at the outset.

Yet, the concept of customary law modification faces a threshold problem. Scharf argues that divergence in state practice signals the creation of a new customary rule rather than the modification of an existing one. The International Law Commission (‘ILC’), in Conclusion 11 of its Draft Conclusions on the Identification of Customary International Law, identifies only three routes for treaty-custom interaction: codifying pre-existing custom, crystallising emerging custom, or generating new custom through widespread subsequent practice accepted as law, notably omitting modification of an existing principle.

Where practice is both widespread and accompanied by opinio juris, an organic customary law modification remains possible, as illustrated by the post-World War II development of maritime delimitation. Here, the doctrine of “natural prolongation”, originally the governing principle of continental shelf entitlement, was progressively displaced through accumulated state practice. The United Nations Convention on the Law of the Sea ultimately codified a hybrid regime combining natural prolongation with a 200 nautical mile distance criterion. What had initially appeared as departures from the prior rule crystallised into modified custom once states came to treat the emergent practice as legally obligatory.

Deviatory State Practice

Whether the externalisation practice constitutes sufficient state practice and opinio juris to modify existing custom becomes the critical question. The International Court of Justice (“ICJ”) held in Nicaragua that

“[..] instances of State conduct inconsistent with a given rule should generally have been treated as breaches of that rule, not as indications of the recognition of a new rule.”

In this light, Article 1(3) of the UK-France Agreement makes the right to seek asylum in the UK conditional on the deportation of another person to France, requiring numbers to be “regularly balanced”, thus transforming refugees from “subjects of rights” under the Refugee Convention into accounting units. Under Article 4(1), France must accept readmission once seven criteria are met, without further formality, even where protection claims remain pending under Article 3. Article 4(6) then requires expeditious disposal of outstanding applications to enable transfer, risking the individualised scrutiny customary international law requires.

On state practice, externalisation arrangements vary considerably in design, are confined to a small group of states, and face sustained contestation from international organisations, courts, and human rights bodies. The African Commission on Human and Peoples’ Rights issued a 2025 resolution explicitly cautioning against this trend, urging states to protect migrant rights amid rising bilateral deals with the EU/US. State practice is therefore not uniform. Critically, opinio juris does not support modification as states are not asserting that non-refoulement is outdated or claiming a right to return refugees to persecution. They instead rely on stretching the “safe third country” doctrine and thinning the “individualised assessment” to bring de facto violations within the appearance of de jure compliance. What is at stake is not non-refoulement itself but the related procedural rights that give it practical effect.

Human rights bodies have interpreted non-refoulement progressively to prohibit transfers exposing individuals to torture, inhuman treatment, arbitrary deprivation of life, or enforced disappearance. The principle operates as a broader framework requiring procedures capable of preventing such outcomes, constraining the domestic designation of “safe third country” practices, and requiring that such designations be preceded by an individualised risk assessment. In M.S.S. v. Belgium and Greece, the European Court of Human Rights (“ECtHR”) held that states cannot rely on transfer mechanisms where systemic deficiencies exist in the receiving state’s procedures. In Hirsi Jamaa v. Italy, extraterritorial obligations were imposed on states where migrants remained under effective state control beyond their territorial borders. Similarly, the Inter-American Commission of Human Rights, in John Doe v. Canada, went further, ruling that Canada’s “direct back” policy violated asylum seekers’ rights by failing to conduct individualised risk assessments before returns to the US, rejecting blanket reliance on third-country safety under arrangements like the Safe Third Country Agreement.

Subsequent Practice-Based Interpretation

Given the failure to meet the modification threshold, the ILC suggests a presumption in favour of interpretation under VCLT Article 31(3)(b). i.e., subsequent state practice establishing the parties’ agreement on treaty meaning. Externalising states argue that their practice clarifies or narrows, rather than modifies, what “safe third states” requires, consistent with Conclusion 7(3) of the ILC’s 2018 Draft Conclusions on Subsequent Agreements and Subsequent Practice.

Article 33(1) of the Refugee Convention is silent on how a state satisfies itself that a receiving territory is safe. The “safe third country” doctrine is not in the Convention. Still, it is an interpretive elaboration developed through state practice from the 1980s and codified in the EU Asylum Procedures Directive. Originally, under the United Nations High Commissioner for Refugees’ guidance and ECtHR jurisprudence, classification as a “safe third country” required no risk of persecution in the third country, admission of the migrant and access to a fair asylum procedure, and no exposure to onward removal.

States are now progressively diluting what “safe” requires. The UK–France Agreement treats France as automatically safe without individual review of admission, fair process, or onward removal. The EU–Turkey Statement treated Turkey as categorically “safe” for Syrians without a case-by-case assessment. This reflects a broader shift where “safety” is defined unilaterally and capriciously to include any state that is formally a signatory to the Refugee Convention with a functioning asylum system, regardless of individual risk assessment. Similarly, individualised assessment, though not explicitly mentioned in Article 33, has been reduced in practice to cursory screenings, the presumption of safety, and accelerated procedures that do not meaningfully evaluate risks.

Four constraints on such an interpretative exercise exist: First, the Refugee Convention’s object and purpose, the protection from persecution or return to face such persecution, creates a material standard below which interpretation cannot descend. Second, subsequent practice that operates to the detriment of third-party beneficiaries (refugees) who have no role in generating it cannot constitute a legitimate interpretation, i.e., the Buga constraint. Third, subsequent practice cannot be used to interpret a convention in a manner that undermines its effective enjoyment, as affirmed in Golder v. United Kingdom. Fourth, Article 31(3)(b) requires practice of all parties, failing which, recourse to Article 32 supplementary means remains available, but its interpretive efficacy is limited. Therefore, the limited practice of externalising states currently constitutes de facto violations, though, given the evolutive nature of custom, this assessment may not hold indefinitely.

Conclusion

Externalisation practices are typically analysed through the lens of breach or doctrinal change. That framing is increasingly inadequate. What the preceding analysis demonstrates is not an attempt to displace non-refoulement as a rule, but a shift in how its satisfaction is structured. This shows that courts and scholars are identifying violations or tracing the formation of new norms but are less equipped to address regimes in which obligations are formally preserved while their performance is redistributed across jurisdictions and actors. If courts treat contemporary practices as permissible interpretations of non-refoulement, their scope may be narrowed without formal amendment. If not, their prevalence does not change their illegality and, simultaneously, raises issues of state responsibility and attribution. The distinction is critical: norm evolution calls for reinterpretation or reform, whereas systematic normative avoidance shifts the focus to enforcement, jurisdiction, attribution, and access to effective remedies. Whether the prohibition on indirect circumvention will be meaningfully enforced remains uncertain.