AXA Insurance UK PLC and another v Commissioners of Inland Revenue and another [2026] UKSC 24

AXA group companies, follower claimants in tax group litigation, sought to rely on a Prudential test case decision later shown to be legally wrong. The Supreme Court held that courts may, exceptionally, order that such a decision does not bind follower claims, and dismissed the appeals.

Background

The appeal concerned the Controlled Foreign Companies and Dividend Group Litigation Order (the GLO), made in July 2003. The claims on the group register alleged that UK tax provisions treating dividends from non-UK companies less favourably than dividends from UK-resident companies were incompatible with EU law. The AXA proceedings were issued on 8 April 2003, entered on the register and stayed while three test cases proceeded. The most relevant of these was the Prudential test case, finally disposed of by the Supreme Court on 25 July 2018.

The appellants were two claimants in the AXA proceedings:

• AXA Insurance UK plc (AXAIUK) claimed restitution of corporation tax paid on portfolio dividends from non-UK companies. Its claims went back to the accounting period ending 31 December 1995, so for periods before 8 April 1997 it needed to rely on section 32(1)(c) of the Limitation Act 1980.
• Guardian Royal Exchange Assurance plc (GREA) claimed compensation for the time value of unlawfully levied advance corporation tax (ACT) that had been set off against lawful mainstream corporation tax before proceedings began (utilised ACT). This was the claim for the prematurity period. GREA sought simple interest under section 35A of the Senior Courts Act 1981, agreed at £5,645,286.72. The Revenue said GREA was confined to the materially smaller statutory remedy under section 85 of the Finance Act 2019.

In Prudential HC (Henderson J, judgment of 24 October 2013; order of 28 January 2014), the Revenue conceded, on the strength of Sempra Metals HL, that compound interest was payable on utilised ACT for the prematurity period (category (a)). Henderson J extended compound interest to categories (b) and (c). That legal landscape was later overturned:

• Littlewoods SC held that EU law did not require compound interest.
• Prudential SC rejected the Sempra Metals unjust enrichment analysis. It said that Prudential’s category (a) claim would also have been rejected had the Revenue not conceded it.
• FII SC 3 held that there is no restitutionary claim at common law for the prematurity period, so claimants are confined to section 85 of the Finance Act 2019.
• FII SC 2 overruled DMG on when time starts to run under section 32(1)(c). Time now runs from when the claimant could with reasonable diligence recognise that a worthwhile claim arises, not from the date of an authoritative judicial ruling.

Richards J held that neither issue had been decided as a GLO issue in the Prudential test case. The Court of Appeal ([2024] EWCA Civ 1430) took a different view on the Set Off Issue. It held that this issue had been decided as a GLO issue, but exercised its power under CPR r 19.23(1)(a) to order that the decision did not bind GREA’s claim. It agreed with Richards J that the Limitation Issue had not been decided as a GLO issue.

The issues

The Set Off Issue

Should the court exercise its power to “order otherwise” under CPR r 19.23(1)(a), so that the Prudential HC decision recognising a common law claim for interest over the prematurity period did not bind GREA’s follower claim?

The Limitation Issue

Did Prudential HC decide, as a GLO issue binding on follower claims, the date from which the limitation period ran under section 32(1)(c) on the DMG test?

CPR r 19.23(1)(a) provides:

“(1) Where a judgment or order is given or made in a claim on the group register in relation to one or more GLO issues— (a) that judgment or order is binding on the parties to all other claims that are on the group register at the time the judgment is given or the order is made unless the court orders otherwise”

The parties’ arguments

The appellants

Mr Bremner KC argued for a very narrow reading of the power to order otherwise. His primary submission was that the rule should be read as permitting only a limitation contained in the test case judgment or order itself. He gave four examples of appropriate use:

• advance orders for sub-groups that differ from the test claimant;
• unsuitability of the test case becoming apparent during or after the hearing;
• procedural unfairness or error in the test case;
• an overbroad ruling.

He further contended that:

• the power should not be used to allow points of law to be re-argued;
• the Revenue had made a tactical choice not to appeal category (a);
• it was unfair that Prudential kept the benefit of the decision while follower claimants did not;
• follower claimants had contributed to Prudential’s common costs.

On limitation, the appellants said Henderson J had determined that time did not run before 12 December 2006 (the date of FII CJEU 1), or at least not before 14 July 1998.

The Revenue

The Revenue stressed the court’s duty to apply the law correctly. It drew an analogy with the exception to issue estoppel recognised in Arnold v National Westminster Bank plc. On limitation, it argued that Prudential HC had made no determination of the start date at all, so CPR r 19.23(1)(a) was not engaged.

The Court’s reasoning

The purpose of the GLO regime

Lady Rose and Lord Richards (with whom Lord Hodge and Lord Hamblen agreed) identified the overall purpose of the regime as facilitating:

“the economic and efficient disposal of claims raising the same issues, for the benefit of the parties and, indeed also, for the benefit of the court, its funding and its other users, and thereby to facilitate access to justice.”

Scope and timing of the power to order otherwise

The court rejected the argument that the power could only be exercised within the test case order itself. It held that the court may make such an order in the follower proceedings after the test case has been disposed of. Nor was the power confined to cases where the test case proved unsuitable.

The court nonetheless emphasised that the discretion is not wide:

“having regard to the importance of not undermining the efficacy of the GLO regime, there will need to be some exceptional quality about the circumstances before it becomes appropriate for the court to consider whether to exercise the power.”

The issue estoppel analogy

The court found the issue estoppel cases helpful but treated the analogy as imperfect. Issue estoppel does not raise the additional concern of wider repercussions for the GLO regime. Two key considerations therefore had to be balanced:

• the integrity of the GLO regime and the binding force of test cases; and
• a party’s right to have its dispute decided according to a correct understanding of the law.

On the second consideration, the court said:

“it will ordinarily be unjust for a person to have their rights determined otherwise than in accordance with the right legal rules. In general, it would be contrary to the rule of law for a court to make an order imposing an obligation on a party to proceedings when it is clear that, as at the date that the order is made, there is no legal basis for imposing such an obligation.”

The threshold for disapplying a test case decision

From this balance the court derived a stringent threshold, stricter than in issue estoppel:

“a challenge to the binding effect of the GLO test case judgment can only be mounted where it has been clearly established that the legal basis for that judgment was wrong. It is not enough that the follower claimant or the unsuccessful defendant in the test case wishes to raise different—potentially better—arguments on the law than were raised and decided in the test case.”

Doubts expressed in later dicta would not suffice. Test case decisions also bind follower claims whether they result from full argument or from concession. Any unfairness this might cause is removed by the extended appeal rights in CPR r 19.23(2).

Factors once the threshold is met

Even where the threshold is met, the court must still weigh whether ordering otherwise would be unjust. Relevant factors include:

• whether disapplying the test case would impose lengthy and costly relitigation;
• any particular unfairness to follower claimants whose claims were stayed, balanced against the advantages they enjoy as followers;
• any tactical manipulation by the party seeking the order, for example through concessions made or points not pursued on appeal.

Application to the Set Off Issue

The Revenue accepted that the Set Off Issue was a GLO issue and that GREA’s claim was indistinguishable from Prudential’s. The court nonetheless held the stringent criterion met because of:

“the extraordinary nature of the developments of the law in the 20 years since these proceedings were started.”

Both planks of the Prudential HC decision had been definitively held wrong by the Supreme Court. Because judgments are declaratory, no legal basis had ever existed for them. The court found no countervailing factor:

• No relitigation. The law was settled by Supreme Court authority:
  

  “There is no question of a second bite at the cherry.”

  This made the case stronger than Arnold, where the construction point remained to be argued.

• No abuse in the Revenue’s conduct. Drawing on FII SC 3, the court held that the decision not to challenge category (a) was not abusive. It concluded:
  

  “It would be a serious injustice for the Revenue if they now have to meet claims valued at many millions of pounds when there is no legal basis for them.”

• Inconsistency with Prudential was not decisive. Differing outcomes between test and follower claims are an inevitable result of the power to order otherwise. Complete consistency across the register is not assured in any event, given CPR r 19.23(1)(b).
• Costs did not tip the balance. The appellants’ costs contributions did not outweigh the injustice of a substantial liability without legal basis. AXA would also have recovered its share of Prudential’s costs paid by the Revenue on the standard basis.

The Limitation Issue

Following the decision on the Set Off Issue, GREA’s utilised ACT claim did not involve a mistake of law cause of action. Section 32(1)(c) therefore mattered only for AXAIUK.

Paragraph (B) of the original GLO issues (“from what date does the applicable limitation period start to run?”) was capable of covering section 32(1)(c). However, no section 32(1)(c) reply had been pleaded in Prudential. The agreed trial list asked only to what extent the claim was barred by the six-year period and by section 320 of the Finance Act 2004.

The court warned against hindsight when interpreting Henderson J’s judgment and order:

• Under DMG, the date of reasonable discovery would have been the same for everyone.
• The only date Henderson J mentioned, 14 July 1998, was specific to Prudential Holborn Life’s joinder.
• His reference to December 2006 at para 257 was an assumption expressed as “probably”.
• The bald declaration that the claims were “in time” mentioned no date.

The court agreed with the Court of Appeal that the date of discoverability “was simply not an issue”. It added that agreed or conceded answers can bind, but:

“A party cannot attempt to keep their powder dry by conceding something which is then decided against them in the test case in the hope of arguing it out in full in a different case on the register. But that is not what happened here.”

AXAIUK’s claim is therefore governed by FII SC 2. The parties might be able to agree a start date in the light of BAT Industries.

Lord Leggatt (concurring)

Lord Leggatt agreed that the appeals should be dismissed, but on a narrower footing. He placed the rule within the wider context of multi-party procedures. He also emphasised that the binding effect of a test case judgment is the quid pro quo of the right of follower parties to seek permission to appeal under CPR r 19.23(2). He agreed that the rule contains no time limit for an application to order otherwise, contrasting rule 18 of the First-tier Tribunal (Tax Chamber) Rules.

He did not find the Arnold analogy persuasive. In his view, a GLO concerns parallel claims whose causes of action have already arisen:

“If the law later changes, it is in principle unjust to permit the parties to some, but not all, materially identical claims to reopen the common issue. To allow this is contrary to the underlying precept of one for all, all for one.”

He stated expressly:

“I have explained why the fact that there has been a material change in the law since the Prudential HC judgment was given is not enough to justify making an order otherwise.”

In his view an order was justified here only because neither purpose of the binding rule (efficiency and consistency) was undermined:

• No further argument was needed to decide the issue.
• Assessing GREA’s category (a) interest on the Sempra Metals basis, while its category (b) and (c) claims were assessed on a different basis, would be irrational and incoherent. This was all the more so because the category (c) principal derives from category (a).

He described this palpable irrationality as:

“far more offensive to the basic principle that like cases should be decided alike than the inconsistency involved in assessing such interest on a different basis from Prudential’s category (a) claim”

He noted that the Revenue’s delay would normally be a compelling factor against an order otherwise. However, it was excused in this exceptionally complex litigation, and no serious argument of unreasonable delay had been advanced.

Practical significance

• The court may disapply a test case decision after the event. The decision confirms that the court may order, after a test case has concluded, that the test case decision does not bind follower claims on the register.
• The threshold is exceptionally high. Where the challenge rests on a later change in the law, the legal basis of the test case must have been clearly established to be wrong. Better arguments or doubting dicta will not suffice. Even then, the court must weigh relitigation, prejudice to followers, tactical conduct, costs and delay.
• Concessions bind follower claims. Concessions in test cases bind follower claims, discouraging parties from keeping arguments in reserve.
• Careful definition of GLO issues matters. Whether an issue was actually decided as a GLO issue turns on a hindsight-free reading of the pleadings, the agreed issues, the judgment and the order. This underlines the need to define and plead GLO issues carefully.
• A divergence of emphasis among the Justices. Lord Leggatt’s concurrence treats a change in law, by itself, as insufficient and puts consistency at the centre. Practitioners should note this difference in emphasis about how rarely, and on what basis, the power should be exercised.
• Practical outcome for these claims. GREA is confined to the statutory remedy under section 85 of the Finance Act 2019. AXAIUK’s limitation position falls to be determined on the FII SC 2 test.

Verdict: The appeals were dismissed unanimously on both issues. On the Set Off Issue, the Supreme Court upheld the Court of Appeal’s order under CPR r 19.23(1)(a) that the Prudential HC decision recognising a common law claim to interest for the prematurity period is not binding on GREA’s claim, so GREA is confined to the statutory remedy under section 85 of the Finance Act 2019. On the Limitation Issue, the court held that Prudential HC did not decide the start date of the limitation period under section 32(1)(c) as a GLO issue, so AXAIUK’s claim is governed by the current law in FII SC 2. Lord Leggatt concurred, giving separate reasons.

Source: AXA Insurance UK PLC and another v Commissioners of Inland Revenue and another [2026] UKSC 24