BACKGROUND

Article 292 TFEU provides that the Council, the Commission and, in some circumstances, the European Central Bank ‘shall adopt recommendations’. The recommendation is one of the five types of legal act referred to in Article 288 TFEU. Like opinions, recommendations are described by that provision as having ‘no binding force’. They therefore lack the swagger of the regulation, the muscularity of the decision, the mystique of the directive. Yet the decision of the Grand Chamber of the Court of Justice in Kingdom of Belgium v European Commission,Footnote ¹ which concerned a challenge to the validity of a recommendation adopted by the Commission, revealed that there is more to recommendations than meets the eye.

The contested recommendation was concerned with ‘principles for the protection of consumers and players of online gambling services and for the prevention of minors from gambling online’.Footnote ² After a preamble of 30 paragraphs, its first substantive provision stated:

Oddly for a measure having no binding force, the recommendation then stated that it ‘does not interfere with the right of Member States to regulate gambling services’, as though this reflected an exercise of restraint by the Commission.

After a section on definitions, the remainder of the recommendation comprised ten sections dealing respectively with information requirements; minors; player registration and accounts; player activity and support; time out and self-exclusion; commercial communication; sponsorship; education and awareness; supervision; and reporting.

The section on supervision invited Member States ‘to designate competent gambling authorities when applying the principles laid down in this Recommendation to ensure and monitor in an independent manner effective compliance with national measures taken in support of the principles set out in this Recommendation’. The section on reporting invited Member States ‘to notify the Commission of any measure taken pursuant to this Recommendation by 19 January 2016…’ It also invited them ‘to collect reliable annual data for statistical purposes’ on a range of matters and to communicate them to the Commission by 19 July 2016.

THE PROCEEDINGS BEFORE THE GENERAL COURT

On 13 October 2014, Belgium brought an action before the General Court for the annulment of the contested recommendation under Article 263 TFEU.Footnote ³ This was a bold step, since the first paragraph of Article 263 expressly excludes recommendations from the class of acts which may be challenged in annulment proceedings. That exclusion is evidently connected with the failure of Article 288 to confer binding force on recommendations. Be that as it may, Belgium argued that the contested recommendation reflected an intention on the part of the Commission to harmonise the application to gambling of Articles 49 and 56 TFEU and that it therefore constituted ‘a hidden directive’.Footnote ⁴ It produced ‘indirect legal effects’Footnote ⁵ because the duty of sincere cooperation would require Member States to use their best endeavours to comply with it and national courts would have to take it into account where relevant.

The General Court dismissed the action as inadmissible. It pointed out that ‘any measures adopted by the institutions, whatever their form, which are intended to have binding legal effects, are regarded as actionable measures within the meaning of Article 263 TFEU…’Footnote ⁶ In order to establish whether a measure was ‘capable of having legal effects’,Footnote ⁷ it was necessary to examine its wording and context, its substance and the intention of its author. Although Article 288 made it clear that recommendations did not have binding force, the choice of form could not alter the nature of a measure. Therefore ‘the mere fact that the contested recommendation is formally designated as a recommendation and was adopted on the basis of Article 292 TFEU cannot automatically rule out its classification as a challengeable act’.Footnote ⁸

The General Court noted that the contested recommendation was ‘worded mainly in non-mandatory terms’.Footnote ⁹ This provided ‘a clear and specific indication’ that its content was ‘not intended to have binding legal effects…’Footnote ¹⁰ It did not include ‘any explicit indication’ that the Member States were ‘required to adopt and apply the principles set out therein’.Footnote ¹¹ Moreover, the Council, the European Parliament and the Commission had agreed that it was not appropriate at this juncture to propose specific EU legislation to regulate online gambling. Although the recommendation had been published in the L series of the Official Journal, this had ‘no bearing on whether the act at issue is capable of having binding legal effects…’Footnote ¹² The same was true of the detailed nature of the principles set out in the recommendation.Footnote ¹³

The General Court acknowledged that the Court of Justice had held in Grimaldi Footnote ¹⁴ that recommendations could not

However, that limited legal effect could not be enough to render recommendations reviewable, for that would be inconsistent with the terms of Article 263. For the same reason, claims based on the principle of conferral, institutional balance, effective judicial protection and fundamental rights were also rejected. An argument that the contested recommendation resulted ‘in unlawful harmonisation and liberalisation of the market in the online gambling sector’ was dismissed by the General Court as ‘based on a manifestly incorrect reading of the recommendation’.Footnote ¹⁵

THE APPEAL TO THE COURT OF JUSTICE

Belgium brought an appeal against the decision of the General Court before the Court of Justice. The Court of Justice declared that the purpose of Article 288 TFEU was ‘to confer on the institutions which usually adopt recommendations a power to exhort and to persuade, distinct from the power to adopt acts having binding force’.Footnote ¹⁶ It could not be established that a recommendation was reviewable under Article 263 TFEU merely by showing that, although it did not produce binding legal effects, it contravened certain principles or procedural rules.

The Court of Justice explained that ‘any provisions adopted by the institutions, whatever their form, which are intended to have binding legal effects are regarded as “challengeable acts” for the purposes of Article 263 TFEU…’Footnote ¹⁷ Thus, an act entitled a recommendation which was intended to have binding legal effects would not constitute a genuine recommendation and would be open to review under Article 263. Moreover, Article 267 TFEU gave the Court jurisdiction to give preliminary rulings ‘on the validity and interpretation of all acts of the EU institutions without exception…’Footnote ¹⁸ None of this was of any assistance to the appellant, however, and the appeal was dismissed.

MUCH ADO ABOUT NOTHING?

On one level, the outcome of the case was not surprising. Articles 263 and 288 TFEU are explicit. Indeed, the Belgian Government must have been advised in advance that its challenge was very unlikely to succeed. Why, then, did it go ahead?

It is possible that Belgium objected to the content of the contested recommendation. While it would have been aware that the recommendation was not binding, it might have been reluctant simply to ignore it. Another possibility is that Belgium objected to the form of the contested recommendation, which might have left the uninitiated uncertain whether it was binding or not.

Although the General Court described the contested recommendation as ‘worded mainly in non-mandatory terms’,Footnote ¹⁹ it did concede that certain provisions were in some language versions ‘drafted in more mandatory terms’.Footnote ²⁰ The versions concerned included those drawn up in Dutch and German, which are official languages of Belgium.Footnote ²¹ Was Belgium perhaps concerned about the discrepancy between those versions and the version in French, another of its official languages, which used non-mandatory terminology throughout?

Then there was the question of why the contested recommendation had been published in the L series of the Official Journal. The point may seem a trivial one, but presumably this did not happen by accident. It must have been the result of a deliberate choice. One is entitled to ask who made it and why. Was it perhaps intended to disguise the recommendation’s real status?

These were among the issues considered by Advocate General Bobek in a detailed OpinionFootnote ²² that concluded by recommending that the order of the General Court should be set aside and the case referred back to it for a decision on the merits.

The Advocate General made two powerful opening points.Footnote ²³ The first was that, ‘in view of the changing legislative landscape of (not only) EU law, which is marked by a proliferation of various soft law instruments, access to the EU courts should be adapted in order to respond to those developments’. The Court should, he said, recognise ‘the fact that there are norms generating significant legal effects that find themselves beyond the binary logic of binding/non-binding legal rules’. The second was that a normative instrument such as the contested act ‘that in the light of its logic, context, purpose and partially also language, can reasonably be seen as setting rules of behaviour, ought to be subject to judicial review, irrespective of the fact that it is somewhat disguised as a set of mere “principles” in a recommendation’.

Advocate General Bobek pointed out that, when the contested recommendation was adopted, it was accompanied by a Commission memorandumFootnote ²⁴ which stated:Footnote ²⁵

The memorandum added:

The wording of the memorandum is striking. Although the Commission does not conceal the fact that recommendations are not binding, the reference to ‘sending a clear message to Member States’ suggests that there will be adverse consequences if they fail to do what is expected of them, thereby impeding the development of uniform principles. In fact, with the exception of the acknowledgement that recommendations are not binding, the remainder of the description of the recommendation given by the Commission could apply equally well to directives. From the perspective of the Commission, recommendations even have the advantage over directives of not requiring any involvement by other institutions.

On the substance of the case, Advocate General Bobek took the view that the General Court had not interpreted the effects of the contested recommendation correctly. It had consequently failed to address properly the admissibility of the application. In the ERTA case,Footnote ²⁶ the Court of Justice had held that annulment proceedings should be available for the purpose of challenging ‘all measures adopted by the institutions, whatever their nature or form, which are intended to have legal effects’. In its subsequent case law, the Court of Justice had made it clear that, when deciding on the admissibility of an action for annulment, the substance of the contested measure was more important than its form. One of the problems with the ERTA test, however, was that over time it was ‘effectively becoming narrower and narrower’ and was ‘falling out of sync with the evolution of the EU’s normative landscape’.Footnote ²⁷ The Advocate General declared: ‘In a world where various instruments of soft law are, in fact, becoming much more numerous and significant than in 1971 [when ERTA was decided], the conditions for standing and judicial review should react to such developments’.Footnote ²⁸

Advocate General Bobek identified two particular challenges facing the ERTA test. One was ‘the rise of various forms of soft law that strictly speaking do not have binding force but at the same time generate legal effects...’. The other was that ‘recommendations are in practice likely to generate a number of legal effects, often quite significant ones, on both the EU level as well as the national level…’Footnote ²⁹ The higher courts in a number of Member States, he said, had been opening up judicial review to include acts that were not strictly binding. This had happened where, for example, an act was perceived as effectively binding because it contained incentives, or when its author had the power to impose sanctions, or when it produced significant effects on its addressees. The Advocate General observed:Footnote ³⁰

According to Advocate General Bobek, recommendations generally fitted that description. Although said to be non-binding, they could ‘generate considerable legal effects, in the sense of inducing certain behaviour and modifying normative reality. They are likely to have an impact on the rights and obligations of their addressees and third parties’.Footnote ³¹ They were also ‘likely to be used in legal interpretation, in particular in order to give meaning to indeterminate legal notions contained in binding legislation’.Footnote ³² They ‘could be used as more than just tools for advancing policies that are politically (lack of consensus) or legally (no specific powers to that effect) gridlocked. They could also potentially be used as a tool to circumvent the same legislative processes’.Footnote ³³ They might even pre-empt the legislative process by shaping

Advocate General Bobek therefore proposed a readjustment of the ERTA rule based on three factors.

1. (1) The first factor was ‘the degree of formalisation (does the EU measure take on the form of a legal act?) and of definitiveness of the measure (has it been adopted at the end, as the culmination of a consultation or, more generally, in a ‘soft-law making’ process?)’.Footnote ³⁵ As far as the form of the act in question was concerned, to be reviewable ‘it must appear to be like a legal text so that it could be reasonably perceived as producing legal effects. In this respect, an act will appear to be a legal act if, for instance, it is divided into articles or at least sections, if it is published in the Official Journal (certainly in the L series, where legislation is supposed to be published)’.Footnote ³⁶

2. (2) The second factor related ‘to the content and overall purpose of the contested act: how precise are the “obligations” contained therein? What is the general aim pursued?’Footnote ³⁷

3. (3) The third factor concerned enforcement. ‘Does the measure contain any clear and specific compliance, enforcement, or sanction mechanisms? Naturally, this does not aim only at direct enforcement, which is very unlikely to be present, but at indirect mechanisms or enforcement, both structural and institutional’.Footnote ³⁸ Structural compliance mechanisms might include ‘reporting, notification, monitoring, and supervision. Elements of peer pressure might also be of relevance, such as the publishing of performance tables, reports involving public naming and shaming, and so on’.Footnote ³⁹

Finally, the Advocate General argued that the contested act would be likely to induce compliance if the adopting institution had the power to impose sanctions on the addressees in the same or related fields.

Advocate General Bobek concluded that the contested recommendation went ‘considerably beyond what might be expected from a document that simply recommends certain principles. In this specific case, it can indeed be argued that that Recommendation is bound to produce legal effects and that reasonable addressees are likely to modify their behaviour in order to comply, at least partially, with the Recommendation’.Footnote ⁴⁰ He said that ‘the level of detail and precision of concrete provisions of the Recommendation is striking. Far from setting out mere “principles”, the Recommendation lays down rather clear and precise rules’.Footnote ⁴¹

Should the Court of Justice reject that approach in favour of a strict application of Article 263, Advocate General Bobek said that it should take that view to its logical conclusion. This would mean recognising that

It would also be necessary for the Court of Justice to revisit Grimaldi and state clearly that, while national courts were at liberty to take recommendations into account if they wished, they were under no obligation to do so. An act that was not binding could not impose obligations on national courts.

THE COURT OF JUSTICE HAS ITS CAKE AND EATS IT

It is a matter of concern that the Court of Justice declined to follow the Opinion of Advocate General Bobek. Not only did it apply the conventional reading of the Treaty that he had so effectively dismantled, but it also failed to underline that recommendations produced no legal effects or to reverse Grimaldi. The outcome is patently inconsistent: recommendations cannot be the subject of annulment proceedings because they have no binding force, but national courts are bound to take them into consideration.

More generally, the decision of the Court of Justice disturbs the institutional balance by offering the Commission a way of circumventing the decision-making processes laid down by the Treaty. Instead of submitting a formal proposal which might have to be amended to secure the approval of the Council or the European Parliament, the Commission may unilaterally issue recommendations in the knowledge that some Member States will comply with them and national courts will be required to take them into account. Moreover, third parties may sometimes consider themselves bound by recommendations. This may be because a recommendation may look like a binding act or contain apparently mandatory language or the third party does not wish to damage its standing in the eyes of the Commission. It seems self-evident that this sort of off-balance-sheet law-making is highly detrimental to the integrity of the Union legal order, particularly in the way it marginalises two of the three institutions mentioned in Article 10(2) TEU (the other is the European Council) presented as evidence of the Union’s democratic legitimacy. It is significant that Article 10(2) does not mention the Commission.

There are two ways in which the Court of Justice could have avoided so perilous an outcome. It could have said that the contested act was not a true recommendation: it was too detailed; the use of mandatory language in some of the language versions threatened legal certainty; it was published in the L series of the Official Journal. It might then have been treated as an innominate act for the purposes of the ERTA line of case law. The Court of Justice stated in that case that what is now Article 263 TFEU ‘treats as acts open to review by the Court all measures adopted by the institutions which are intended to have legal force’.Footnote ⁴³ It went on to add that an action for annulment had to be available ‘in the case of all measures adopted by the institutions, whatever their nature or form, which are intended to have legal effects’.Footnote ⁴⁴ As Advocate General Bobek pointed out, that test subsequently became somewhat stricter as the term ‘legal effects’ came to be replaced by ‘binding legal effects’, the term used by both the General Court and the Court of Justice in the Belgium case. Advocate General Bobek observed that binding legal effect was a much narrower category than legal effect.Footnote ⁴⁵ However, for present purposes, the Court could have found that the contested act satisfied the narrower test because of the ruling in Grimaldi that national courts were bound to take recommendations into consideration. It would have made no sense to exclude from that rule a fictitious recommendation being treated as an innominate act for the purposes of the ERTA test.

The drawback of the approach described in the previous paragraph is that it would have done little to discourage the (ab)use by the Commission of fictitious recommendations. A better approach would therefore have been to follow the second solution proposed by Advocate General Bobek and to deny recommendations any legal effect whatsoever.

That the Court of Justice did not adopt either of those solutions is grim news for those who believe in the importance of judicial review in protecting the rule of law and individual rights. Having shut down access by private parties and out-sourced much of the responsibility for upholding their rights to the national courts of the Member States, there are now signs that the European Court of Justice is taking a stricter approach to the notion of a reviewable act under Article 263 TFEU.

The Belgium case is not an isolated example. In NF v European Council,Footnote ⁴⁶ the applicant sought the annulment of the so-called EU-Turkey statement of 18 March 2016 designed to reduce irregular migration from Turkey to Greece. That statement attracted a wave of criticism on humanitarian grounds.Footnote ⁴⁷ Its central plank was that irregular migrants from Turkey seeking to enter Greece would be returned and that, for every migrant of Syrian nationality returned to Turkey, another Syrian would be resettled in the EU.

The statement was published in the form of a press release, according to which the members of the European Council and their Turkish counterpart had met and the EU and Turkey had agreed that certain actions should be taken.Footnote ⁴⁸ As the General Court acknowledged, this ‘could, admittedly, imply that the representatives of the Member States of the European Union had acted … in their capacity as members of the “European Council” institution [sic]…’Footnote ⁴⁹ However, the General Court concluded that, ‘notwithstanding the regrettably ambiguous terms of the EU-Turkey statement … it was in their capacity as Heads of State or Government of the Member States that the representatives of those Member States met with the Turkish Prime Minister on 18 March 2016 in the premises shared by the European Council and the Council…’Footnote ⁵⁰ The statement could not, therefore, be considered an act of the European Council.

Was this outcome compatible with Article 47 of the Charter of Fundamental Rights (right to an effective remedy and a fair trial)? On this point, the General Court noted that Article 47 was ‘not intended to change the system of judicial review laid down by the Treaties…’Footnote ⁵¹ The authority given for that proposition was the judgment of the Court of Justice in Inuit Tapiriit Kanatami v Parliament and Council.Footnote ⁵² That judgment refers to the Explanation on Article 47 of the Charter, which does indeed make this point. But let us be clear: the outcome was the result of a deliberate policy choice by the Court of Justice. The Explanations on the CharterFootnote ⁵³ are not themselves part of the Treaties and are merely required to be given ‘due regard’ when the Charter is applied.Footnote ⁵⁴ They state on their face that ‘they do not as such have the status of law’. The Court of Justice could have chosen a different path in Inuit and the General Court should have done so in NF, where some of the values on which the Union is said by Article 2 TEU to be founded were at stake. One can understand why the Court of Justice is so defensive about the prospect of Union accession to the European Convention on Human Rights.Footnote ⁵⁵

From a doctrinal point of view, the case law on the action for annulment is incoherent.Footnote ⁵⁶ The Court of Justice has frequently departed from the text of the Treaty. ERTA is arguably an example, indubitably so are Les Verts Footnote ⁵⁷ and Chernobyl Footnote ⁵⁸ (the latter case reversing the decision in Comitology Footnote ⁵⁹ delivered less than two years previously). Further departures, in Extramet Footnote ⁶⁰ and Codorniu,Footnote ⁶¹ were necessary to deal with the chaotic state of the case law concerning the concept of a decision and direct and individual concern under Article 173 EEC. Then there is the apparent absence of any legal basis for the requirement to show an interest in bringing proceedings, a requirement that is not mentioned in Article 263 and does not apply to privileged applicants.Footnote ⁶²

These developments have made it very difficult for the Court of Justice to hide behind the wording of the Treaty when seeking to justify its decisions in this field. Following Codorniu, the case law seemed to have been stabilised, but an abrupt change of direction occurred in UPA.Footnote ⁶³ That case led to a Treaty change at Lisbon, relaxing the standing rules applicable to private parties, but the relaxation was narrowly construed in Inuit.Footnote ⁶⁴

Although the action for annulment is self-evidently more effective than the preliminary rulings procedure as a means of reviewing the validity of Union acts, in Inuit the Court of Justice continued the process of delegating responsibility for undertaking that task to the national courts. This may lead to at least the appearance of inconsistency. In Gauweiler and Others v Deutsche Bundestag,Footnote ⁶⁵ the German Bundesverfassungsgericht sought guidance on the validity of certain decisions taken by the European Central Bank and set out in a press release. Although the decisions in question had yet to be implemented and this would not be possible until further legal acts had been adopted, the Court of Justice declared the reference admissible. The judgment followed the Opinion of Advocate General Cruz Villalón, who observed that the alternative

The validity of the decisions set out in the press release was upheld by the Court of Justice.

The circumstances of the Gauweiler case were unusual in that it seems to have been the European Central Bank’s intention in issuing the press release to produce an effect on the markets, a ruse which appears to have worked. However, one is entitled to ask whether the Court of Justice would have taken the same approach to admissibility had the issue arisen in annulment proceedings.

One should not need to ask the question. It arises because in direct actions the Court of Justice and the General Court now sometimes seem content to collude with other institutions to evade the requirements laid down by the Treaty and deny applicants the right to an effective remedy. The Courts’ reluctance to engage directly in effective judicial review is no longer confined to cases brought by private applicants. Privileged applicants may also be caught in the net. Perhaps the cases concerned are outliers and do not presage the start of a new and unwelcome trend in the case law. But alarm bells are ringing.