The Fiction of Non-Entry Meets the Fiction of Remaining: AG Emiliou in Sedrata

 

 

Andreina De Leo*, Post-doctoral researcher, Maastricht University

 

* Funded by the European Union (ERC, SoftEn project, 101165167, PI: Lilian Tsourdi). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.

 

Photo credit: Fred Romero, via Wikimedia commons

 

Introduction

 

On 23 April 2026, Advocate General Nicholas Emiliou delivered his Opinion in Sedrata, one of two pending preliminary references concerning the Italy–Albania Protocol. The Opinion marks a significant development in the ongoing proceedings before the Court of Justice of the European Union, addressing for the first time the question of whether asylum and return procedures may be conducted in a third country while remaining fully under the jurisdiction of a Member State.

 

While affirming that EU law continues to apply wherever a Member State exercises jurisdiction, and that the relocation of procedures outside EU territory does not, in itself, remove them from the scope of the acquis, the Opinion does not exclude, in principle, the possibility of delocalising asylum and return procedures to a third country, provided that all guarantees under EU law are fully respected.

 

This post examines the key elements of the Opinion and highlights the tensions that arise from its underlying logic.

 

EU Law Beyond Territory: Applicability and Conditional Compatibility in the AG Opinion

 

A central feature of the Opinion is its unequivocal rejection of a strictly territorial understanding of EU migration law. The AG finds that both the Return Directive and the Asylum Procedures Directive apply directly to the situations at issue, not merely by virtue of their incorporation into national law, but as a matter of EU law itself. In his view, the transfer of individuals to centers located in a third country cannot have the effect of rendering those directives inapplicable or of prejudicing the application of common standards, insofar as the procedures clearly fall within the material scope of EU law (see, in particular, paras. 34 and 48).

 

More broadly, the Opinion endorses a jurisdiction-based understanding of the scope of EU law, echoing principles familiar from international law. Where a Member State exercises authority over individuals, it remains bound to ensure compliance with all applicable EU rules, irrespective of the geographical location in which those activities are carried out. In this respect, the Opinion directly contradicts the position advanced by the European Commission during the hearing, which emphasized the territorial anchoring of the asylum acquis and denied its direct applicability outside the Union. Instead, the Advocate General affirms that EU law does not “switch off” when procedures are relocated beyond EU borders (para. 47).

 

Having established that EU law applies, the Opinion adopts a relatively concise approach to compatibility. As regards the Return Directive, the AG considers that nothing in its provisions expressly precludes the operation of detention facilities outside the territory of a Member State. In his view, the Directive regulates the grounds, duration, and conditions of detention, but remains silent on the geographical location of such facilities. From this silence, he infers that EU law does not, in principle, prohibit transfers to centers situated in a third country. The Opinion recognizes that such a scenario was likely not envisaged by the EU legislature. On this basis, and in the absence of explicit regulation, Member States retain a margin of discretion as to the organization of detention, including its possible extraterritorial dimension (paras 58-59, and 64).

 

At the same time, the Opinion draws a crucial conceptual distinction: the transfer to Albania cannot be qualified as a “return” or a “removal” within the meaning of the Directive. It neither brings the return procedure to an end nor constitutes the enforcement of a return decision. This clarification is significant, as it confirms that the Italy–Albania model is conceptually and legally distinct from the idea of “return hubs” currently discussed in the context of the proposed EU Return Regulation. In that framework, the transfer of a third-country national to a third State, other than the country of origin or habitual residence, would amount to the execution of a return decision directed towards that State, a possibility which is not permitted under the existing EU law. By contrast, the transfer to Albania is characterized as an intermediate step within an ongoing procedure: individuals remain fully under the authority and jurisdiction of the Member State, and the return decision is neither executed nor exhausted (paras 60-62).

 

A similar reasoning is adopted in relation to the Asylum Procedures Directive and the Reception ConditionsDirective. The AG considers that Article 9(1) of the Procedures Directive – which enshrines the right to remain on the territory of Member States pending the first instance asylum decision – does not, in principle, preclude the transfer of asylum seekers to a third country, provided that applicants are able to exercise their procedural rights effectively, including appearing before the competent authorities and benefiting from the guarantees laid down in EU law. While acknowledging that a literal reading might suggest that applicants must be brought back to the territory of the Member State once they lodge an application, the AG rejects such an interpretation when the provision is read in context and in light of its purpose. In his view, the notion of “remaining in the Member State” must be understood by reference to the definition in Article 2(p) of the Directive, which already includes border areas and transit zones, thus pointing towards a functional rather than strictly territorial understanding (paras 93-97).

 

On this basis, and supported by a functional and systemic reading, he argues that Member States cannot be regarded as precluded, as a matter of principle, from organizing asylum procedures in facilities located outside their territory, provided that those facilities fall under their jurisdiction and that EU guarantees are fully ensured in practice. The core objective of Article 9(1) is, in his view, to prevent removal before a final decision on the application has been taken and to safeguard the effectiveness of the right to seek asylum, rather than to guarantee physical presence on the territory of the Member State in all circumstances. In situations where those guarantees are maintained and no risk of refoulement arises, the relocation of procedures does not, in his view, undermine the Directive’s purpose. Accordingly, Emiliou concludes that Article 9(1) does not, in principle, preclude national legislation such as that at issue in the main proceedings (paras. 102–103).

 

The Functional Turn: “Magically turning non-EU territory into EU territory via some mystical legal alchemy” (Peers, here)

 

As argued elsewhere (see here, with Celoria, and here with Celoria and Ferri), I concur that EU law does not apply as a matter of voluntary choice, but follows from the fact that the procedures fall within its material scope, thereby ensuring the need to preserve the primacy and uniform interpretation of EU law, and to prevent unilateral circumvention of the acquis. Likewise, the transfer to Albania does not amount to a “return” or a “removal” within the meaning of the Return Directive, nor can it be assimilated to the concept of “return hubs”, as it does not constitute the enforcement of a return decision. It is also plausible that the silence of the EU legislature reflects the fact that such configurations were not contemplated at the time of adoption. However, beyond these points, the reasoning of the AG appears unconvincing, and is affected by internal inconsistencies.

 

The Opinion relies on a broad and functional interpretation of spatial concepts within EU law. The AG adopts an expansive reading of notions such as “territory,” “border,” and “transit zones,” suggesting that their meaning should be determined in light of function rather than physical location. This approach makes it possible to treat facilities located outside the Union as functionally equivalent to border or transit zones situated within Member State territory.

 

This is not entirely new: for instance, in the recent Danané (discussed, in this blog here), the Court, on the basis of an Opinion by the same AG, accepted that procedures traditionally associated with entry control may operate through a legal fiction of non-entry and be carried out in facilities located inland where national law designates them as “border” or “transit” settings. In that sense, the underlying logic of functional interpretation of geographical concepts is not unprecedented.

 

However, those cases remain anchored, in practice, within the geographical space of the Union, understood in a legally fragmented but still territorially contained manner. What is distinctive in the present context is the further conceptual step that is being contemplated: the extension of this legal fiction beyond the geographical borders of the Union itself. This marks a qualitative shift, as the “border” is no longer merely redefined within the territory of the Member State, but effectively displaced outside it.

 

This development raises significant concerns. By effectively allowing Member States to shape the spatial reach of EU procedures, it risks extending derogatory regimes beyond their traditionally circumscribed scope and weakening the safeguards attached to them. Moreover, the Opinion itself appears to expose tensions regarding the internal coherence and normative limits of this reasoning.

As mentioned, the interpretative ambiguity stemming from the absence of an explicit prohibition on relocating EU asylum and return procedures outside the Union is addressed through an expansive reading of “territory” and of “border” or “transit zones,” extending these concepts to extraterritorial settings. In other words, in the absence of a clear treaty or legislative definition clarifying that “territory” refers to the geographical territory of the Union rather than what a Member State unilaterally designates as such, the question arises as to whether Member States retain any discretion to determine the spatial scope of EU asylum law in this field.

 

EU institutions, including the Commission during the hearing, have consistently rejected such an approach, clarifying that asylum claims cannot be lodged or processed outside the Union, as this would require an extraterritorial application of EU law considered neither legally feasible nor politically acceptable. At most, Member States may transfer individuals to third countries in compliance with the principle of non-refoulement, but without retaining responsibility for the examination of asylum applications after disembarkation. The European Parliament’s Legal Service has adopted a similar position, recognizing that jurisdiction may, in certain circumstances, extend extraterritorially, while stressing that this does not transform such locations into EU territory for the purposes of asylum law. The institutional position thus converges on a key distinction: while jurisdiction may extend beyond borders in limited situations, the notion of “territory” under EU asylum law remains autonomous and confined to the geographical territory of the Union (for more info, see De Leo & Celoria, p. 604).

 

The AG introduces, however, a preliminary contextual remark noting that the agreement underlying the Italian legislation concerns Albania, a geographically close State to Italy, separated only by the Adriatic Sea, which is also an ECHR Contracting Party and an EU candidate country engaged in accession negotiations. These elements are treated as relevant factors in assessing the lawfulness of the scheme under EU law (paras 68-69). This is where a significant internal incoherence emerges.

 

If “territory” under EU asylum and migration law is not an autonomous legal concept, meaning that Member States are free to define its spatial reach at their discretion, then the assessment of compatibility with EU law in abstracto cannot, without contradiction, be made to depend on contextual variables such as geographical proximity or an ex ante assumption of compliance with fundamental rights. The attempt to anchor the analysis in Albania’s proximity and its presumed rights compliance therefore sits uneasily with the jurisdictional reading endorsed elsewhere in the Opinion. It effectively introduces an ad hoc relational criterion into what is otherwise presented as a non-autonomous legal concept under EU law, thereby blurring the distinction between a principled determination of the spatial scope of EU migration and asylum law and a pragmatic, case-by-case assessment of legality. Either the asylum and migration acquis is territorially bound to the Union’s geographical space, including only its legally constructed exceptions such as borders and transit zones, or it is not territorially constrained at all. In the latter scenario, the question would no longer concern abstract spatial limits but the concrete application of EU guarantees in individual cases, leaving national courts to assess compliance with EU law obligations without any role for EU law in further delimiting the spatial reach of “territory” itself.

 

The result is a conceptual inconsistency: what is framed as a question of legal scope, i.e. whether “territory” under EU law is autonomous or subject to Member State discretion, is made dependent on operative factors based on contextual and geopolitical considerations. These considerations, however, have no clear legal basis in determining the applicability of the acquis and end up suggesting, implicitly, a degree of autonomy in the concept of territory that the initial premise simultaneously denies.

 

This tension is further reinforced by the Advocate General’s acknowledgment that, while compatibility in abstracto may be preserved, the extraterritorial location of detention facilities inevitably generates practical and structural difficulties in ensuring full compliance with EU guarantees, thereby highlighting how relocation outside the Union introduces an inherent and additional layer of complexity in the effective protection of rights that would not arise within the territorial space of the Union.

 

More concretely, these difficulties are linked to structural factors such as geographical distance, the need to cross an international border to access the centers, and dependence on both third-country authorities and those operating within the centers for the enjoyment of key guarantees, including family visits and effective communication with legal counsel. The AG emphasizes that this does not, as such, imply a violation of EU law, but rather that Member States may be required to adopt specific organizational and logistical measures to ensure compliance with EU standards (para 80).

 

The limits of this reasoning become particularly visible in the AG’s own discussion of the requirement of “immediate” release under the Return Directive. In assessing the meaning of this guarantee, the Opinion accepts that the term “immediately” cannot be interpreted in an overly rigid or literal manner, but nevertheless acknowledges that national authorities must be able to organize, within a short timeframe, the practical modalities necessary to give effect to a judicial order ordering release, including transport by air or sea where required (para. 82). Even accepting this flexible interpretation, however, the Albanian model reveals the structural impossibility of equivalence introduced by extraterritorialisation. In a purely domestic setting, a finding that detention is unlawful triggers immediate release within the legal and physical space of the Member State, without any further material impediment to the restoration of liberty. By contrast, in the Albanian setting, even where a judicial authority orders release, the individual remains subject to continued physical constraint until the completion of a cross-border transfer back to Italy, thereby introducing an additional and legally significant phase of deprivation of liberty which is absent in a territorial context and stems solely from the extraterritorial location of the detention facility. This example illustrates, more broadly, that the issue is not confined to a single instance of unequal treatment. Rather, it reveals a pattern of structural frictions affecting a range of guarantees, i.e. access to legal assistance, the possibility of family visits, and the effectiveness of judicial protection, which, taken together, seriously undermine the claim of functional equivalence.

 

Seen in this light, the reasoning concerning the conditions for compatibility appears particularly unconvincing. The difficulty is not simply that individual guarantees may, in specific cases, be ineffectively implemented. It is that the extraterritorial design itself generates structural constraints that systematically hinder their effective exercise. These are not incidental or remediable deficiencies, but structural features of a system that operates outside the territorial and institutional framework in which the EU acquis was conceived. The issue, therefore, is not whether compliance can be demonstrated case by case, but whether a model that introduces additional layers of dependency, delay, and fragmentation can, as such, meet the standard of effectiveness and uniform applicability required by EU law, irrespective of any ad hoc organizational measures adopted by Member States. As such, this interpretation departs not only from a literal reading of the relevant provisions, but also from their teleological interpretation, which is precisely to ensure the effective and practical enjoyment of the rights they enshrine.

 

Furthermore, accepting the relocation of procedures outside the Union leaves unresolved a set of central issues concerning the Dublin system and mutual trust (a dimension addressed in detail in our previous analysis here), which the Opinion largely sidelines, merely noting that its applicability follows from that of the Asylum Procedures Directive and that its provisions must be interpreted coherently with it (para. 49). These questions therefore remain insufficiently developed in the reasoning, plausibly because they were not directly raised by the referring court, but are likely to be clarified in pending related proceedings (Joined cases C-706/25 Comeri and C-707/25 Sidilli), which more broadly concern whether the Protocol may adversely affect the proper functioning of internal EU rules, in potential breach of the Union’s exclusive competence in the light of the ERTA doctrine.

 

Conclusion

 

Overall, the Opinion of Advocate General Emiliou in Sedrata adopts an ambivalent approach to the legality of the de facto extraterritorial application of EU asylum and migration law.

 

On the one hand, it clearly affirms that EU law continues to apply wherever Member States exercise jurisdiction, thereby ruling out any attempt to circumvent obligations through geographical relocation. At the same time, it preserves the compatibility of the Protocol with EU law through a functional redefinition of the notion of “territory”. On the other hand, the delocalization of asylum and return procedures to third countries is accepted only under strict conditions: full compliance with EU standards, coupled with an uncertain reliance on contextual factors such as geographical proximity and an ex ante assumption of adequate fundamental rights protection based on formal international commitments. Yet the AG himself acknowledges that ensuring such compliance is significantly more demanding in an extraterritorial setting, while also leaving open a degree of Member State discretion in shaping the spatial understanding of “territory” and “border”.

Taken together, this gives rise to a structural tension and an internal incoherence that ultimately weakens the persuasiveness of the reasoning. From this perspective, coherence and effectiveness can in my view only be preserved by recognising that “territory”, for the purposes of EU asylum and return law, is an autonomous concept confined to the geographical territory of the Union, including its legally constructed exceptions such as border areas and transit zones. This implies, in practical terms, that both the examination of asylum applications under ordinary and accelerated border procedures and the detention phase under the Return Directive, prior to the enforcement of removal to the country of origin or to a third different country (should the proposed Return Regulation be adopted), must take place within the Union’s geographical space, irrespective of any unilateral qualification by Member States.

This view is grounded in the structural logic of the asylum and return acquis prior to removal, whose safeguards are designed to operate within the Union’s territorial framework. It is within this framework that supervision, enforceability, and effective protection are meant to be ensured. Accordingly, such guarantees cannot be transposed outside the Union through a functional extension of jurisdiction without altering their operation and effectiveness, even where formal complianc appears to be ensured.

Ultimately, it is now for the Court of Justice to decide whether this logic is to prevail, or should it be rewritten.

European Union Product Liability Law and its (Uncertain) Future – Some Thoughts on LF v. Sanofi Pasteur case

Magdalena Tulibacka, Director of the Center of International and Comparative Law, Visiting Assistant Professor of Practice, Emory Law School 

Photo credit: Pexels, via Wikimedia commons

 

It is difficult to overstate the importance of the forthcoming judgment of the European Court of Justice (ECJ) in LF v. Sanofi Pasteur.¹ The judgment, if it follows the recent opinion of ECJ’s Advocate General (AG) Medina,² is likely to create a major shift in the system of liability established by the EU’s Product Liability Directive (PLD).³

Three questions referred to the ECJ by the Court of Appeal of Rouen arose in a case involving a vaccine produced by Sanofi Pasteur. The case concerns a type of a health injury referred to by the French court as a progressive condition.⁴ The claimant started experiencing pain and other health complaints after being vaccinated. Her condition kept getting progressively worse. Even accounting for the continued worsening of LF’s health, however, it is somewhat puzzling how long it took for the case to reach the courts.⁵ It appears that LF waited 11 years from the time of appearance of first symptoms and 7 years from the diagnosis of her condition to commence proceedings before a compensation scheme, which rejected the claim because of lack of causal link,⁶ and 3 further years to bring a lawsuit against Sanofi.⁷ As held by two French courts considering this case, following Article 1245-15 and 1245-16 of Code Civil (implementing the provisions of the PLD on limitation periods), LF’s actions were thus time-barred.⁸

The Cour de Cassation, however, decided to refer the case back to a Court of Appeal of Rouen to test some of the recent arguments that appeared in its own jurisprudence and in judgments of other courts against the application of limitation periods in other domestic regimes when progressive conditions are concerned.⁹

In order to enable LF’s case against Sanofi to proceed, the Court of Appeal asked first of all whether a potential victim of a defective product could bring a case against its manufacturers based on fault in lack of vigilance and lack of information about the risks of using the products. This is very difficult to reconcile with the PLD - a system that does not allow any other general product liability system functioning on the same basis to exist. The primary reason that this question arose in the case seems to be that French law has more generous time limits for general tort liability claims, and thus LF could still arguably bring her claim against Sanofi.

The two remaining questions relate to the system of liability established by PLD and implemented into French law, and specifically its limitation periods and their application in cases of progressive diseases. The questions can be summarized as follows:

Does the PLD’s 10-year long stop period comply with the right to an effective remedy as provided in Article 47 of the EU’s Charter of Fundamental Rights,¹⁰ when it applies to cases of progressive conditions?

Does the Directive allow for the 3-year limitation period to start, in cases of progressive injuries, when the injury or condition has stabilized, not when the claimant knew or should have known about it, as indicated in PLD?

These questions push the boundaries of what the system of the Directive established. With its strict liability for property damage, personal injury and death caused by defective products, the PLD introduced one of the world’s most influential and comprehensive product liability systems. The Directive aims to be a complete system of liability rules, carefully balancing the variety of interests involved: the interests of victims of defective products, the industry, and even society as a whole. Recently, the PLD experienced a major overhaul, and a new Directive was enacted, bringing the European product liability law into the digitalized, increasingly complex market reality.¹¹ The new rules are not yet in force and do not apply in the case at hand, but they do reflect the current consensus on the whole system. The changes related to limitation periods will be described below.

The Directive continues to rely on national laws of EU Member States for interpretation and application of major elements of liability, such as defect, damage, and causal link. It also co-exists with other national contractual and non-contractual liability systems as well as compensation schemes for redress of damage caused by defective products. It is at this juncture – the interaction with the national systems of laws and remedies – that many questions arise, this case being a good example.

In response to the French court’s questions, the AG suggests that the ECJ ought to conclude as follows:

-          Victims of defective products can bring actions against the producers of these products using national tort-based, fault-based liability, as long as the alleged fault consists of factors not exclusively related to defectiveness of the product (like failures in vigilance).¹²

-          The ten-year long stop introduced in the Directive is invalid in the light of Article 47 of the EU’s Charter of Fundamental Rights, ‘in so far as its application has the effect of extinguishing the right to claim compensation of injured persons suffering from a progressive disease who, according to medical evidence, due to the progressive nature of their medical condition, cannot fully evaluate the damage caused to them and have therefore been unable to initiate proceedings against the producer within that period, thereby depriving those persons of their right of access to a court’.¹³

‘In the situation of a progressive disease, the three-year limitation period established in that provision starts to run on the date of stabilisation of the damage, defined as the moment from which, according to medical evidence, the condition of the injured person is no longer evolving.’¹⁴

For some context: the French system continues to challenge the maximum harmonization the PLD was meant to ensure. As regards the implementation of the PLD, France was reprimanded by the ECJ for late and then improper implementation and needed to amend its law.¹⁵

Further, the French tort liability system, and to some extent even its contractual liability, historically presented a more attractive option for victims of defective products.16 The French product liability law was developed by courts in the overall consumer-friendly climate, where the key role of liability rules was for victims of defective products to be compensated, as mostly strict liability of manufacturers and suppliers.¹⁷ French courts were open to abandoning the requirement of proving fault, as well as to adopting presumptions of causation and defect in product liability cases – the approach that is not common among other EU Member States.¹⁸ As mentioned above, the general tort liability system in France provides more generous limitation periods as well. It is thus not difficult to understand why some victims may wish to resort to this general liability system instead of the one established by the implementation of the PLD.

 

Reflecting on question 1.

In reference to the interaction between the PLD and the national liability systems, the AG’s opinion reflects a unique conundrum. While the Directive was not meant to be the only system of liability where victims of defective products could recover compensation, it is the only system of objective (‘defect-based’)¹⁹ liability for such products. Within the scope of liability as set out by the Directive, its rules constitute the threshold and the ceiling. Article 13 PLD provides that the Directive does not affect any rights an injured person may have according to the rules of contractual and non-contractual liability or a special liability system. This provision was further elucidated in ECJ’s jurisprudence. In Gonzalez Sanchez the Court highlighted that liability systems based on other grounds: such as fault or a warranty in respect of hidden defects, could remain it operation.²⁰ In Commission v France, a general product liability system different from that provided by the Directive was held not permissible.²¹ According to Gonzalez Sanchez, as confirmed by the AG in the current case, the Directive is the exclusive source for cases where the liability follows a defective product (as defined by PLD) causing damage or injury a person.²² Fault is not relevant in this system. The notion of defect as provided by the Directive focuses on lack of safety.²³

The AG’s argument that resorting to general tort liability rules should be possible if it could be established that the defendant was at fault by failing in its vigilance duties, however, is not very convincing. If the essence of the defendant’s fault is lack of vigilance over the product and the alleged lack of reaction (perhaps by pulling the vaccine off the market), is not the lack of safety in the product the condition si ne qua non here? Safety and vigilance seem to be very closely tied in European law. Within the context of product liability, courts cannot deem that there was a failure in vigilance without also assessing whether there was a defect (lack of safety). The AG suggests that fault in this case consists of factors not exclusively related to defectiveness of the product.²⁴

But surely, monitoring of the product’s safety once it enters the market is a requirement that features heavily in European Union laws on product safety. Pharmaceutical products carry a uniquely stringent set of requirements as regards monitoring. Further, in the context of product liability, failure to warn of potential risks, rather than being some additional criterion or requirement, is a feature of defect. Even though the PLD does not recognize the US-style ‘boxing’ of types of defects (into manufacturing, failure to warn and design defects), those three types are widely recognized in literature and jurisprudence across the EU and beyond.

In my opinion, the AG’s recommendations in this matter do not comply with the text of the Directive and the nature and spirit of product liability law as established by it.

 

Reflecting on Questions 2 and 3:

Let us now move to the question of limitation periods under the Directive. Generally, legal systems distinguish two types of limitation periods:

-          What we can call ‘ordinary’ limitation periods - procedural in nature and often, in product liability cases, dependent on subjective discovery of the injury/damage and the person responsible or liable to redress it. This moment of subjective discovery is also when the cause of action accrues.

-          Long-stops (also referred to as preclusion, prescription, long-stop, or repose) - more substantive in nature (in some systems, such as the PLD, a long stop is referred to as a period after which the claimant’s rights are extinguished), and dependent on objective criteria, not on the subjective position/situation of the claimant.

Factors at play in the determination of these periods can be broadly classified into the following categories:

-          Substantive, constitutional and fundamental rights arguments: legal certainty and rule of law, finality, fair trial, access to justice and effective remedy,

-          Procedural and evidential arguments: effect of the passing of time on availability of evidence, and

-          Economic arguments – ‘closing the books’, calculation of risk and liability exposure, obtaining affordable insurance coverage.

In summary: determination of limitation and expiry periods is always a product of compromise between the plaintiff and the defendant interests, but also wider social interests in justice being done on the one hand and in encouraging progress and development of new products on the other hand. Such a compromise has been established in the PLD. It has remained in place till today and survived the comprehensive reform of the Directive in 2024, with some changes.

In the old PLD, claimants can bring product liability suits within 3 years from when they became aware, or should reasonably have become aware, of the damage, the defect and the identity of the producer. Further, the rights under the Directive expire 10 years from when the product was placed into circulation.²⁵ In the new PLD, the 3-year limitation period has been retained virtually unchanged. The 10-year long-stop is now called ‘expiry period’. It has also been retained largely unchanged (with the modification of the start of the period for substantially modified products). However, an exception was introduced where an injured person is not able to initiate proceedings within 10 years due to the latency of her injury. The period in such cases is extended to 25 years.

The French Court’s questions address both the limitation period and what is now called the expiry period. With regard to the latter, the question is whether its very existence is contrary to Article 47 of the EU Charter of Fundamental Rights as regards progressive conditions. The Charter, as part of the Treaties and thus primary EU law, can be the basis for a constitutional review by the ECJ of any legally binding EU measure. Any secondary EU law, including the PLD, which does not comply with its requirement that everyone whose rights guaranteed by EU law were violated should have access to an effective remedy (Article 47), can be declared at least partially void.

The opinion of the AG recommending exactly such an outcome should be looked at in the context of the views of the ECtHR. In its 2020 judgment in Sanofi v. France and in an earlier judgment in Howard Moor v Switzerland the ECtHR held that the long-stop violated the right to access to justice of victims who suffered from latent diseases.²⁶

The recent amendment of the PLD extending the expiry period to 25 years for latent conditions addresses these concerns, albeit unfortunately the new PLD does not provide a definition of latency leaving its determination to national law. It is possible that even this longer period may not always be sufficient, as certain products may well cause injuries and diseases with a much longer latency period than 25 years. Perhaps the period will need to be extended in the future for some latent conditions, but we are not ready to abandon it entirely. We are also not ready to leave the decision whether or not it applies to courts if the conditions are already present but, according to the claimant’s doctors, may not have stabilized. The long-stop (expiry period) constitutes one of the fundamental elements of the balance of consumer-business interests set out by the PLD. Any amendments to it will no doubt cause a push-back from the industry side and calls for rebalancing the whole system. Especially amendments that have their roots in undefined legal and medical concepts.

What is particularly concerning in the case at hand is that the AG is willing to allow the unique approach adopted by some French judiciary for progressive conditions to be used in cases based on the Product Liability Directive, potentially extending its effect across other EU Member States. The question of progressive injuries is itself problematic. Further, this approach focuses on an (as yet undefined) concept of ‘stabilization’.

I will now address both these concepts and their application to limitation periods.

Progressive conditions mean that the injury is known for some time, albeit not in its entirety. The French argument is that the claimant will not be able to assess the full extent of the damage while her condition keeps developing, and thus, no limitation period should be running until the claimant’s condition ‘stabilizes’.²⁷ The notion of a progressive condition or disease is not defined in the Directive. It is also unlikely that the ECJ would provide such a definition, as it is rather a medical and a case-specific term. Thus, the exact meaning would be determined by doctors within each EU state – and that’s what the AG seeks to avoid.

If one were to agree with the opinion of the AG addressing questions 2 and 3, it would mean that for progressive diseases the only limitation period is three years from the moment when the claimant’s doctors conclude that her condition has stabilized (as the ten-year period would be disapplied). The AG’s argument that an EU-level approach is needed in such cases and the response cannot be left to Member States, should be commended but what she suggests is, in my view, the wrong way to approach this.

It seems from the submissions by the German government in the case that the problems with the progressive diseases and the resulting limitations in access to remedy may well be a problem with the French system. While it is a fact that provisional damages are normally not available in civil law systems, such systems have other ways in which victims can bring suits before their injuries present themselves in their entirety (such as bringing declaratory actions). Suits can be brought while the claimant’s condition is developing, and courts can assess potential future damages as part of the compensation award. Additional suits can potentially be brought as follow-on actions when new conditions develop, or the existing condition worsens beyond what was initially predicted. The text of the AG’s opinion contains statements to the effect that French law, in contrast to other legal systems in the EU, does not allow claims for compensation of future damages or provisional damages. It seems to also indicate that the claimant may not be able to bring follow-on claims. If that is indeed the case, there is a significant systemic problem in French rules of damages. But a more likely answer is that, in France, there are options for assessing of future damages and taking them into account in the final determination of compensation amounts due to the claimant, and there are also possibilities for bringing follow-on claims.

The application and enforcement of substantive rights granted by EU law has always been dependent on national remedial and procedural rules, subject to the requirements of effectiveness and equivalence. Here this principle is applied backwards – the AG is allowing the perceived deficiencies in these national systems to cause the whole, carefully constructed compromise of PLD to be shaken. Would it not be a better idea for the ECJ to request that only French law is clarified in this respect, thus maintaining the coherence of the whole PLD system across the EU?

The French requirement that the limitation period does not start running until the claimant’s disease has ‘stabilized’, was already subject to a judgment by the ECtHR, where Sanofi challenged France alleging that Article 6.1 of the European Convention of Human Rights (right to fair trial) was violated. The ECtHR concluded that France was within the limits of the discretion (margin of appreciation) granted by the Convention when it allowed the limitation period to be approached in this way by its courts.²⁸ This judgment was issued in a different legal context – specifically one where the ECHR is known for granting the states-parties a certain margin of appreciation. Here, in the context of the EU PLD, the discretion is severely limited.

Keeping the start of the limitation period flexible and subject to the decision of the doctor and the court in each case to the extent suggested by the AG can potentially create uncertainty. It is problematic from the perspective of access to evidence, the question of who has the power to determine ‘stabilization’, and the rights of the defendants.²⁹ It can also lead to potentially undesirable consequences of depriving the victim of a defective product of access to a remedy. The way the start of the ordinary limitation periods is usually understood is that it means the cause of action accrued. If the cause of action does not accrue until the person’s condition ‘stabilizes’, could it not be argued that patients who do not wish to wait for this moment do not have a claim? Would this not deprive many victims of access to remedy for a potentially long time, and often for the rest of their life?

Another point worth noting is that if we allow victims of defective products to wait with bringing cases until their condition has stabilized, which can mean after their death, we may be leaving the defendants in a position where they are to expect lurking lawsuits, usually by families of people who died. Thus, if the ECJ were to adopt the AG’s suggestion in this respect, perhaps it would be advisable to introduce a notification requirement into the PLD for such situations, such as the one in the Payment Services Directive, recently interpreted by the ECJ.30 If we take the need for an effective remedy into account, it may be a more effective, and more proportionate, response to require the claimant to notify the defendant of their injury, even if the latter is still progressing. Such a step would provide the defendant with the knowledge of the claim and some level of clarity, as well as provide the claimant with the certainty that their claim will not be time barred. Further, the notification could provide an encouragement for the parties to settle.

 

Conclusions:

In spite of the recent comprehensive reform of the PLD, some very important parts of the European Union product liability system are by no means settled, thus creating uncertainty for potential plaintiffs and defendants. This comment cautions against following the AG’s opinion because, while it may assist the claimant (LF) in her case, it could create a destabilizing effect on the EU-wide product liability system. The AG’s recommendations in this case are a concern. In a reality where European legal systems - through their legislators and courts – continue to face difficult cases of injuries, they are bound to experiment, introduce new systems, new rules, and new ways of interpretation and application of the existing rules. Perhaps it is not realistic to expect that PLD will fully achieve its objective of harmonization, and PLD is playing ‘catch up’ with national systems – the latest reform is an example. But the system should remain stable, and such national experimentation can threaten this stability.

The arguments raised in this comment are particularly notable after the new EU Directive on Representative Actions came into force. Product liability cases can now be brought using a pan-EU representative procedure. With some major private international law issues still unresolved, there will be scope for forum shopping. If we allow national peculiarities to remain part of the product liability system to the extent suggested by the AG in LF v. Sanofi Pasteur, we increase the risk of forum shopping.

 

¹ Case C-338/24, LF v. Sanofi Pasteur SA, Request for a preliminary ruling from the Cour d’appel de Rouen (France) lodged on 7 Mary 2024, https://eur-lex.europa.eu/eli/C/2024/4716/oj/eng.

₂ Opinion of AG Medina delivered on 19 June 2025.

³ Council Directive 85/374/EEC on the approximation of the laws, regulations and administrative provisions of the Member States concerning liability for defective products, OJ L 210, 7.8.1985, pp. 29-33, recently replaced by Directive (EU) 2024/2854 of the European Parliament and of the Council of 23 October 2024 on liability for defective products and repealing Council Directive 85/374/EEC, OJ L 2024/2853, 18.11.2024; in the text of this comment: ‘PLD’ or ‘the Directive’ and ‘new PLD’ or ‘new Directive’.

⁴ No precise definition of such a progressive disease is provided, except for it being a condition that develops over a long period (para. 53 of the AG Opinion, see note 2).

⁵ LF received the Revaxis vaccine manufactured by Sanofi Pasteur in 2003. One explanation advanced in the case indicates that a residual amount of aluminum hydroxide (detected in LF’s body), used in certain vaccines, could have caused her condition.

⁶ The Board for Conciliation and Compensation for Medical Accidents is a French no-fault compensation scheme: Commission de Conciliation et d’Indemnisation des Accidents Medicaux (CCI). Her claim was rejected by CCI following an assessment by the expert appointed in the case who did not find a causal link.

⁷ LF’s condition was deemed stabilized 8 years after her diagnosis. In 2020 LF brought a lawsuit against Sanofi Pasteur, based on tort liability (fault liability - Article 1240 Code Civil), and strict product liability (Article 1245 Code Civil). The lawsuit was dismissed by the Court of Alençon and then by the Court of Appeal of Caen because LF’s claims were time-barred. Under the Code Civil, in the provisions implementing the PLD (Articles 1245-15 and 1245-16, implementing Articles 10 and 11 PLD) a claimant in a product liability case must bring a claim within three years from the date on which she was aware or ought to have been aware of the defect, the damage and the identity of the defendant. Further, in actions based on strict product liability there is a strict ten-year time limit (starting on the date when the product was placed on the market). The Cour de Cassation overturned the Court of Appeal’s decision and referred the case to the Court of Appeal of Rouen.  

⁸ As per Articles 10 and 11 of the current PLD, see below for analysis.

⁹ As per some previous judgments: Cass 1ère civ. 1 June 1999, B. 178; Cass. 2ème civ., 4 May 2000, no. 97-21.731; Cass 2ème civ. 11 July 2002, no. 01-02.182). The Cour de Cassation’s held that, in the event of an action for damages seeking compensation for bodily injury, the limitation period could only start running on the date when the damage has ‘stabilized’. Only then, according to the Court, the claimant would be able to assess the complete scope of her damage or injury.

¹⁰ Article 47 of the Charter provides for the right to an effective remedy: ‘Everyone whose rights and freedoms guaranteed by the Law of the Union are violated has the right to an effective remedy before a tribunal …’.

¹¹ See n. 3.

¹² C-338/24, AG Opinion, para. 43.

¹³ C-338/24, AG Opinion, para. 105.

¹⁴ C-338/24, AG Opinion, para. 124.

¹⁵ For instance, in Case C-52/00 Commission v. France, Judgment of 25 April 2002.

¹⁶ J.S. Borghetti, The development of product liability in France, in S. Whittaker (ed.) The development of product liability, Volume 1, Cambridge University Press, 2010, pp. 87-113., at p. 98.

¹⁷ Ibid. J.S. Borghetti, The development of product liability in France.

¹⁸ This approach was accepted by the ECJ as in line with the PLD in C-621/15 NW, LW, CW v. Sanofi Pasteur MSD SNC, judgment of 21 June 2017. Presumptions of defect and causation were also introduced in the new PLD: see Article 10.

¹⁹ H. Taschner, ‘Product liability: Basic problems in a comparative law perspective’, in Fairgrieve, D. (ed.), Product Liability in Comparative Perspective, Cambridge University Press, Cambridge, 2005, pp. 155 to 166, at p. 161.

²⁰ Case C-183/00 Gonzalez Sanchez v. Medicina Asturiana SA, Judgment of 25 April 2002, para. 31.

²¹ Case C-52/00 Commission v. France, Judgment of 25 April 2002.

²² AG Opinion, para. 36. See note 2.

²³ As defined in Article 6 of the old PLD.

²⁴ C-338/24, AG Opinion, (para. 48).

²⁵ Articles 10 and 11 old PLD, Articles 16 and 17 new PLD.

²⁶ ECtHR in Sanofi Pasteur v France, 2020. In Howard Moor, the court held that, ‘where it is scientifically proven that a person is unable to know that they are suffering from a certain illness, such circumstances should be taken into account when calculating the limitation period or statute of limitations.’ Elimination of a long-stop is not merely a European idea: for instance some U.S. states’ supreme courts struck down as unconstitutional the periods of repose in all product liability cases or only in cases of personal injuries. See: “50-State Survey of Statutes of Limitations and Repose in Prescription Product Liability Cases”, JD Supra, 2020.

²⁷ LF – the claimant in the French case – has indeed, according to her doctors, stabilized.

²⁸ This case involved the general tort liability system.

²⁹ For instance – it would be more difficult to assess liability risks for the purpose of obtaining insurance.

³⁰ Case C-665/23 IL v. Veracash SAS, Judgment of 1 August 2025.

 

 

 

Brexit means Brexit for UK nationals and EU citizenship: analysis of an Advocate-General’s opinion

 

Professor Steve Peers, University of Essex

Did British citizens retain their EU citizenship after Brexit? The EU and the UK government assume not, but some Brits have begged to differ.

The issue has reached the EU courts in several ways. Three cases went directly to the EU General Court, challenging the EU Council’s decision to conclude the withdrawal agreement, on the grounds that the agreement (in the applicants’ view) wrongly removed their EU citizenship. Last August, the EU General Court rejected all these cases for lack of standing; the applicants have all appealed to the CJEU. (See my compilation of Brexit litigation for further details).

Two cases reached the CJEU via national courts sending questions about the interpretation of EU law and the validity of the decision to conclude the withdrawal agreement. This route does not raise standing problems. The first of these cases (Case C-673/20 EP) has led to an opinion of an Advocate-General of the CJEU yesterday – which argued that British nationals have lost their EU citizenship as a result of Brexit, and the decision concluding the withdrawal agreement was valid.  

Summary of the opinion

The opinion begins by quoting Article 9 TEU: “Every national of a Member State shall be a citizen of the Union. Citizenship of the Union shall be additional to and not replace national citizenship” – and noting that Article 20(1) TFEU is nearly identical. It then summarises the specific issue in this case: the loss of local election voting rights of a UK citizen in France, due to her loss of the status of EU citizen (which entails such voting rights) as a consequence of Brexit. Interpreting these Articles in the context of the Treaties as a whole, the Advocate-General states that:

Article 9 TEU and Article 20(1) TFEU provide that Union citizens must be nationals of a Member State. Union citizenship is additional to, and does not replace, nationality conferred by the Member States. In that context, it is important to acknowledge that the Member States could have decided to pool their competences and to confer on the European Union the power to determine who is entitled to become a Union citizen. That explicit choice by the Member States not only renders the European Union powerless to create Union citizenship independently from nationality as conferred by the Member States, but also raises a constitutional barrier to such a power being implied under Union law. [para 22]

This is an answer to those who argue that the EU could confer EU citizenship as such on Brits, Russians or any other nationals of non-EU countries. It leads to his next point, dismissing the relevance of the previous CJEU judgments on the limits of the loss of EU citizenship which were referred to by the applicant (Rottmann, Tjebbes and Wiener Landesregierung – see discussion here) on the grounds that they all concern the loss of EU citizenship via means of loss of nationality of a Member State.

By comparison, the UK’s withdrawal from the EU took place on the basis of Article 50 TEU, which had the consequence of the UK ceasing to be a Member State. Article 50(3) TEU states explicitly that ‘the Treaties shall cease to apply’ to a State that leaves the EU when a withdrawal agreement enters into force, or ‘failing that’ two years after its notification to leave (unless that time frame is extended by mutual agreement). The opinion interprets this to mean that:

…In consequence, the United Kingdom no longer resolved, inter alia, to establish a citizenship common to that of the nationals of EU Member States or to create an ever closer union among the peoples of Europe. Since the existence of Union citizenship depends upon the acquisition and possession of the nationality of a Member State, and the United Kingdom voluntarily abandoned that status in the manner prescribed by Union law, British nationals ceased to be Union citizens. [para 28]

He then examined – and dismissed – the specific arguments made by the applicant. First, she argued that she retained her EU citizenship despite the UK leaving the EU. In his view, since the EU ‘has freely determined that acquisition of citizenship of a Member State is a condition precedent for the conferral of Union citizenship’, her argument that links to a State ground a claim to that State’s nationality fail. Equally, in his view, her argument that her links with France prevent deprivation of her EU citizenship must fail, as she could always claim French nationality (and therefore EU citizenship): “[w]ithout her possession of the nationality of a Member State, France cannot recognise EP as a Union citizen.”

Although EU citizenship, according to case law, aims to integrate EU citizens into the society of the host State, this did not mean, in the Advocate-General’s view, that the history of such integration in an individual case prevented the loss of EU citizenship:

…In addition to the objection that EP, like all other British nationals, ceased to meet the essential precondition for the enjoyment of the status and benefits of Union citizenship, namely Member State nationality, as a direct consequence of the United Kingdom’s sovereign decision to withdraw from the European Union, it may further be observed that the integration of third-country nationals into the societies of the Member States is not among the goals furthered by Union citizenship. [para 34]

A brief reference in the Rottmann opinion to the independence of EU citizenship from Member State nationality was not relevant either, as that opinion also made it clear that nationality of a Member State was a condition to have the status of EU citizen: “far from supporting the proposition advanced by EP, point 23 of the Opinion of Advocate General Poiares Maduro in Rottmann entirely undermines it” [footnote omitted].

…It describes the rationale for Union citizenship being contingent upon holding the nationality of a Member State as the Member States’ mutual commitment to construct a new form of civic and political allegiance on a European scale. By its sovereign decision to leave the European Union, the United Kingdom signalled its clear determination to repudiate that commitment. In the context of that act by a sovereign State, an individual cannot seek to rely upon his or her British nationality to assert a claim either to Union citizenship or to its benefits. [para 37]

A technical argument about the scope of the transition period in the withdrawal agreement (discussed here) failed too: the opinion concludes that only some EU law continued to apply to Member States as regards UK nationals during that period, and this did not include voting rights in local elections.

The opinion then returns to the argument that the judgments in Rottmann, Tjebbes and Wiener Landesregierung – which require that the loss of EU citizenship via the loss of Member State nationality require an individual decision subject to proportionality, legitimate expectations, and the right to be heard – apply to Brexit. Here, the Advocate-General reiterates more fully that the earlier rulings do not apply by analogy to a State withdrawing from the EU:  

…The principle of proportionality requires that a competent authority balance conflicting rights and norms before taking a decision affecting an individual. The circumstances of this case require no balancing by the deciding authority that would take account of EP’s personal circumstances. As a direct result of the sovereign decision of the United Kingdom to withdraw from the European Union, a person in EP’s position lost the right to vote and to stand as a candidate in municipal elections in France, her Member State of residence. It is that sovereign decision, not that of a Member State or any authority thereof, that caused EP to lose the benefit of those rights. [para 42]

The applicant could not compare herself to a stateless person, as she remains a national of the UK:

…She can address any issue that she may have concerning her status or rights as a British national to the United Kingdom authorities. France or the European Union are incapable of playing any role in such a dispute. [para 43]

The same applied to her argument based on legitimate expectations:

…Any breach of legitimate expectations that EP may wish to ventilate concerning her status as a Union citizen is to be addressed to the United Kingdom, which has withdrawn from the European Union, and not to either the French authorities or to the European Union. [para 44]

And also to the argument that ‘the French authorities stripped her of the right to vote and to stand as a candidate in municipal elections, thereby depriving her of the right to participate in the democratic process’:

…Any deprivation of her right to participate in the democratic process as a British national arises exclusively as a consequence of United Kingdom law. [para 45]

Nor could she rely upon the principle of non-discrimination on grounds of nationality, in Article 18 TFEU, because although that principle applied to UK citizens during the transition period in general, the specific issue of voting rights was excluded from it by the withdrawal agreement, and non-EU citizens are in a different position than EU citizens. However, it remains to Member States, if they so choose, to allow non-EU citizens to vote in local elections.

Finally, in light of his analysis, the Advocate-General answers the specific questions referred by the national court to the effect that:

-          due to the UK’s decision to leave the EU and the withdrawal agreement, UK nationals are no longer EU citizens; any legal questions arising from this are a matter for the UK, outside the CJEU’s jurisdiction;

-          various Treaty articles and the withdrawal agreement do not preserve EU citizenship rights for UK nationals, because the UK has ceased to be a Member State and ‘the exercise of rights conferred by Union law does not furnish any legal basis upon which an individual’s status as a Union citizen falls to be determined’; and

-          the decision concluding the withdrawal agreement is valid, because the loss of voting rights follows from the UK’s decision to leave the EU, considering that the EU (according to prior case law) has a wide discretion in external relations and did not exceed the margins of that discretion.

More fundamentally, as to whether the withdrawal agreement ‘infringes certain principles underlying EU identity and is disproportionate since it contains no exception to the rule that British nationals lose the rights attaching to Union citizenship’, the opinion argues:

…Since the United Kingdom’s sovereign choice to leave the European Union amounts to a rejection of the principles underlying the European Union, and the Withdrawal Agreement is an agreement between the European Union and the United Kingdom to facilitate the latter’s orderly withdrawal from the former, the European Union was in no position to insist that the United Kingdom fully adhere to any of the European Union’s founding principles. Nor could the European Union secure rights that, in any event, it was not bound to assert on behalf of persons who are nationals of a State that has left the European Union and who are therefore no longer Union citizens. Finally, since Union citizenship depends upon the possession of Member State citizenship, no response other than the exclusion of British nationals from the definition of Union citizens was possible whilst remaining within the scope of the Treaties. [para 75]

Comments

Advocate-General’s opinions are not binding on the CJEU judges, so this is not the final word on the legal issues yet. But let’s examine what the opinion would mean if the judges follow it.

The opinion is, first and foremost, a complete rejection of the argument that UK citizens retain EU citizenship after Brexit. It explicitly applies whether those UK citizens have exercised free movement rights or not; and although it is focussed on voting rights (the subject-matter of the case) there is no logical reason to limit its effect to voting rights only. So the free movement rights of EU citizenship have also been lost – after the end of the transition period. (The applicant’s technical arguments about whether the transition period exceptions applied to Member States – even if those arguments had been successful – would anyway be moot now, following the end of that period).

The Advocate-General mentions the prospect of the applicant obtaining French citizenship (therefore EU citizenship), but not all UK citizens would find that route so easy. However, this is not crucial to the main logic of his legal argument, which is simply that EU citizenship depends on holding the nationality of a Member State – and the UK is no longer a Member State.

Moreover, in the view of the Advocate-General, the responsibility for the UK no longer being a Member State is solely that of the UK, given that the UK took a unilateral decision to leave the EU. This is consistent with prior CJEU case law (see Wightman and Shindler, discussed here and here). It follows implicitly that this case was not about whether Brexit itself was legally valid, since that was a matter for UK law (and that issue was not even raised by the national court). This is apparently sufficient in itself to defeat the argument relating to legitimate expectations, without needing to examine whether the EU created such expectations.

The Advocate-General does not discuss the common line of argument that ‘there is no provision explicitly detailing what happens to EU citizenship when a Member State leaves the EU’, presumably because he considers his conclusion as to the requirement of a link to the nationality of a Member State sufficient. If necessary, though, the response to that argument is simply that (as the opinion points out), Article 50 states that the Treaties cease to apply to the withdrawing Member State; and the Treaties create the concept of EU citizenship.

As for the analogy which the applicant sought to draw between Brexit and the prior CJEU judgments on the loss of EU citizenship, the opinion convincingly argues that the prior judgments confirm the link between Member State nationality and EU citizenship, not weaken it. For if EU citizenship could exist independently of holding Member State nationality, on the basis that the person concerned had once held the nationality of a Member State, then those judgments would have reached their conclusions on an entirely different basis. The whole point of the prior judgments is that the loss of Member State nationality fell within the scope of EU law because it led to the loss of EU citizenship.

The applicant’s argument turns on an assumption that EU citizenship is a status linked to individuals, not a status linked to a state’s membership of the European Union. But while there could well be an argument that EU citizenship should be based on the former approach, the wording of the Treaties (as the opinion confirms) points inescapably to the latter. Member States simply did not choose to create a status of EU citizenship independently from that of nationality of a Member State.

A further argument, not discussed by the Advocate-General, is that a comparison with the means of acquisition of EU citizenship is unhelpful to the applicant too. Brexit is a collective loss of EU citizenship, not an individual loss; so the better analogy is not with the case law on individual loss, but with the mechanism of collective acquisition. And the means by which nationals of the last 16 Member States to join the EU collectively obtained EU citizenship was those States’ accession to the EU on the basis of Article 49 TEU. It logically follows that the collective loss of EU citizenship follows logically from the parallel process of leaving the EU set out in Article 50.  

Photo: Garry Knight, via Wikimedia commons

Barnard & Peers: chapter 26