Dillon: the Supreme Court on the Windsor Framework and the Legacy Act

A legal analysis

On 7 May 2026, the Supreme Court handed down a unanimous judgment in Dillon v Secretary of State for Northern Ireland [2026] UKSC 15 that clarifies several important questions about the post-Brexit constitutional settlement for Northern Ireland, and in doing so reverses a number of conclusions reached by the Northern Ireland courts below. Although the political consequences of the ruling are partially blunted by the Labour Government’s commitment to repeal and replace the Northern Ireland Troubles (Legacy and Reconciliation) Act 2023 (the “2023 Act” or “Legacy Act”), the legal consequences extend well beyond the legacy context.

The Supreme Court has handed down a judgment of constitutional significance relating to the scope of the power of courts in Northern Ireland to disapply, or treat as of no force and effect, primary legislation where directly effective rights protected through Article 2(1) of the Windsor Framework and section 7A of the European Union (Withdrawal) Act 2018 are infringed.

1. What the Court decided

Three points stand out:

1. Article 2(1) of the Windsor Framework is justiciable, but on far narrower terms than the Northern Ireland courts had allowed. It cannot be married to the broad, aspirational language of paragraphs 1, 11 and 12 of the “Rights, Safeguards and Equality of Opportunity” (“RSEO”) chapter of the Belfast Agreement to generate directly enforceable rights at large. Article 2(1) requires an identifiable EU-law right or obligation which falls within the ambit of the RSEO chapter and which, together with Article 2(1), satisfies the test in Case 12/86 Demirel v Stadt Schwabisch Gmund [1987] ECR 3719 for direct effect.
2. The Charter of Fundamental Rights does not provide free-standing justiciable rights in Northern Ireland post-Brexit. It applies only where Member States are “implementing Union law” within the meaning of article 51(1), and the engagement of a mere EU competence is not enough.
3. An ab ante (prospective) human rights challenge to the architecture of the ICRIR fails on the high “all or almost all cases” threshold. The Court held that it will rarely be possible to determine whether a particular ICRIR review satisfies the Article 2/3 investigative obligation until the review has been completed.

The Supreme Court did not disturb the Human Rights Act declarations concerning the immunity provisions, section 8, section 43(1), or the Court of Appeal’s additional declarations concerning section 43(2) and section 45. Those matters were not live before it: the Secretary of State had abandoned the appeal against the High Court declarations before the Court of Appeal gave judgment, and did not appeal the Court of Appeal’s further declarations concerning section 43(2) and section 45 to the Supreme Court. (The judgment also records that the Court of Appeal’s inclusion of section 40 in the immunity declaration was a typographical error, which the Supreme Court corrected.)

2. The Windsor Framework: a deliberate narrowing

(a) The “container” theory rejected

Perhaps the central doctrinal move in Dillon is the rejection of what one might call the “container theory” of Article 2(1) of the Windsor Framework. Counsel for the applicants, Mr Larkin KC, argued that the RSEO chapter functioned as a container into which any EU-derived right operative on 31 December 2020 could be poured. On that view, the non-diminution obligation could attach to any such right without the need to demonstrate that the right itself was “set out” in the RSEO chapter with the specificity required for direct effect.

The Supreme Court was unequivocal in rejecting that approach. Reading Article 2(1) in conjunction with the RSEO chapter, the Court held that “[i]t is necessary to identify a clear and precise obligation in EU law by reference to the RSEO chapter. It is only in this way that the obligation not to diminish rights, safeguards or equality of opportunity acquires any content.” The high-level affirmations in paragraphs 1, 11 and 12 – “civil rights”, “the suffering of the victims of violence”, a “right to remember” – are simply “too general to give rise to any directly enforceable rights in relation to inquests, civil proceedings or prosecutions.”

Importantly, the Court did not hold that Article 2(1) is never directly effective. It expressly recognised that the provision may operate in conjunction with a directly effective provision of one of the Annex 1 Directives, or with another EU instrument falling within the ambit of an RSEO right, provided the Demirel requirements are satisfied in respect of the obligation imposed. The operative question for future Article 2(1) cases will therefore be whether the EU instrument in question falls “within the ambit of” one of the bullet-point rights in paragraph 1 of the RSEO chapter (or, less clearly, paragraphs 11 or 12). The phrase “within the ambit of” is borrowed (consciously or otherwise) from the article 14 ECHR jurisprudence and is famously elastic; further litigation as to its meaning in the Windsor Framework context is almost certain.

(b) The Victims Directive: a closing of the door

The applicants relied on articles 11 and 16 of Directive 2012/29/EU (the Victims Directive) – respectively the right to review of a decision not to prosecute and the right to a decision on compensation from the offender in criminal proceedings. The Court accepted, in principle, that Article 2(1) of the Windsor Framework could operate in conjunction with directly effective EU instruments to generate enforceable rights. But it held that the Victims Directive could not bear the weight the applicants placed on it, because:

• Articles 11 and 16 are concerned with the conduct of individual prosecutions and the exercise of prosecutorial discretion in particular cases. They do not engage with whether, as a matter of policy, prosecutions should be available at all for an entire class of cases;
• A legislative decision by Parliament to confer immunity on a class of offences is not a decision “taken by a prosecutor” within the meaning of Recital 43;
• The ICRIR, not being a prosecutorial authority, cannot itself engage the Directive.

This is a doctrinally clean answer, but its practical reach is significant. It strongly suggests that the Victims Directive will not usually assist where the challenge is to a broad legislative policy decision about whether prosecutions may be pursued, as distinct from a decision by a prosecutorial or investigative authority in an individual case.

(c) Subsequent practice and the limits of treaty interpretation

A subsidiary but important strand of the judgment concerns the role of “subsequent practice” under article 31(3)(b) of the Vienna Convention on the Law of Treaties 1969 as an aid to interpretation. The applicants and the Equality Commission relied on a series of UK Government statements (the August 2020 “Explainer”, a Walker letter to Professor McCrudden, Lord Duncan’s written answer of 28 January 2020) which suggested that Article 2(1) was capable of direct effect generally. The Court dismissed these as “unilateral” statements that “cannot be considered to be subsequent practice in the application of the treaty which establishes the agreement of the parties regarding its interpretation.”

For practitioners, the practical point is sharp: government policy documents and Ministerial assurances about the scope of post-Brexit instruments will carry limited interpretive weight before the courts, unless they reflect a mutually agreed position with the EU. The Court reinforced that the question of direct effect is governed by EU law and the Demirel test, and that practice cannot rewrite that test.

(d) Constitutional significance

The constitutional importance of this aspect of the decision is hard to overstate. The Northern Ireland courts’ interpretation of the Windsor Framework had exposed Acts of Parliament to potential disapplication on Article 2(1) grounds, including where EU-law rights were said to overlap with Convention-type protections in the legacy context. The Northern Ireland High Court had also used Article 2(1) to disapply legislation in other subject areas, including immigration. The Supreme Court has now narrowed the broader, freestanding use of Article 2(1) – though, importantly, it has not closed it off entirely. Where a sufficiently precise EU instrument can be identified, falling within the ambit of an RSEO right, disapplication of primary Westminster legislation remains a live possibility in Northern Ireland. This continues to mark Northern Ireland out as a jurisdiction in which Parliamentary sovereignty operates under a unique constitutional qualification.

In this author’s view, the decision can be read as part of a discernible pattern in which the courts, faced with broad rights commitments made by the UK Government to the people of Northern Ireland in the post-Brexit settlement, have construed those commitments more narrowly than the political rhetoric surrounding their adoption might have suggested. Whether one accepts that characterisation or not, the practical scope of Article 2(1) has plainly narrowed.

3. The Charter cross-appeal: no free-standing rights

The applicants’ cross-appeal that the Charter of Fundamental Rights of the EU should apply on a free-standing basis was given short shrift. The reasoning is orthodox EU law but its application here is consequential.

The Court reaffirmed the “anchor” requirement: the Charter applies to Member States only when they are “implementing Union law” (article 51(1) of the Charter), as construed in Åkerberg Fransson (Case C-617/10) and consistently applied since. The applicants’ reliance on CG v Department for Communities in Northern Ireland (Case C-709/20) for the broader proposition that the Charter is applicable “in all situations governed by EU law” was rejected as a misreading: in CG there were two clear “anchors” – Union citizenship and the freedom of movement under article 21 TFEU. As the Court put it, CG “can be explained as an application of the ‘settled case law’ on the Charter’s scope to the specific facts, involving both Union citizenship and EU free movement rights and their implementation.”

The Court was also notably wary of the constitutional consequences of the applicants’ position. To allow Convention rights, repackaged as Charter rights, to be enforced via Article 2(1) of the Windsor Framework would, on the Secretary of State’s submission (which the Court endorsed), permit the “limited language of article 2(1)’s reference to the diminution of rights” indirectly to:

“…[lead] via a reading into the RSEO chapter of the [Belfast Agreement] of the entirety of the [Charter] (which did not exist at the time), a general and free-standing source of civil, political, economic and social rights accompanied by the most powerful remedial consequences known to domestic law (namely the disapplication of primary legislation of the sovereign Parliament).”

Such a result, the Court held, would “surely require the clearest possible statutory language” — and there was none.

The practical significance is straightforward: the Charter cannot be used in Northern Ireland as a freestanding source of rights through Article 2(1) of the Windsor Framework. It remains relevant only where an applicable EU-law provision supplies the necessary anchor, in which case it continues to operate as an aid to interpretation.

4. The ab ante challenge: the high “all or almost all cases” threshold

(a) Setting the threshold

For practitioners considering pre-emptive challenges to legislative schemes, the Dillon judgment provides the most detailed statement to date of how the ab ante test from R (Bibi) v Secretary of State for the Home Department [2015] UKSC 68, Christian Institute v Lord Advocate [2016] UKSC 51 and In re Abortion Services (Safe Access Zones) (Northern Ireland) Bill [2022] UKSC 32 (“Safe Access Zones“) operates in the procedural article 2/3 context.

The applicants had to establish that the legislative scheme would give rise to a failure to comply with one or more elements of the investigative obligation in “all or almost all cases”. The Court emphasised the dual requirement: it is not enough to show that one or more of the essential parameters of an effective article 2/3 investigation might be breached. The challenger must show both (i) that the breach will occur in all or almost all cases, and (ii) that, even on that assumption, the investigation taken as a whole will be ineffective in all or almost all cases.

That dual requirement, married to the granular, case-by-case nature of the Tunç v Turkey [2016] Inquest LR 1 “essential parameters” analysis, makes ab ante challenges to investigative schemes very difficult indeed. The Court held that it will rarely be possible to determine whether a particular ICRIR review satisfies the Article 2/3 investigative obligation until the review has been completed. That guidance is likely to be cited frequently in future. The practical advice to claimants will increasingly be to wait for a concrete case in which the asserted defect actually crystallises.

(b) Reframing the comparator

A subtle but important move is the Court’s rejection of the Court of Appeal’s tacit comparison between ICRIR reviews and legacy inquests. The Court of Appeal had relied heavily on its “collective experience” of legacy inquests, in which legally aided lawyers cross-examined witnesses, and concluded that the absence of equivalent provision in the 2023 Act militated against effectiveness. The Supreme Court held that this approach was wrong in principle. The article 2/3 obligation does not require an adversarial procedure with publicly funded representation and cross-examination. Police Ombudsman investigations – wholly inquisitorial, with no hearings – have been treated as capable of discharging the procedural obligation: in Dalton v United Kingdom (2025) 81 EHRR SE5 the Strasbourg Court considered that any Article 2 duty had been satisfied by the Police Ombudsman investigation in conjunction with the ongoing civil proceedings against the police (a point reiterated in In re Dalton [2023] UKSC 36). The proper comparator is the Convention itself, not the system the new scheme replaces.

This will matter beyond legacy. Where Parliament replaces an adversarial procedure with an inquisitorial alternative, the question is not whether the new procedure mimics the safeguards of the old, but whether – assessed jointly across all the Tunç “essential parameters” – it is capable of producing an effective investigation.

(c) The “security veto”: Wiley resurrected

The Court’s treatment of the so-called “security veto” in Schedule 6, paragraph 4 of the 2023 Act is perhaps the most consequential element of this part of the judgment for ongoing practice.

The Court of Appeal had characterised the Secretary of State’s power to prohibit ICRIR disclosure of sensitive information as “an effective veto” – placing the final say in the hands of the Secretary of State. The Supreme Court disagreed, in terms which substantially reshape the operation of paragraph 4. The Secretary of State is held to be subject to:

• The ordinary rationality, evidence and relevancy requirements of public law;
• A Tameside duty (per Secretary of State for Education and Science v Tameside Metropolitan Borough Council [1977] AC 1014) to inquire (including obtaining the ICRIR’s views on the effect of non-disclosure on the administration of justice);
• An obligation to carry out the Wiley balancing exercise (R v Chief Constable of West Midlands Police, Ex p Wiley [1995] 1 AC 274) – balancing the public interest in national security against the public interest in the administration of justice;
• The article 2/3 procedural obligation under section 6(1) of the Human Rights Act, breach of which renders any decision unlawful; and
• Judicial review supervision under the duty of candour, with the court having the final say on whether the Secretary of State’s decision should be quashed (rather than on the underlying merits of disclosure).

The substantive move here is less novel than it might first appear. Implying ordinary public law duties – rationality, Tameside, the duty of candour, section 6 HRA – into the exercise of a statutory power is the default position in administrative law, not a feat of judicial ingenuity. The more substantive element is the importation of the Wiley balance into a statutory regime which, on its face, frames the Secretary of State’s prohibition power solely around his or her view of the risk to UK national-security interests. That step does meaningful work: it converts what might have been read as a relatively unstructured ministerial determination into a structured public-law decision-making exercise in which the public interest in the administration of justice must be weighed alongside national security.

Whether this judicial reconstruction will hold up when tested in concrete disclosure disputes remains to be seen. In this author’s view, the practical effect is to shift the locus of scrutiny: rather than upholding a declaration of incompatibility, the Court has placed significant weight on subsequent judicial review proceedings to give the Wiley-structured regime its bite. Whether that proves sufficient will depend on how robustly the courts apply the framework in concrete disclosure disputes.

(d) The independence challenge

The applicants’ challenge to the ICRIR’s institutional independence in relation to disclosure decisions was rejected on essentially the same basis as the disclosure challenge. Drawing on the Court’s earlier analysis in In re McQuillan [2021] UKSC 55, the Court reiterated that the Strasbourg case law does not require “absolute independence” or “complete hierarchical or institutional disconnection” — what is required is sufficient independence, judged in light of the circumstances of the specific case. The Secretary of State’s role in disclosure is structurally constrained, and the courts retain supervisory jurisdiction. Moreover, decisions about national security are constitutionally and institutionally an executive matter (citing In re Secretary of State for Northern Ireland [2025] UKSC 47): it is “constitutionally impossible to remove the executive from a decision-making process involving national security.”

5. The reconciliation exception: an obiter sketch

Although the Secretary of State’s abandonment of the immunity appeal in the Court of Appeal meant the question did not fall to be decided, the Supreme Court took the opportunity – at the invitation of the Northern Ireland Veterans Movement – to consider whether the Strasbourg Court recognises a “reconciliation exception” to the general rule that breaches of articles 2 and 3 must be prosecuted.

The analysis (Dujardin v France (Application No 16734/90) (unreported) 2 September 1991, Tarbuk v Croatia (Application No 31360/10) (unreported) 11 December 2012, Ould Dah v France (2009) 56 EHRR SE17, and Marguš v Croatia (2014) 62 EHRR 17) is careful and important. The Court’s conclusion is finely balanced: the Strasbourg jurisprudence “leave[s] open the possibility that there may be circumstances in which the Strasbourg court could recognise an exception to the general rule … but it has not done so to date.” Applying the so-called “mirror principle” from R (Ullah) v Special Adjudicator [2004] UKHL 26 and R (AB) v Secretary of State for Justice [2021] UKSC 28, the Court declined to develop the law in advance of Strasbourg.

The implications for the political process are real but should not be overstated. After Dillon, the compatibility of a conditional immunity or amnesty scheme has not been finally determined by the Supreme Court or Strasbourg. But domestically, no reconciliation exception has been recognised, and the declarations of incompatibility concerning the 2023 Act’s immunity provisions remain in place. The Court left open only the possibility that Strasbourg might in a future case recognise some form of reconciliation exception. It did not formulate or endorse a domestic test under which a future amnesty would be lawful.

In practice, this means that the Government’s forthcoming Northern Ireland Troubles Bill cannot be assumed to be the only Convention-compatible answer; equally, those advocating for a return to conditional immunity cannot point to Dillon as judicial cover. The Court has been careful to leave the substantive question to be resolved elsewhere.

6. Where the decision sits against earlier authority

The case can usefully be read as the latest in a line of decisions in which the Supreme Court has resisted broader readings of the Belfast Agreement-derived constitutional architecture:

• In re Allister [2023] UKSC 5 – the Supreme Court upheld the constitutional propriety of the Northern Ireland Protocol against challenges based on the Acts of Union. Dillon now narrows the reverse-facing argument: even where Article 2(1) is engaged, its effect is more constrained than the Northern Ireland courts had suggested.
• In re McQuillan [2021] UKSC 55 – the Court’s careful articulation of the article 2/3 investigative obligation is the doctrinal scaffolding on which Dillon‘s treatment of independence and effectiveness rests.
• In re Dalton [2023] UKSC 36 and Dalton v United Kingdom (2025) 81 EHRR SE5 – confirm that the procedural obligation can be discharged by non-curial mechanisms (notably Police Ombudsman investigations, in conjunction with other available proceedings such as civil claims). Dillon draws explicitly on this to deny the Court of Appeal’s implicit assumption that an inquest is the gold standard.
• Safe Access Zones [2022] UKSC 32, Bibi [2015] UKSC 68, Christian Institute [2016] UKSC 51, In re JR123 [2025] UKSC 8 – the ab ante line. Dillon applies the doctrine in a context where, perhaps for the first time, the Supreme Court has expressly mapped the “all or almost all cases” test onto the multi-factorial Tunç essential parameters.
• R (AB) v Secretary of State for Justice [2021] UKSC 28 – the most recent restatement of the Ullah mirror principle. Dillon applies it cautiously to deny a “reconciliation exception” without ruling it out forever.
• In re Secretary of State for Northern Ireland [2025] UKSC 47 – the Court’s recent restatement of the Wiley principles in a national security PII context. The reading of paragraph 4 of Schedule 6 in Dillon effectively imports Wiley into a statutory disclosure regime that does not, on its face, require it.

7. Practical implications

For practitioners advising victims and families

• Pre-emptive challenges are now much harder. The combination of the ab ante threshold and the Court’s strong steer that effectiveness is rarely assessable in advance means challenges to forthcoming investigative schemes will generally need a concrete case.
• Concrete disclosure decisions are now the litigation forum. The Court has redrawn the statutory disclosure regime as a structured Wiley exercise subject to judicial review. Practitioners should be alive to the Tameside obligations the Court has read into paragraph 4 of Schedule 6 – including the requirement to seek the ICRIR’s views on the impact of non-disclosure on the administration of justice. Decisions made without those steps will be vulnerable.
• Legal representation in ICRIR reviews remains an open issue. The Court has not held that legal aid is required across the board, but it has not closed off case-specific challenges. A claim made on the basis of a particular review (rather than the scheme as a whole) is still arguable.

For Government and the ICRIR

• The judgment is more an organising framework than a vindication. Although the Secretary of State largely won, the Court has subjected the disclosure regime to a substantially heavier procedural and substantive burden than the statutory text suggests on a literal reading.
• The forthcoming Troubles Bill must be drafted with care. The Court’s narrowing of Article 2(1) of the Windsor Framework does not eliminate the risk that future legacy legislation could fall within it: if any future statute touches an EU instrument that does meet Demirel and falls within the ambit of an RSEO right, disapplication remains on the table.

For constitutional and public law

• Article 2(1) of the Windsor Framework remains a live disapplication doctrine – but a narrower one. Practitioners across all subject areas (immigration being the most obvious) should expect renewed argument as to what falls “within the ambit of” the bullet-point rights in paragraph 1 of the RSEO chapter, and as to whether a particular EU instrument provides the necessary anchor.
• The Charter is no longer available as a free-standing source of rights in Northern Ireland. It survives only as an interpretive aid where there is an implementing EU measure.
• Subsequent practice arguments under the Vienna Convention will face stiff headwinds where the practice relied upon is unilateral.

For victims’ organisations and the public

• The declarations of incompatibility relating to the immunity provisions, section 8, and section 43 stand – but they do so because they were not the subject of a live appeal, not because the Supreme Court has affirmed the underlying reasoning. The doctrinal foundation is, in that sense, untested at this level.
• The Government’s policy choices remain its own. The Supreme Court has not given Ministers cover for any particular policy on legacy; it has cleared away some of the legal undergrowth but the substantive political and moral choices remain to be made.

8. Concluding thoughts

Dillon is a difficult case to characterise. In one sense it is a careful, technically orthodox application of established doctrines – the Demirel test, Åkerberg Fransson, Tunç, the ab ante line, Ullah. In another sense it is a quietly significant work of constitutional reconstruction. Article 2(1) of the Windsor Framework has been redrawn so as to be operable but constrained. The Charter has been firmly placed back in its box. The “security veto” in Schedule 6 has been judicially re-engineered as a structured Wiley exercise. And the ab ante doctrine has been emphatically reaffirmed as a high – perhaps very high – hurdle for prospective challenges to investigative regimes.

For solicitors and barristers, the immediate task is to absorb the redrawn Article 2(1) framework, the renewed Wiley analysis in the disclosure context, and the case-by-case orientation the Court has imposed on the ICRIR challenge. For Masters students and academics, Dillon offers a remarkably rich seam: the relationship between the Belfast Agreement and EU law, the survival of the Charter, the reach of the Ullah principle in evolving Strasbourg territory, and the constitutional propriety of judicial re-reading of statutory ministerial powers.

For victims of the Troubles and those affected by the 2023 Act, the position is unhappier. Some declarations of incompatibility survive – but for procedural reasons rather than because the Supreme Court has endorsed them. The principal substantive question, whether conditional immunity is compatible with articles 2 and 3 of the Convention, remains formally open. The repeal-and-replace Bill will move that question into a new political arena, but the legal question is not gone. It has merely been deferred.

What is certain is that Dillon will be the gateway authority on the Windsor Framework, on the Charter in post-Brexit Northern Ireland, and on the limits of ab ante challenges, for some time to come.

Articles and content on this site are for informational purposes only and do not constitute legal advice. Do not rely solely on this information.