Of Snowballs and Yoyos: Some Conceptual Reflections on the Autonomy of the EU Legal Order

Table of Contents: 1. Introduction. – 2. Snowballing: expanding and mushing together autonomy and other principles. – 3. Yoyoing: pivoting between ‘thick’ and ‘thin’ autonomy. – 4. Zooming in: different understandings of autonomy in action. – 4.1. Investment arbitration as the hotbed of recent EU autonomy debates. – 4.2. The EU’s accession to the ECHR – will the third time really be a charm? – 5. Conclusion.

Abstract: Despite its constitutional importance and prominent role in several high-profile decisions of the Court of Justice of the European Union (CJEU), the EU’s external autonomy remains one of the most elusive (structural) principles of EU law. The present contribution argues that this elusiveness can be at least to some extent explained by two conceptual tendencies which can be identified in the CJEU’s relevant case law: snowballing and yoyoing. The former concerns the gradual expansion of the principle over time and its mushing together with other key tenets of the EU constitutional framework. The latter concerns the pivoting between different levels of scrutiny, which can be referred to as a ‘thick’ and ‘thin’ autonomy. To get a better understanding, the contribution then looks at the CJEU’s arbitration-related case law, in particular Opinion 1/17, providing an in-depth analysis of how the tendencies identified have shaped the Court’s reasoning. Finally, the contribution turns to the revised agreement on the accession of the EU to the European Convention on Human Rights, arguing that some solutions may only pass muster if the Court will apply a narrow understanding of autonomy, which is all but certain.

Keywords: Achmea – EU accession to the ECHR – external dimension of autonomy – thick and thin autonomy – Komstroy – Opinion 1/17.

1.   Introduction

Autonomy is one of the key tenets of the European Union (EU) legal order. It is variously considered as a ‘constitutional’,[1] ‘structural’,[2] or even ‘existential’[3] principle of EU law, whose roots can be traced back to the early landmark judgments of the (now) Court of Justice of the European Union (CJEU) in Van Gend en Loos and Costa v ENEL.[4] Today, the focus is mostly on the external dimension of the autonomy of the EU legal order, i.e. vis-à-vis public international law.[5] This external dimension of the autonomy of the EU legal order has become a defining – and often constricting – factor in the legal interaction of the EU and its Member States with the wider world. In recent years, the CJEU has addressed autonomy concerns in several high-profile decisions, typically involving some form of international dispute settlement.[6] And yet, the autonomy of the EU legal order remains a ‘highly contested concept’.[7]

For some, it is ‘a very part of the DNA of that legal order’[8] and a key ‘means of protecting the core values and structural tenets on which the EU is founded’.[9] Others take a more critical stance, decrying autonomy as a self-serving tool of a court jealously guarding its own prerogatives,[10] or at least lamenting its persisting elusiveness.[11] Autonomy has been described as a ‘shapeshifter’,[12] remaining unpredictable and in flux.[13] To some extent, of course, this may be unavoidable. After all, autonomy is an unwritten principle that has been discovered and gradually developed by the Court. But the same is true of other core principles, such as primacy and direct effect, whose normative content nevertheless appears to be much less of an ‘enigma’.[14]

Hence, how can one explain the persisting elusiveness of the autonomy principle? In the present contribution, I argue that the elusiveness is to a large part owed to two conceptual tendencies in the Court’s relevant case law that I refer to as snowballing and yoyoing. Section 2 addresses the first tendency, ‘snowballing’, which concerns the gradual expansion of what EU external autonomy requires in the Court’s case law as well as the mushing together of those requirements with other key tenets of the EU constitutional framework into an all-encompassing argumentative amalgam.[15] Section 3 turns to the second tendency, ‘yoyoing’, arguing that in its relevant case law the Court pivots between different levels of scrutiny, which are referred to here as ‘thick’ and ‘thin’ (understandings of) autonomy. Section 4 takes a closer look at how these differences play out in practice. Section 4.1 revisits the Court’s investment arbitration-related case law. Section 4.2 assesses the revised draft agreement on the EU’s accession to the ECHR, identifying autonomy issues that may arise depending on the level of scrutiny the Court would apply. Section 5 concludes.

2.   Snowballing: expanding and mushing together autonomy with other principles

The external dimension of the autonomy of the EU legal order has been discovered and gradually developed by the CJEU over the course of several decades since the 1960s. Thus, the Court’s case law is inevitably key to understanding what autonomy entails.[16] A closer look suggests that over time a ‘minimum core’[17] has crystallized. For example, the Court has repeatedly taken issue with international dispute settlement mechanisms that could affect the allocation of powers between the EU and its Member States, that could interpret or apply EU law, or that would feature organic links with the CJEU. And yet, the essence of external autonomy cannot be captured by an – albeit useful[18] – ‘checklist’ of requirements found in the Court’s autonomy-related decisions.[19] This is because the process of the evolution of the EU’s external autonomy has neither been straightforward, nor is it yet complete. In particular, such a checklist-approach would not adequately reflect a tendency that can be observed in the Court’s case law and that the present contribution calls ‘snowballing’. ‘Snowballing’ comprises two aspects: the Court (1) keeps expanding existing criteria and requirements while adding new ones; and (2) mushes together those new elements with existing ones as well as key tenets of the EU constitutional framework into an all-encompassing argumentative amalgam.

At least to some extent, the first aspect is of course an unavoidable consequence of the gradual development of the autonomy principle in the Court’s case law. While the number of relevant decisions has remained relatively limited thus far, they often concerned hard cases raising fundamental and highly context-specific issues.[20] Moreover, the European Commission (EC) and Member States usually did not make the same mistakes more than once or twice. A case in point is the composition of international tribunals. In Opinion 1/76, which concerned the draft agreement on the establishment of a European laying-up fund for inland waterway vessels, the Court had taken issue with the proposal that some CJEU judges should also sit on the envisaged Fund Tribunal, since, in its view, such organic entanglements could affect the judges’ impartiality.[21] In Opinion 1/91, the CJEU emphatically rejected a similar arrangement, arguing that it would make it ‘very difficult, if not impossible’ for judges sitting on both courts ‘to tackle questions with completely open minds’.[22] This time, the Court’s argument hit home. The EC and Member States learnt their lesson and subsequently no longer tried to establish any organic links between the CJEU and external dispute settlement bodies. However, the way how to comply with the requirements formulated by the Court is not always so obvious.[23] In fact, whether and why a specific treaty arrangement is or is not compatible with the autonomy of the EU legal order can often only be fully understood with hindsight. Snowballing significantly complicates this trial-and-error way of learning: in almost each new case, the Court adds bits and pieces to established criteria, introduces new elements to the mix, or (re-)combines existing requirements in novel ways.[24] Consequently, treaty drafters more than once had to go back to the drawing board and figure out how to accommodate the CJEU’s new red lines.

The second aspect that characterizes the Court’s snowballing – the mushing together of these various elements as well as key tenets of the EU constitutional framework into an argumentative amalgam – is perhaps even more significant. Again, the fact that the Court piles various lines of reasoning and/or legal principles upon each other to justify a certain outcome is per se nothing new.[25] And yet, with the increasing reliance on autonomy in combination with other key tenets and more or less vague principles of EU law, its evolution has arguably taken on a completely different quality.[26] As Lenaerts et al recently noted, the CJEU has explained that autonomy ‘relates to the constitutional framework of the EU, the nature of EU law, the principle of mutual trust between the Member States, the system of fundamental rights protection provided for by the Charter, the substantive law of the EU that directly contributes to the implementation of the process of European integration, the principle of sincere cooperation, and the EU system of judicial protection’.[27] But how exactly is autonomy related to these very different aspects, and what are the normative consequences? So far, the CJEU’s explanations have remained quite vague and apodictic, as its recent observation that autonomy ‘resides in the fact that the Union possesses a constitutional framework that is unique to it’ may illustrate.[28] At the same time, commentators find it increasingly ‘hard to distinguish in earnest between the different requirements which the Court originally discerned’ as they ‘more or less melt into each other’.[29] In my view, this development has been exacerbated by the Court’s practice – starting with Opinion 2/13 – to explicitly or implicitly split its relevant reasoning into two distinct parts. The first part contains preliminary considerations, setting out key provisions and principles pertaining to constitutional foundations and the ‘idiosyncratic nature’[30] of the EU in abstract terms. The second part then contains the Court’s concrete legal assessment, without however, necessarily connecting all the dots and/or explaining how the abstract considerations relate to its concrete findings.

By charging autonomy with ever more constitutional gravitas, the Court seems to signal a sense of the end justifying the means, even where its legal reasoning may lack in nuance and/or doctrinal rigor.[31] For example, prior to the judgment in Achmea there had been much debate about the proper construction of Article 344 TFEU. Most commentators, arbitral tribunals, and even AG Wathelet, in his opinion in Achmea, had considered Article 344 TFEU to apply only to disputes between Member States. By contrast, the Court adopted a broader reading, covering also disputes between Member States and private parties based on inter se treaties concluded between Member States. Consequently, it found that Articles 267 and 344 TFEU ‘must be interpreted as precluding’ a provision such as the arbitration clause in the intra-EU BIT at issue.[32] Curiously, however, the judgment contains no – explicit – interpretation of Article 344 TFEU. The Court briefly mentions the provision in its preliminary considerations, but subsequently does not engage with its wording, its telos, or its normative links with Articles 4(3) and 19 TEU and 267 TFEU. In the concrete case, the Court brushing over these intricate doctrinal issues allowed critics to dismiss its decision as essentially ‘political’.[33] On a more general level, such a lack of transparent reasoning is an important factor contributing to the elusiveness and unpredictability of the autonomy principle. Snowballing, thus, not only expands the scope of the principle of autonomy but also creates room for ‘strategic ambiguity’,[34] which is closely connected to the second conceptual tendency the present contribution identifies, i.e. ‘yoyoing’.

3.   Yoyoing: pivoting between ‘thick’ and ‘thin’ autonomy

When carefully reading the Court’s relevant case law, one cannot help but get the distinct impression that ‘autonomy means different things in different contexts’.[35] More precisely, I will argue in this section that the Court applies the same criteria differently in different contexts, pivoting between more or less rigid levels of scrutiny – which I call ‘thick’ and ‘thin’ (understandings of) autonomy.

‘Thick’ autonomy involves a scrupulous assessment of any potential impacts of the ‘external’ legal system or dispute settlement mechanism on the autonomy of the EU legal order. It goes way beyond the question whether an external adjudicative body has jurisdiction to formally interpret or apply EU law, let alone whether EU institutions would be legally bound by such an interpretation.[36] Rather, a threat to the EU’s autonomy could also arise, for example, where a substantive overlap between provisions in an envisaged agreement and EU law entails the risk that an adjudicative body under such an agreement could effectively – or even only hypothetically – encroach on the uniform interpretation and application of EU law and/or the Court’s ‘exclusive jurisdiction’. Similarly, when applying a thick understanding of autonomy, the ‘very existence of […] a possibility’ that disputes between Member States or between Member States and the EU over matters of EU law involving an alleged violation of an international treaty (such as the ECHR) could be submitted to an external adjudicative body or court may be found to undermine Article 344 TFEU.[37] As is well known, the Court relied on this exacting standard in Opinion 2/13 and thereby, compared to its earlier case law, significantly ‘rais[ed] the bar of compliance’.[38]

A more recent instance reflecting a thick understanding of autonomy is the Court’s broad reading of what constitutes a dispute pertaining to the interpretation or application of EU law in Achmea. Even though the Court acknowledged that arbitral tribunals were only called upon to rule on possible infringements of the BIT, even the incidental interpretation or application of EU law – which was not ruled out by the BIT, but which would also clearly not be binding on the Court – sufficed to trigger autonomy concerns. Admittedly, in the specific case, the applicable law clause in the BIT clearly covered EU law.[39] This made it easy for the Court to assume at least a potential risk of repercussions for the uniform and consistent interpretation and application of EU law in the Member States.[40] In Komstroy, however, the Court subsequently expanded its reasoning to the ECT, despite its much narrower applicable law clause – and it did so on grounds that appear squarely at odds with Opinion 1/17 on CETA, a prime example of the Court applying a thin understanding of autonomy.[41]

Such ‘thin’ autonomy is characterized by a greater willingness on part of the Court to ‘overlook’ the practical implications of an external adjudicative body’s treatment of EU law, so long as it has no formal jurisdiction to interpret and apply EU law ‘as law’ or formally bind EU courts to its understanding thereof. Similarly, when the Court applies a thin understanding of autonomy, the mere possibility that Member States can avoid infringing its exclusive jurisdiction may be sufficient to comply with Article 344 TFEU.[42] Finally, when applying a thin understanding of autonomy, the Court appears more inclined to acknowledge practical necessities and go along with efforts made by treaty negotiators to pre-empt potential autonomy concerns. As just mentioned, a recent example of the Court applying a ‘thin’ understanding of autonomy is Opinion 1/17, which will be discussed in more detail in the next section. However, Opinion 1/17 was not the first or only instance where the Court took a more conciliatory approach.[43] Another example was Opinion 1/92, where it gave the green light for the revised system of legal supervision under the EEA Agreement.

One aspect of the CJEU’s reasoning illustrating its thin understanding of autonomy in Opinion 1/92 concerns its own role in the EEA context. In Opinion 1/91, the Court had taken issue, inter alia, with the envisaged possibility that EFTA country courts could be authorized to ask the Court to ‘express itself’ on a question of interpretation of provisions in the EEA agreement, which were substantially identical to provisions in the EU Treaties. The Court’s interpretation, however, would not have been binding on these courts – a fact the Court found to be an unacceptable encroachment on the nature of its function as conceived by the EU Treaties.[44] The revised EEA agreement instead provides that EFTA states may ‘allow a court or tribunal to ask the [CJEU] to decide on the interpretation of an EEA rule’.[45] This means that an interpretation rendered by the CJEU would now be binding on the referring EFTA country courts. In reality, of course, the Court still has no control over what EFTA country courts make of its interpretations. Thus, the fact that the formal recognition of the binding nature of its decisions was enough to alleviate its autonomy concerns (regarding this particular point), in my view, indicates a ‘thin’ understanding of autonomy.[46]

In light of the above, it becomes clear that yoyoing is an important factor contributing to the elusiveness of the EU’s external autonomy, because it makes it difficult to gauge in advance, which level of scrutiny the CJEU will apply in a concrete case. Of course, some expectations in terms of consistency and predictability may be exaggerated and overlook the different ways in which a wide range of heterogenous, complex treaty arrangements interact with various ‘essential characteristics’ of the EU and EU law. Indeed, it is submitted that a certain flexibility, and thus yoyoing between thick and thin autonomy, is neither surprising, nor per se objectionable. The Court could, however, do a better job in explaining why it chooses to apply a more or less exacting standard in a particular case. Put bluntly, what is problematic is not that the CJEU applies different standards, but the impression that it applies double standards.

4.   Zooming in: different understandings of autonomy in action

4.1.  Investment arbitration as the hotbed of recent EU autonomy debates

In recent years, extensive and controversial discussions about the autonomy of the EU legal order took place in the context of the Court’s investment arbitration related case law, in particular, its decisions in Achmea and Komstroy as well as its Opinion 1/17 on the investment arbitration mechanism in CETA. Moreover, the Court’s rulings had significant practical repercussions. Achmea eventually led to the termination of almost 200 intra-EU BITs;[47] Komstroy led to a series of actions by the EU and its Member States trying to put an end to intra-EU investment arbitration under the ECT;[48] by contrast, Opinion 1/17 gave the green light for the new Investment Court System (ICS), which has been inserted in several of the EU’s new generation trade agreements and which underlies the EU’s proposal for the establishment of a multilateral investment court.[49]

In the present context, Opinion 1/17 is of particular interest, since it brings together elements of both tendencies described above. In the run-up to the Court’s decision, considering the Court’s uncompromising approach in Achmea, many commentators had expressed doubt as to the compatibility of the CETA ICS mechanism with EU autonomy.[50] Yet, already the opening paragraph of the Court’s analysis left no doubt that the general gist of Opinion 1/17 would be quite different from Achmea. The Court recalled

‘that an international agreement providing for the creation of a court responsible for the interpretation of its provisions and whose decisions are binding on the European Union, is, in principle compatible with EU law’ and that ‘the competence of the European Union in the field of international relations and its capacity to conclude international agreements necessarily entail the power to submit to the decisions of a court that is created or designated by such agreements as regards the interpretation and application of their provisions’.[51]

Per se, this was nothing new. As is well known, the CJEU has repeated this point mantra-like since Opinion 1/91.[52] However, in Achmea, it had appeared almost as an afterthought at the very end of the Court’s reasoning. Putting the same statement at the top of the principles underpinning the Court’s autonomy assessment, from the outset signalled a markedly different approach.

4.1.1. Thin autonomy at its finest: EU law as fact

On substance, the Court found that the compatibility of the CETA ICS mechanism with the autonomy of the EU legal order depended on two conditions: First, that the envisaged tribunal would have no jurisdiction to interpret and apply rules of EU law other than provisions of the CETA itself. And second, that its powers would not otherwise have the effect of preventing the operation of the EU institutions in accordance with the EU constitutional framework. While the second point introduces a new substantive element of EU autonomy (i.e. an element of snowballing), the Court’s analysis on both points clearly reflects a thin understanding of autonomy.

As regards the first condition, the Court, for a number of reasons, was satisfied that the CETA tribunal would have no power to interpret or apply EU law other than the provisions of the CETA itself (on this specific aspect, see below 4.1.2). The CJEU, in particular, relied on Article 8.31.2 CETA, which had been included during the legal scrubbing with the clear aim of circumnavigating potential autonomy concerns. According to this provision, ‘in determining the consistency of a measure with this Agreement, the tribunal may consider, as appropriate, the domestic law of a Party as a matter of fact’ and in doing so ‘shall follow the prevailing opinion given to the domestic law by the courts or authorities of that Party’. Moreover, it provides – stating the obvious – that ‘any meaning given to domestic law by the Tribunal shall not be binding upon the courts or authorities of that Party’.[53] The Court accepted these statements at face value and – without probing whether Article 8.31.2 CETA would effectively prevent the CETA tribunal from interpreting or applying EU law – concluded that an examination of domestic law by the CETA tribunal ‘cannot be classified as equivalent to an interpretation’.[54]

It appears that the Court mainly came to this conclusion because it found that Article 8.31.2 would ‘unequivocally’ say so.[55] Indeed, the second sentence of Article 8.31.2 CETA, clarifies (‘For greater certainty’), that in determining the consistency of a measure with the CETA, the tribunal may consider domestic and EU law, but only as a matter of fact. But obviously, on a thicker understanding of autonomy, this ‘solution’ could have created new problems, not least because in practical terms it is all but clear that treating domestic law as fact ‘has made or would make much difference to [arbitral tribunals’] reasoning process’.[56] And how could it? Irrespective of what one calls the cognitive process, any tribunal that must consider domestic law has to determine its meaning and/or effect – be it as a matter of law or fact – and thus will invariably engage in some form of (de facto, indirect, etc) interpretation.[57]

Admittedly, the third sentence of Article 8.31.2 CETA to which the CJEU also refers, aims to minimize this dilemma by providing the CETA tribunal with strict guidance on how to establish the content of the relevant domestic law. More specifically, it provides that when taking domestic law into account, the tribunal ‘shall follow the prevailing interpretation given to the domestic law by the courts or authorities of that Party’.[58] In principle, this alignment should allow for a significant reduction of unavoidable de facto interpretations by the tribunal. However, it is still no panacea. An obvious limitation exists, for example, where the domestic courts have not addressed the issue at hand or no prevailing interpretation has crystallized yet.[59] A similar problem could arise in cases involving questions of EU law, where the domestic courts have not asked the CJEU for a preliminary ruling. In this regard, it should be kept in mind that only final courts in EU Member States have a duty to ask the CJEU for a preliminary ruling if they have doubts as to the correct interpretation of EU law. But investment arbitration under the CETA does not require the prior exhaustion of domestic remedies. Even if an investor initially pursues domestic remedies, it can submit a claim to arbitration at any time; all it has to do is withdraw or discontinue any existing proceedings before domestic courts (or waive the right to initiate such proceedings).[60] Thus, where a dispute, for example, concerns new regulatory measures or laws based on an EU directive, there may be no domestic rulings, let alone any ‘prevailing interpretation’, which the CETA tribunal could rely on.

It appears that the CETA parties anticipated as much and thus included a final proviso in the third sentence of Article 8.31.2 according to which ‘any meaning given to domestic law by the Tribunal shall not be binding upon the courts or authorities of that Party’. To be clear, from an international law perspective, this seems self-evident. Indeed, as Jarrod Hepburn emphasizes, ‘it has never been suggested that a tribunal’s finding on [the respondent state’s] law would bind respondent state courts, whether those findings are treated as fact or law’.[61] But even so, the proviso does not alter the fact that under CETA, the parties are obliged to comply with and enforce an award, regardless of whether they agree or disagree with a tribunal’s findings on domestic law, including EU law.[62]

In sum, the above assessment shows two things. For one, the CETA drafters carefully crafted the domestic law clause, with a view to avoiding autonomy concerns. For another, despite its careful crafting, the provision remains subject to certain limitations and cannot rule out all potential effects on the interpretation or application of EU law. The Court, however, did not probe any of the practical limitations of the domestic law clause. Instead, it limited itself to a rather atypical, text-bound ‘checkbox’ style assessment, reflecting a thin understanding of autonomy.

4.1.2. Blatant contradictions? The dual nature of international agreements in CETA vs Komstroy

A second aspect that deserves mention in this context concerns the Court’s conciliatory approach to accommodating the dual nature of the CETA as an international agreement, but also as an integral part of EU law. The Court pointed out that in the context of international agreements entered into by the EU, ‘the jurisdiction of the courts and tribunals specified in Article 19 TEU to interpret and apply those agreements does not take precedence over either the jurisdiction of the courts and tribunals of the non-Member States with which those agreements were concluded or that of the international courts or tribunals that are established by such agreements’.[63] More specifically, the Court found that it was

‘precisely because of the reciprocal nature of international agreements and the need to maintain the powers of the Union in international relations that it is open to the Union […] to enter into an agreement that confers on an international court or tribunal the jurisdiction to interpret that agreement without that court or tribunal being subject to the interpretations of that agreement given by the courts or tribunal of the Parties’.[64]

Leaving aside the peculiarity of the Court’s reasoning from an international law perspective, its focus on whether the CETA tribunal had ‘the power to interpret or apply provisions other than those of the CETA’[65] was key for finding that the ICS mechanism was compatible with EU autonomy.

Yet, only a few months later, the yoyo swung back again. In Komstroy the Court found that the rationale underlying its Achmea decision also applied to intra-EU investment arbitration under the ECT.[66] A basic precondition for any autonomy concerns to arise in the first place, however, was that EU law could be (potentially) applicable in an investment arbitration under the respective investment treaty. As discussed above, in Achmea the Court could rely on the broad applicable law clause of the BIT to jump this hurdle.[67] In the context of the ECT, however, the issue was less clear-cut. This is because the applicable law clause in Article 26(6) ECT – similarly to Article 8.31.1 CETA – provides that ‘[a] tribunal […] shall decide the issues in dispute in accordance with this Treaty and applicable rules and principles of international law’.[68]

To overcome the hurdle, AG Szpunar had suggested that in light of its dual nature, which the CJEU had confirmed in Achmea, EU law would constitute such ‘applicable rules and principles of international law’ and thus come within the scope of Article 26(6) ECT.[69] The problem with the AG’s reasoning was that it focused solely on one side of the equation, i.e. the nature of EU law, without engaging with the meaning of the phrase ‘applicable rules and principles of international law’ in the context of the specific provision of the ECT.[70] Perhaps this is why the CJEU took a different – but unfortunately no less convincing – approach: the Court argued that ‘the ECT itself is an act of EU law’, and consequently arbitral tribunals established under Article 26 ECT are inevitably ‘required to interpret, and even apply, EU law’.[71] As Jed Odermatt has pointed out, while ‘technically accurate from an EU law perspective’, this understanding conveniently overlooks that the ECT is also – one may add, primarily – a multilateral treaty under international law.[72] Even more strikingly, the Court’s reasoning regarding the ECT is squarely at odds with Opinion 1/17, where the fact that the CETA Tribunal had no jurisdiction to interpret or apply rules of EU law other than the provisions of the CETA, was key to finding that the ICS mechanism posed no threat to the autonomy of the EU legal order. The Court’s U-turn in Komstroy without any meaningful explanation is certainly puzzling. In my view, the Court should have made an effort to either distinguish the ECT from the CETA or explain why EU law other than the provisions of the ECT could, in its view, be applicable in ECT-based arbitrations. By contrast, the lack of any transparent reasoning on such a crucial point not only weakens the persuasiveness of the Court’s decision but also fuels the elusiveness of the autonomy principle.

4.1.3. Adding new substantive dimension: safeguarding the operation of the EU institutions in accordance with the EU constitutional framework

The second major autonomy issue in Opinion 1/17 concerned the question whether factual pressures on EU policy making resulting from investor-state dispute settlement under CETA could ‘[undermine] the capacity of the Union to operate autonomously within its unique constitutional framework’.[73] As the Court explained, the autonomy of the EU legal order would be adversely affected if arbitral tribunals

‘were to have jurisdiction to issue awards finding that the treatment of a Canadian investor is incompatible with the CETA because of the level of protection of a public interest established by the EU institutions, [since] this could create a situation where, in order to avoid being repeatedly compelled by the CETA Tribunal to pay damages to the claimant investor, the achievement of that level of protection needs to be abandoned by the Union’.[74]

This line of reasoning was remarkable, since it added a new substantive dimension to the autonomy of the EU legal order: the level of protection of public interests enshrined in EU legislation. Just as remarkable, however, is the Court’s subsequent explanation, why CETA would pose no such risk, which once again reveals a thin understanding of autonomy.

At the outset, the Court was satisfied that in light of the general exceptions clause in Article 28.3.2 CETA, tribunals would have ‘no jurisdiction to declare incompatible with the CETA the level of protection of a public interest established by the EU measures […] and, on that basis, to order the Union to pay damages’.[75] This formulation is curious. The general exceptions clause, which is modelled on Article XX of the GATT, allows for the potential justification of measures that would otherwise violate the agreement; but clearly does not limit tribunals’ jurisdiction to decide on the matter.[76] More importantly, the general exceptions clause expressly only applies to violations of the non-discrimination standards (national treatment, most-favoured-nation treatment),[77] but not the practically more relevant standards of fair and equitable treatment (FET) and the prohibition of direct or indirect expropriations without compensation enshrined in Section D of Chapter 8.[78] The Court recognized the limited scope of the general exceptions clause but found that as regards the latter obligations, Article 8.9.1 CETA ‘[i]n the same way’ states that the parties have a right to regulate to achieve legitimate policy objectives. Moreover, the Court cited Article 8.9.2 CETA, which provides for greater certainty that ‘the mere fact that a Party regulates, including through a modification to its laws, in a manner which negatively affects an investment or interferes with an investor's expectations, including its expectations of profits, does not amount to a breach of an obligation under [Section D]’. But contrary to what the Court seems to suggest (without further explanation), it is far from clear whether these provisions really create a safe harbour for any ‘legitimate’ public interest regulation.[79] Arbitral tribunals frequently recognize that sovereign states have a right to regulate their domestic affairs.[80] But this does not mean they give those states carte blanche to regulate as they see fit without having to compensate affected investors – an aspect the Court seems to have overlooked.[81]

Next the Court reiterated parts of Points 1(d) and 2 of the Joint Interpretative Instrument,[82] but once again failed to explain how, in its view, the parties’ shared understanding that CETA would not lower certain regulatory standards would be relevant in the context of investor-state arbitration.[83] And yet, without further ado, it concluded that in light of the provisions mentioned it was ‘apparent […] that the discretionary powers of the CETA Tribunal and Appellate Tribunal do not extend to permitting them to call into question the level of protection of public interest determined by the Union following a democratic process’.[84] Moreover, touching briefly on the (notoriously vague) substantive investment protection standards enshrined in CETA regarding indirect expropriations as well as fair and equitable treatment, the Court found that the textual clarifications contained in CETA would further ensure that ‘the required level of protection of a public interest, as established following a democratic process, is not subject to the jurisdiction conferred on the envisaged tribunals’.[85] Irrespective of whether one thinks that the Court overestimates the effectiveness of the textual clarifications, its reasoning in any case seems slightly beside the point. In investment arbitration, tribunals generally do not ‘call into question’ democratically determined levels of protection of public interests. Rather, tribunals (may) grant investors compensation for damages incurred due to breaches of their substantive rights – regardless of the branch of government involved, including damages resulting from democratically sanctioned policy changes in the public interest.[86] Nothing in CETA seems to fundamentally change this.

4.1.4. Key take-aways: pragmatism, but no paradigm shift

In sum, Opinion 1/17 shows two things. For one, that the CETA drafters had done their homework. They skilfully provided the necessary – and sufficient – formal reassurances to convince the Court that the investment arbitration mechanism in CETA would not pose a threat to the autonomy of the EU legal order. For another, that it takes two to tango. The Court’s atypically text-bound ‘check-box’ approach and the identified gaps in its reasoning clearly reflect a ‘thin’ understanding of autonomy. As such it attests more to the Court’s willingness to take the bait, than to the legal effectiveness of the precautions built into the CETA. The Court’s remarkable mention of the ‘need to maintain the powers of the Union in international relations’ only reinforces the impression of the Court’s pragmatism.[87] But while perhaps the most candid admission on part of the Court that political calculations may affect its legal assessment, the wider implications of Opinion 1/17 for the EU’s external autonomy are far from clear.[88] In the aftermath of Opinion 1/17, commentators have suggested that the Court tends to adopt a stricter understanding of autonomy in the context of agreements, which (also) apply between Member States (such as intra-EU BITs or the ECT) and not only vis-à-vis third countries (such as CETA).[89] While this seems generally plausible, it is important to note, however, that the Court did not simply dismiss the autonomy concerns raised against the ICS by distinguishing CETA from intra-EU BITs and the ECT based on mutual trust considerations. What this means for the hundreds of extra-EU BITs which Member States have concluded with third countries, and which contain no safeguards comparable to CETA, currently remains an open question.[90]

On a more general level, nothing indicates that Opinion 1/17 would mark a broader paradigm shift towards a more lenient approach to EU external autonomy. As shown above, the Opinion 1/17 was not the first instance where the Court applied a thin understanding of autonomy; and soon after it pivoted back to a thicker understanding in Komstroy. Moreover, the extra-legal considerations that appear to have had an impact on the Court’s light-touch assessment of EU autonomy concerns in the CETA-context, related specifically to the EU’s role as an actor in the arena of international trade and investment law. Thus, Opinion 1/17 tells us little about which legal safeguards or political necessities (perceived or real) would prompt the Court to adopt a similarly pragmatic approach in other contexts such as migration, environmental policy or human rights protection.[91] And yet, using the example of the EU’s renewed attempt to accede to the ECHR, the next section argues that much will depend on whether the Court applies a thin or thick understanding of autonomy.

4.2.  The EU’s accession to the ECHR: will the third time really be a charm?

Prior to Achmea, the most hotly debated decision on the autonomy of the EU legal order had been Opinion 2/13, in which the Court infamously found the then envisaged agreement on the EU’s accession to the ECHR to be incompatible with the EU Treaties.[92] After this ‘bombshell’,[93] negotiations on the EU’s accession, as mandated by Article 6(2) TEU, resumed only in 2020 and were concluded on a technical level in March 2023, still leaving the intricate issue of alleged human rights violations in the context of the CFSP to be resolved internally by the EU.[94] Eventually, in July 2025, the Commission decided to request an opinion of the Court regarding the compatibility of the revised draft accession agreement (revised DAA) with the EU Treaties pursuant to Article 218(11) TFEU.[95]

Thus, the question arises whether on the third attempt the Court will finally give the green light for the EU’s accession to the ECHR. The following subsections outline major changes in the revised DAA which seek to respond to the Court’s main concerns in Opinion 2/13 and identify potential shortcomings. The aim of the present contribution is obviously not to try and predict how the Court will ultimately decide. Yet, it argues that whether some changes envisaged in the revised DAA will be successful likely depends on whether the Court will adopt a thick or thin understanding of autonomy.[96]

4.2.1. Changes in the revised DAA

Specific characteristics of the EU legal order: The CJEU’s first major concern in Opinion 2/13 was that the draft accession agreement did not sufficiently take into account the ‘specific characteristics’ of the EU legal order. More specifically, the Court raised issues pertaining to the lack of coordination between Article 53 ECHR and Article 53 CFR; mutual trust; and the possible repercussions of advisory opinions under Protocol 16 to the ECHR for the preliminary reference procedure under Article 267 TFEU.[97]

The Court’s concern over the coordination of Article 53 ECHR and Article 53 CFR has been addressed in Article 1(9) of the revised DAA, which provides that Article 53 ECHR ‘shall not be construed as precluding High Contracting Parties from jointly applying a legally binding common level of protection of human rights and fundamental freedoms, provided that it does not fall short of the level of protection guaranteed by the Convention and, as relevant, its Protocols, as interpreted by the European Court of Human Rights’. This provision does two things: it underlines that the level of protection guaranteed by the ECHR as interpreted by the ECtHR sets the common minimum standard for all contracting parties, but at the same time signals a ‘hands-off approach’ when some parties wish to go beyond this minimum level. To avoid autonomy concerns, the revised DAA plays the ball back into the EU’s field, where differences about higher levels of rights protection should be resolved.

As regards the issue of mutual trust, Article 6 of the revised DAA provides that ‘[a]ccession of the European Union to the Convention shall not affect the application of the principle of mutual trust within the European Union. In this context, the protection of human rights guaranteed by the Convention shall be ensured’.[98] In my view, this ‘solution’ is one of the most problematic aspects of the revised DAA. As Daniel Halberstam has argued convincingly, mutual trust had been one of the CJEU’s ‘biggest concerns’ in Opinion 2/13, due to ‘a very practical tension between the case law of the [ECtHR] and the [CJEU], especially in matters of asylum and family law’ and reflecting a ‘profound clash between the [CJEU’s] constitutional and the [ECtHR’s] intergovernmental vision of the Union’.[99] While the case law of both courts may now be more converging than it was ten years ago, Article 6 of the revised DAA does not appear to substantially address the underlying systemic concerns. Rather, the provision attempts to formally ‘square the circle’, by recognizing the principle of mutual trust, while emphasizing at the same time that the protection of human rights guaranteed by the ECHR must be ensured.[100] How this would work in practice, however, is far from clear.[101] Thus, the drafters seem to bank on the Court applying a ‘thin’ understanding of autonomy, which focuses more on formal assurances than practical effects, when scrutinizing the revised draft accession agreement. However, this is all but certain and therefore a risky bet.

Finally, the CJEU’s concern that the possibility of the highest Member State courts to request an advisory opinion of the ECtHR on the interpretation or application of the ECHR would undermine the autonomy and effectiveness of the preliminary reference procedure is addressed in Article 5 of the revised DAA. According to this provision, where questions relating to the interpretation or application of the ECHR fall within the scope of application of EU law, the relevant Member State courts ‘shall not be considered as a highest court or tribunal of a High Contracting Party for the purposes of [Article 1(1) of Protocol No 16]’. In other words, Article 5 of the revised DAA prevents such courts from requesting an advisory opinion.[102] Prima facie, preventing these Member State courts from requesting advisory opinions pursuant to Protocol 16 should suffice to allay the CJEU’s concerns over the autonomy and effectiveness of the preliminary reference procedure.[103] However, this solution raises a new autonomy issue: who gets to decide whether a question put forward by a highest domestic court falls within the field of application of EU law. If push comes to shove, i.e. if a qualified domestic court requests an advisory opinion, by the looks of it, this could only be the ECtHR; however, in light of the CJEUs case law, giving the Strasbourg court the power to determine the scope of EU law – even if only for the purpose of Protocol 16 – could very well clash with its claim to exclusive jurisdiction over the definitive interpretation of EU law.[104]

Inter-state cases: The second major issue identified in Opinion 2/13 concerned the possibility of inter-party cases between EU Member States and between the EU and its Member States, which the Court found to infringe Article 344 TFEU. This issue is addressed in two new paragraphs that have been added to Article 4 of the revised DAA. According to the new Article 4(3), the EU and its Member States ‘in their relations with each other shall not avail themselves of Article 33 [ECHR]. Nor shall the member States of the European Union avail themselves of Article 33 [ECHR] insofar as a dispute between them concerns the interpretation or application of [EU] law’. Moreover, to avoid the suspicion that the ECtHR would determine on its own whether and to what extent an inter-party dispute between EU Member States concerns the interpretation or application of EU law, Article 4(4) of the revised DAA stipulates that the ECtHR ‘shall provide the European Union upon request with sufficient time to assess, as a matter of priority’ exactly this question. It should not be overlooked that Article 4(4) does not explicitly address the effect of the EU’s assessment in the proceedings before the ECtHR. But unlike the solution provided in Article 5 of the revised draft regarding Protocol 16, Article 4(4) at least recognizes that the EU should play some role in determining whether a dispute falls within the scope of EU law.

Co-respondent mechanism: The CJEU’s third major concern in Opinion 2/13 pertained to the so-called co-respondent mechanism. Article 3(5) of the revised DAA provides that the EU or its Member States may become a co-respondent, ‘either by accepting an invitation from the Court or upon their own initiative’ and that the ECtHR ‘shall admit a co-respondent by decision if a reasoned assessment by the European Union sets out that the conditions justifying the application of the co-respondent mechanism set out in paragraphs 2 or 3 of this article are met’.[105] Importantly, the revised DAA does away with the previously envisaged plausibility test, which the CJEU had found to interfere with the division of powers between the EU and its Member States.[106] Otherwise, however, Article 3 of the revised DAA prima facie seems to do little more than sugar-coat the fact that it is ultimately the ECtHR who decides over the admission or termination of a party’s role as co-respondent. In particular, the provision leaves open what exactly the ECtHR is supposed to do with a ‘reasoned assessment’ provided by the EU. This issue is addressed, however, in the revised draft explanatory report, according to which ‘[t]he conclusion of this assessment by the EU will be considered as determinative and authoritative’.[107] This appears to put in place an ‘almost mechanical automaticity’,[108] making the EU itself the ‘procedural gatekeeper’,[109] and thus reducing the ECtHR’s power to admit a co-respondent to a mere formality.[110] While perhaps not a model in terms of transparent treaty drafting, in sum these changes may be enough to convince the CJEU that the ECtHR – in the context of the co-respondent mechanism – will not assess the rules of EU law governing the division of powers between the EU and its Member States. By contrast, if the ECtHR could plausibly assume any power to second-guess the EU’s reasoned assessment, autonomy issues could once again arise.[111]

The second issue the CJEU had raised regarding the co-respondent mechanism concerned the attribution of responsibility. In this regard, the revised DAA scraps the possibility that the ECtHR may determine the sole responsibility of either co-respondent entirely. Instead, where the ECtHR establishes a violation in proceedings involving a co-respondent, it shall hold both respondent and co-respondent jointly responsible for that violation. That said, in order to avoid concerns that such joint responsibility may unduly affect EU Member States’ reservations made in accordance with Article 57 ECHR (and thus Article 2 of Protocol No 8 to the EU Treaties), Article 2(3) of the revised DAA stipulates that such reservations ‘shall retain their effect’ where the Member State is a co-respondent.[112] In doing so, the revised DAA avoids the ECtHR’s involvement in the internal attribution of responsibility between the EU and its Member States. This should suffice to meet the CJEU’s previous concerns.[113]

Prior involvement procedure: The CJEU’s fourth major concern was with the procedure for its own prior involvement in cases where it would not previously have had the opportunity to assess the compatibility of a relevant provision of EU law with the rights guaranteed by the ECHR. In Opinion 2/13, the CJEU took issue with two points: it criticized that the ECtHR would determine whether the CJEU had already given a ruling on the same question of law, which would be tantamount to granting Strasbourg jurisdiction to interpret the Luxembourg court’s case law. It also found it unacceptable that the envisaged procedure would not allow it to rule on the interpretation of provisions of secondary law. Thus, prima facie, it may seem surprising that the revised DAA has not substantially changed the provision on the prior involvement procedure, but again seeks to resolve the issues at the level of the revised draft explanatory report instead.[114] At least as regards the Court’s second concern, which had specifically pertained to the narrow definition of the phrase ‘[a]ssessing the compatibility’ in the original draft explanatory report, however, solving the issue at that level is certainly stringent.[115] As regards the CJEU’s first concern, the revised draft explanatory report now provides that ‘[d]etermining whether it is necessary to initiate the prior involvement of the CJEU […] depends on a finding by the EU of whether the CJEU has already undertaken the [relevant] assessment […]’.[116] As in the context of the co-respondent mechanism, such a finding by the EU is once again to be considered as ‘determinative and authoritative’.[117] What could be problematic from an autonomy perspective, however, is that the CJEU’s assessment would not be formally binding on the ECtHR.[118] In this regard it may be recalled that in Opinion 1/91 the Court found it ‘unacceptable’ that the treaty at issue envisaged a procedure that would result in answers of the Court that were ‘purely advisory and without any binding effect’.[119]

4.2.2. Will better be good enough?

The revised DAA clearly reflects the negotiators’ serious efforts to accommodate the CJEU’s autonomy concerns expressed in Opinion 2/13. While it therefore seems unlikely that the CJEU will engage in snowballing and torpedo the revised DAA by adding new and unforeseen autonomy requirements, the EU’s accession to the ECHR does not appear to be a done deal either. The ‘outsourcing’ of some of the more critical issues to the revised draft explanatory report appears to be a creative and politically feasible way to deal with some of the CJEU’s autonomy concerns. Whether it will also be legally effective remains to be seen – and much may depend on whether the Court applies a thick or thin understanding of autonomy. This is because, as shown above, the revised DAA may still fall short of fully resolving three of the CJEU’s autonomy concerns voiced in Opinion 2/13.

First, the newly included provision on mutual trust in the revised DAA attempts to resolve the autonomy issue by formally recognizing the principle of mutual trust, while at the same time emphasizing that the protection of human rights guaranteed by the ECHR must be ensured. While aiming to reconcile the approaches of the two Courts, the provision, however, falls short of the CJEU’s demand in Opinion 2/13 to generally prevent situations where the ECHR would ‘require a Member State to check that another Member State has observed fundamental rights, even though EU law imposes an obligation of mutual trust between those Member States’.[120] Thus, under the revised DAA, mutual trust would not automatically override the obligations under the ECHR. Moreover, as Tobias Lock has pointed out, the ECtHR would ‘continue to be the final arbiter of the limits of mutual trust where mutual trust comes up against fundamental rights’.[121] In light of these severe limitations, the viability of the ‘solution’ envisaged in the revised DAA seems very much to depend on the level of scrutiny that the CJEU applies. In other words, it may only suffice to resolve the autonomy issue stemming from the principle of mutual trust, if the Court applies a thin understanding of autonomy, focusing on formal assurances and ignoring practical effects.

Second, the revised DAA seeks to allay the CJEU’s concerns over the autonomy and effectiveness of the preliminary reference procedure by limiting the possibility for highest Member State courts to request advisory opinions pursuant to Protocol No 16 regarding questions falling within the scope of application of EU law. However, this attempt to avoid potential circumventions of Article 267 TFEU seems to create a new autonomy issue: when assessing whether a question put forward by such a court falls within the field of application of EU law, it appears that the ECtHR would nolens volens be required to determine the scope of EU law, even if only for the purpose of Protocol No 16.[122] This could clash with the CJEU’s claim to exclusive jurisdiction over the definitive interpretation of EU law, at least if the Court applies a thick understanding of autonomy.

Third, the fact that the CJEU’s assessment in a prior involvement procedure is not binding on the ECtHR seems to sit uncomfortably with the EU’s autonomy and, in particular, the nature of the function of the Court. It is true that the ECtHR would have various procedural options to accommodate a decision by the CJEU and put an end to a case, if it finds that the CJEU’s prior involvement has resolved the matter.[123] But still, the revised draft explanatory report explicitly states that the ECtHR is not bound by CJEU’s assessment. This could be seen as reducing the CJEU’s involvement to a purely advisory function – a role it emphatically rejected as ‘unacceptable’ in Opinion 1/91.[124] From this perspective, the solution envisaged in the revised DAA could raise autonomy concerns even if the Court applies a thin understanding of autonomy. Finally, it should be recalled that the CJEU’s concern pertaining to the CFSP was not addressed in the revised DAA. Whether the Court’s interpretation of the scope of its jurisdiction in CFSP matters in KS and KD[125] as well as Neves 77[126] has resolved the issue and thus paved the way for the EU’s accession to the ECHR remains to be seen.[127]

5.   Conclusion

Despite its immense importance for the EU legal order, EU autonomy continues to be plagued by vagaries regarding its normative content. Spurred by this discrepancy, the present contribution explores how we can explain the persisting elusiveness of the EU’s external autonomy. It identifies two conceptual tendencies in the Court’s relevant case law, which it refers to as snowballing and yoyoing. Both tendencies make autonomy an effective, but rather unpredictable tool.

Snowballing concerns not only the expansion of the principle over time, but also its mushing together with other key tenets of the EU constitutional framework into an increasingly all-encompassing argumentative amalgam. While the first aspect is owed to the gradual development of EU autonomy in the Court’s case law, it still means that what autonomy requires can often only be fully understood with hindsight. Moreover, by charging autonomy with ever more constitutional gravitas the Court signals a sense of the end justifying the means, even where its legal reasoning may lack in nuance and/or doctrinal rigor.

The second tendency, yoyoing, concerns the Court’s pivoting between different levels of scrutiny, referred to here as a ‘thick’ and ‘thin’ autonomy. Thick autonomy involves a scrupulous assessment of any potential impacts of the ‘external’ system on the autonomy of the EU legal order. Thus, when applying a thick understanding of autonomy, a threat could arise, for example, where an adjudicative body under an envisaged agreement could effectively – or even only hypothetically – encroach on the uniform interpretation and application of EU law and/or the Court’s ‘exclusive jurisdiction’. Thin autonomy, by contrast, is characterized by a more formal assessment of specific treaty arrangements. Thus, when applying a thin understanding of autonomy, the Court may, for example, be willing to ‘overlook’ the practical implications of an external adjudicative body’s treatment of EU law, so long as it has no formal jurisdiction to interpret and apply EU law ‘as law’ or formally bind EU courts to its understanding thereof. Since it is difficult to gauge in advance which level of scrutiny the CJEU will apply in a concrete case, the present contribution finds that yoyoing is an important factor contributing to the elusiveness of the autonomy of the EU legal order.

A case in point is Opinion 1/17, in which the Court to the surprise of many commentators adopted a thin understanding of autonomy and gave the green light to the ICS mechanism in CETA. However, the present contribution argues that the Court’s conciliatory approach was an expression of pragmatism, not a broader paradigm shift. Moreover, Opinion 1/17 tells us little about which legal safeguards or political necessities (perceived or real) would prompt the Court to adopt a similarly permissive approach in other contexts such as migration, environmental policy or human rights protection. Specifically with a view to the EU’s accession to the ECHR the contribution thus finds that – depending on the level of scrutiny the Court will apply – the revised DAA may still fall short of fully resolving all autonomy concerns.

On a more general note, the contribution finds that a certain degree of flexibility, including some yoyoing between thick and thin autonomy, is neither surprising, nor per se objectionable. Indeed, a certain pragmatism on part of the Court seems inevitable to ensure the EU’s ability to interact on the international level. The Court could, however, explain better why it applies a more or less exacting standard in a particular case. Put bluntly, what is problematic is not that the CJEU applies different standards, but the impression that it applies double standards.