Dr Eleni Frantziou, Associate
Professor in Public Law and Human Rights, Durham Law School, and Dr Sylvia
de Mars, Reader in Transnational Public Law, Newcastle Law School
Photo credit: Samuel
Lennox, via Wikimedia
Commons
Introduction
Is there anything left to say on
the direct effect of EU law? Yes, according to the UK Supreme Court, which
handed down its much-awaited judgment in Dillon et
al on 7 May 2026. The UK reference above is not a typo:
despite Brexit, aspects of the EU/UK Withdrawal
Agreement (‘WA’) have direct effect under the conditions provided for in EU
law, in line with Article 4 WA. In Dillon, the UK Supreme Court was
asked to interpret whether one provision of this agreement – Article 2(1) of
the Ireland/NI
Protocol (‘Protocol’), which is itself a core part of the WA (Article 182
WA) – is directly effective and, if so, on what terms.
The case concerned the Legacy
(Troubles and Reconciliation) Act 2023, which set out a broad range of
immunities for serious crimes committed in Northern Ireland during the
Troubles. The principal claimants were one direct victim and three relatives of
victims killed in the conflict – a wife, mother and sister who, as O’Donoghue
observes, were but minimally
acknowledged by the UKSC. They challenged the immunities provisions of the
Act under the ECHR and EU law.
In this post, we focus only on
the EU law dimensions of the case and, particularly, on the question of direct
effect. In our view, Dillon raises significant questions about the
correct application of the direct effect conditions, the ultimate arbiter of
which is the CJEU. In particular, we argue that the UKSC’s interpretation of
direct effect in Dillon is inconsistent with EU law, because it creates
an unprecedented requirement of ‘double direct effect’: it applies the direct
effect conditions to both a provision of the Withdrawal Agreement/Protocol and
to measures referenced in this provision.
Why was EU law still relevant
in Dillon in 2026?
One of the most contentious aspects
of the Withdrawal Agreement negotiations was the status of Northern Ireland. It
is widely
documented
that EU membership was one of the principal catalysts for the Belfast/Good
Friday Agreement (‘BGFA’) that officially marked the end of the Northern
Ireland conflict known as ‘the Troubles’. As EU membership by both the UK and
Ireland had ensured free movement between the Irish north and south, the
creation of special arrangements that secured this after Brexit was a crucial
aspect of the UK’s exit negotiations.
These special arrangements were
embedded into the Ireland/NI Protocol (as of 2023 also known as the ‘Windsor
Framework’ in the UK), which was appended to the Withdrawal Agreement as a
core part of it dealing with a specific ‘withdrawal’ issue. The Protocol provides for continued
north/south cooperation, avoiding a hard border and protecting both trade and
cross-border work. Importantly for the purposes of the Dillon case, the Protocol
also notes in recital 7 of its Preamble that ‘Union law has provided a
supporting framework for the provisions on Rights, Safeguards and Equality of
Opportunity of the 1998 Agreement’ (‘RSEO’) and includes a specific non-regression
guarantee with respect to this part of the BGFA. To this end, Article 2(1) of
the Protocol provides:
The United
Kingdom shall ensure that no diminution of rights, safeguards or equality of
opportunity, as set out in that part of the Belfast/Good Friday Agreement entitled
Rights, Safeguards and Equality of Opportunity results from its withdrawal from
the Union, including in the area of protection against discrimination, as
enshrined in the provisions of Union law listed in Annex 1 to this Protocol,
and shall implement this paragraph through dedicated mechanisms.
It was this article of the
Protocol that formed the basis of the EU law claim in Dillon. The
claimants’ argument went as follows: since the UK had been subject to the
Victims’ Rights Directive (‘VRD’) before Brexit and this had been implemented
in NI through domestic measures, the immunities created by the Legacy Act
reduced the level of protection of fundamental rights available during the UK’s
membership of the EU, and were therefore in conflict with Article 2 of the NI
Protocol. This claim was made possible by Article 4(1) WA, which states that
the “Agreement shall produce in respect of and in the United Kingdom the same
legal effects as those which they produce within the Union and its Member
States” and the test for individuals to rely on a provision of the Withdrawal
Agreement before domestic courts is that such a provision should “meet the
conditions for direct effect under Union law.”
In line with the widely known Van
Gend en Loos formula, a provision of EU law has direct effect if it
confers rights to individuals that are clear, precise, and unconditional,
rather than being subject to the adoption of further measures by either Member
States or the EU. The wording of Article
2(1) of the Protocol prima facie meets these conditions with ease. The
right conferred by Article 2(1) is to not have ‘rights,
safeguards or equality of opportunity’ as set out in the BGFA and in Annex 1 of
the Protocol diminished. This paragraph contains a ‘textbook’ negative
obligation. It requires no further measures at all – the UK simply needs to
refrain from introducing new measures that reduce the level of protection of EU
fundamental rights as it stood on 31 December 2020.
The claimants consequently argued
that, since Article 4 WA also provides for the Agreement to have primacy and,
at paragraph 2, requires that domestic courts be capable of disapplying
incompatible provisions, the immunity provisions of the Legacy Act had to be
disapplied. Two
courts
in Northern Ireland had found for the claimants, disapplying the offending
provisions – but the government challenged this on appeal to the UKSC. It
argued that if Article 2 was to have direct effect, it could only do so by
reference to obligations that were themselves clear, precise and unconditional.
The Supreme Court sided with the
government. It found that Article 2 could not always have direct effect, but
only under certain circumstances, namely when the measures cross-referenced in
Article 2 were themselves directly effective.
It is worth setting out its reasoning in some detail.
The UKSC commences, at paragraph
112, by citing the so-called Demirel
test. Demirel is one of the
early cases in which the CJEU considered whether provisions in bilateral
agreements signed by the EU could have direct effect; and it found at paragraph
14 that:
…a provision
in an agreement concluded by the European Union with a non-member country must
be regarded as being directly applicable when, regard being had to its wording
and to the purpose and nature of the agreement, the provision contains a
clear and precise obligation which is not subject, in
its
implementation or effects, to the adoption of any subsequent measure... (emphasis
added)
So far, so good. While, as we discuss further below, the
treatment of the Withdrawal Agreement (or Protocol) as part of ordinary
EU external relations law is questionable, this does reflect the commonly used
test for evaluating the direct effect of bilateral agreements involving the EU.
However, the UKSC proceeds with
the following statement in paragraph 113:
The obligation
imposed on the United Kingdom by article 2(1) of the Windsor
Framework
relates to rights, safeguards or equality of opportunity “as set out in” the
RSEO chapter. It is therefore necessary to consider whether, having regard to
the wording and to the purpose and nature of the RSEO chapter and of article
2(1) of the Windsor Framework, those provisions read together impose a clear
and precise obligation which satisfies the test for direct effect.
In our view, the words “therefore
necessary” make a huge argumentative leap, hitherto unseen in any EU law treatment
of direct effect. In this single short paragraph, the CJEU takes the Demirel
test but applies it simultaneously to the Withdrawal Agreement and
its provisions and the BGFA’s RSEO chapter’s provisions. In other words,
the UKSC reasons that Article 2 of the Protocol cannot have direct effect as
such: for it to be directly effective, anything it references must also be
directly effective. This leads to its
even more confounding finding, which is – ultimately – that Article 2 of the
Protocol sometimes has direct effect – but at other times it does not.
This is because, according to the UKSC, it is impossible to identify a clear
and precise obligation without reference to the RSEO chapter. As the UKSC puts
it at paragraph 116, “[i]t is only in this way that the obligation not to
diminish rights, safeguards or equality of opportunity acquires any content.”
This aspect of the ruling – which
we dub ‘double direct effect’ –presents a deep challenge to settled
understandings of the direct effect of EU law.
The fallacy of the UKSC’s
direct effect
The UKSC’s ruling does not
question the possibility of Article 2 of the Protocol to have direct effect,
nor does it challenge the terms of direct effect. According to the Court, as
per paragraph 112, the relevant conditions are the conditions set out in EU
law. But the choice of EU law is already telling of the ruling’s direction and
tone: rather than going straight for the Van Gend en Loos conditions,
the Court refers to the Demirel test on the direct effect of international
agreements.
While this choice is defensible
to a degree as the WA is, after all, a bilateral agreement, the confidence with
which the UKSC resorts to it is puzzling. Unlike association agreements, such
as the one between the EU and Turkey at stake in Demirel, or even the
UK/EU Trade and Cooperation Agreement, EU competence for which resides in
external relations law (Article 217 TFEU), the Withdrawal Agreement is a very specific
type of agreement. The competence to conclude it is found in the constitutional
part of the Treaty on European Union – Article 50 TEU – and it is Article 50
(2) that sets out that the terms of withdrawal are to be negotiated in
accordance with Article 218(3) TFEU. The Withdrawal Agreement is thus a
category in its own right: it is not about a third state negotiating the terms
of its association with the EU, however deep, but about a Member State
negotiating its departure from the bloc in line with its own constitutional
requirements, as set out in Article 50 TEU.
The exceptionality of the Withdrawal Agreement has been raised indirectly
in litigation before the CJEU and, albeit that the Court has not had specific
occasion to address what this means for direct effect, it has emphasised the significance
of Article 50 TEU as the essential legal context for analysing the provisions
of the Withdrawal Agreement (see, e.g., Préfet
du Gers I, para 54).
But even if we were to accept
that the Withdrawal Agreement should be treated just like any other bilateral
treaty, its own terms make its ‘intention’ clear, given the express wording of
Article 4 WA on the application of the direct effect conditions – a very
unusual stipulation that, to our mind, would not go unnoticed by the CJEU. In
fact, as Gallo
and Labus note, specific exclusions of direct effect are the norm in the
external relations context, as is indeed the case for the TCA, Article 5 of
which preclude direct reliance by individuals on its provisions. The choice of
the Demirel conditions is, therefore, significant for at least two
reasons: first, it is important symbolically. The UKSC makes a choice to treat
the Withdrawal Agreement as one of many international agreements the UK has
signed. This automatically limits any perceived onus on its part to read the
‘purpose and nature’ of the Withdrawal Agreement or the Protocol as resulting
in different obligations than other international law. Second, the
reliance on the Demirel test and its focus on not only the precision and
unconditionality of the relevant provisions, but also the ‘purpose and nature’
of the agreement, enables the UKSC’s further findings about Article 2(1) WA.
For, as already highlighted above, rather than treating ‘purpose and nature’ as
pertaining to the Withdrawal Agreement and its provisions, the UKSC
understands this test as applying to the RSEO section of the BGFA.
The ‘purpose and nature’ of the
BGFA is summarised in paragraph 119 as “establishing peace in Northern Ireland
after decades of sectarianism and civil conflict”. This reference appears to
preclude a reading of the BGFA as aiming to secure cross-community equality in
a broader sense, which makes reliance on Article 2 for the purposes of ensuring
the non-diminution of EU fundamental rights altogether less likely. It
will be far more difficult to link provisions in the RSEO chapter to EU law if it
is essential to prove that their purpose and nature was ‘establishing peace’.
This focus on the Demirel condition of the nature and purpose of the
agreement thus appears to negate, at least to an extent, the practical
relevance of Article 2 as a non-regression clause. Rather than protecting the
‘supporting framework’ of fundamental rights that membership of the EU meant
for the peace process in tune with the Protocol’s Preamble, the interpretation
could prove to be much narrower– a possibility that remains to be fleshed out
further in subsequent case law.
Beyond the type of direct effect
it chooses, though, the UKSC also commits, in our view, a fundamental,
substantive error of EU law in its application of the direct effect conditions.
This error rests on the finding that Article 2(1) is a provision that could
sometimes, but not always, have direct effect – depending on what else it
references. This position does not withstand scrutiny from the perspective of
EU law.
It is trite EU law, including
under Demirel, that the direct effect conditions refer to ‘a provision’.
Leaving the categorical exclusion of certain forms of direct effect for certain
instruments (eg, the horizontal direct effect of directives), the CJEU has
never held that a provision can both have and lack direct effect. The test is a
clear binary: the provision either meets the conditions and it is directly
effective or it does not, and it is not.
In this sense, it may have been
more coherent for the UKSC to state that Article 2 is not directly effective at
all. That, however, would have created the problem that some of the EU law
explicitly referred to in the provision (the Equality directives) are par
excellence directly effective obligations. Finding that Article 2(1) lacks
direct effect altogether would have created a procedurally awkward situation,
whereby the Annex to the Protocol has direct effect and supremacy in accordance
with Article 4(1) WA, but its ‘activating’ provision does not. It was also
clear that the provision was thought to have direct effect when negotiated and
the UK government did not challenge the right of individuals to invoke it with
respect to the annexed Directives; while the UKSC made it clear that UK
government interpretations of its own commitments made when concluding the WA
were of no help in determining of Art 2 of the Protocol was directly effective
(see paragraphs 123-124 of the judgment), it seems to have wanted to avoid
making the existence of that Annex 1 and, indeed, Article 2 itself wholly pointless
from a litigation perspective. Hence, the UKSC resorts to a half-way house: if
the provisions against which non-diminution is sought are themselves directly
effective, then Article 2 can be invoked before domestic courts. If they are
not, it cannot. Much like Schrödinger’s cat, then, direct effect is simultaneously
alive and dead – it all depends on what Article 2 is referencing.
This reasoning in our view
misunderstands (though it does not reference) CJEU case law on the direct
effect of provisions that also refer to other provisions, as well as the
nature of the direct effect conditions as interpreted in EU law, more
generally. For example, Charter provisions regularly give ‘specific expression’
to non-directly effective measures of directives (see, eg, the Grand Chamber rulings
in Egenberger,
Braathens,
KL
v X). Unlike the approach espoused by the UKSC, it is not the
non-directly effective measures detailing the core obligation that suddenly
acquire direct effect. Rather, the presence of a directly effective measure, such
as Art 47 CFR, makes the more specific obligations listed in the directives
invocable in court, where they otherwise would not be. The core obligation is
directly effective regardless of what it cross-references. The CJEU makes this
very clear in Egenberger:
78. [L]ike
Article 21 of the Charter, Article 47 of the Charter on the right to
effective judicial protection is sufficient in itself and does not
need to be made more specific by provisions of EU or national law to confer
on individuals a right which they may rely on as such [emphasis added].
In other words: it is the more
general provision of primary law – Article 47 of the Charter – that gives rise
to a clear, precise and unconditional obligation. This does not require a
cross-reference to a separate provision that is also directly effective in
order to be relied upon. Rather, the only effect of the directives in the above
cases is to show that the scope of EU law – and hence the relevant provision of
the Charter – was engaged.
We found similar reasoning in
CJEU judgments that involve international agreements, where the
cross-referencing to Annexes and Protocols is very common. In Sevince,
for instance, concerning the EEC-Turkey Association Agreement, the CJEU found
in paragraph 22 that provisions in several decisions adopted under this
association agreement had direct effect – and that this
cannot be
affected by the fact that [other provisions] provide that the procedures for
applying the rights conferred onto Turkish workers are to be established under
national rules. Those provisions merely clarify the obligation of the
Member States to take such administrative measures as may be necessary for the
implementation of those provisions, without empowering the Member States to
make conditional or restrict the application of the precise and unconditional
right which the decisions of the Council of Association grant to Turkish
workers. (emphasis added)
The analogy is, in our view,
powerful: similarly to Article 2, this case concerned a standstill obligation
that no further obstacles to the free movement of workers be imposed. Moreover,
like Article 2, there was an expectation that national procedures would be
introduced to ensure the effectiveness of this core obligation. The CJEU clearly
distinguished this procedural conditionality from the primary negative duty.
The idea that the ‘referred-to’
provisions do not themselves have to be directly effective for a primary provision
to be directly effective is also expressly confirmed in the 2022 ruling in Deutsche
Umwelthilfe, which concerned standing requirements in the
implementation of the Aarhus Convention. The CJEU’s reasoning is worth citing
in some detail:
66. [W]hile it
is true that Article 9(3) of the Aarhus Convention does not have direct
effect in EU law and cannot, therefore, be relied on, as such, in a dispute
falling within the scope of EU law, in order to disapply a provision of
national law which is contrary to it, the fact remains that, first, the primacy
of international agreements concluded by the European Union requires that
national law be interpreted, to the fullest extent possible, in accordance with
the requirements of those agreements and, secondly, that Article 9(3) of
the Aarhus Convention, read in conjunction with Article 47 of the Charter,
imposes on Member States an obligation to ensure effective judicial protection
of the rights conferred by EU law, in particular the provisions of environmental
law...
…
78. … the
discretion conferred on the Member States to lay down rules governing the right
to bring proceedings, referred to in [Article 9(3) of the Aarhus Convention],
does not affect their obligation to ensure a right to an effective remedy
enshrined in Article 47 of the Charter, as, moreover, also alluded to in
Article 9(4) of the Aarhus Convention. Article 47 of the Charter is
sufficient in itself and does not need to be made more specific by provisions
of EU or national law in order to confer on individuals a right which they may
rely on as such … Thus, that article may be relied on as a limit on the
discretion left to the Member States under Article 9(3) of the Aarhus
Convention.
This case is also analogous to
what the UKSC had to consider in Dillon: a primary provision that set
out a clear, precise and unconditional right (Article 47 of the Charter) – and
further provisions, covering more specific situations in which such a right
might apply, which are not directly effective (in Article 9(3) of the Aarhus
Convention). The CJEU had every
opportunity to set out that both Article 9(3) of the Aarhus Convention and Article
47 of the Charter had to be directly effective, but did not do so – instead, it
stressed that Article 47 sets out framework conditions for how a conditional,
imprecise referred provision had to function.
This is in our view similar to the function that Article 2 of the
Protocol carries out: it sets out the conditions that apply once the
RSEO and Annex 1 directives need to be considered, in the form of setting a
non-diminution test.
Finally, the UKSC’s understanding
of the core negative obligation in Article 2 as vacuous without a
contemporaneous assessment of the RSEO sets too high a threshold for clarity
and precision. If this was how its reasoning worked, the CJEU would have denied
the direct effect of most provisions of EU primary law. To take one iconic
example, it clearly would have declined to accept the direct effect of Article
157 TFEU in Defrenne back in 1976; as Daniele Gallo puts it, “[t]he
contents of the obligation of [what is now Article 157 TFEU] were considered
sufficiently clear, to the extent that the Member States must ensure ‘the
application of the principle that men and women should receive equal pay
for equal work’, despite … that there was, at the time, relative uncertainty
about the concept of ‘equal work’ as well as that of ‘work of equal
value’. (Gallo, OUP 2025, 83).
To summarise, the RSEO determines
the scope of Article 2 and it is squarely for the Supreme Court to interpret
what this means. As noted above, there may be disagreement about how narrowly
the ambit of the RSEO was drawn in this ruling – disagreement which can be
further discussed in the Specialised Committee on the Windsor Framework. Such
disagreement, however, would concern the substantive question of diminution (i.e.,
was there a relevant RSEO right that has now been diminished?) – and not the
relevant test for direct effect. As a matter of the direct effect conditions,
there can be little doubt, based on CJEU authority on direct effect, that
Article 2 can be invoked before domestic court. In turn, if disagreement arises
about the direct effect conditions, the correct interpretation must be
determined by the CJEU – a point to which we now turn, in concluding.
Is this the final word on the
direct effect of Article 2 NIP?
Assuming that our analysis is
correct, and that the application of the direct effect conditions in Dillon
is inaccurate, can anything be done about it? The UKSC was not entitled to make
a reference to the CJEU on the matter of direct effect, as this is not provided
for under this part of the Protocol. While references are possible with regard
to a limited number of other provisions, primarily relating to trade, they are
not provided for in the context of the non-diminution guarantee. It is to be
hoped, however, that the Joint
Committee overseeing the WA will discuss these issues in its upcoming
meeting, particularly given the significance of the direct effect question for
ongoing litigation on Article 2 at lower levels and, more generally, for the
very ability of the Dedicated Mechanism
set up under the Protocol to contribute to the enforcement of the
non-diminution guarantee through litigation.
Arguably, regardless of the
outcome of the political discussion, a question should now be put to the CJEU
about the correct application of the direct effect criteria by agreement of the
parties. The Dillon case demonstrates the technical difficulties
associated with ongoing Brexit litigation that involves EU law concepts without
the possibility of interpretive clarity through the preliminary reference
structure, and it is in nobody’s interest for different versions of a 60 year old
test to apply in the EU itself and its closest trading partner. Ultimately, though,
if no agreement is reached about what the correct interpretation of the direct
effect conditions is in the context of Article 2 within the Joint Committee, an
arbitration panel may be appointed to determine the issue. Its decision will be
binding on both parties. Crucially, the arbitration panel would be required
to make a reference to the CJEU to determine any issues of EU law before it
reaches its decision. Undoubtedly, the concept of direct effect would be such
an issue.
