Section 68(4) as a One‑Tier “Gatekeeper”: The Court of Appeal Has No Permission Jurisdiction Where the High Court Refuses Section 68 Relief on the Merits

A) Precedents Cited

i) K1 v B [2025] EWHC 2539 (Comm)

This was the first-instance judgment under challenge. Mr Justice Robin Knowles refused the proposed amendment, holding the pleaded “illegality/unenforceability” type complaint did not fall within s.68(2)(g). The Court of Appeal’s decision does not revisit that merits conclusion; it treats it as the relevant “decision under” s.68 for the purpose of s.68(4).

ii) Federal Republic of Nigeria v Process & Industrial Developments Limited [2023] EWHC 2638 (Comm)

The High Court had applied principles from this decision when construing s.68(2)(g), emphasising process-focused scrutiny: the sub-paragraph targets irregularities in how an award is obtained (e.g., fraud in the arbitral process), not simply that the award enforces an allegedly illegal contract. In the Court of Appeal, this mattered indirectly: it framed the High Court’s refusal as a merits application of s.68, rather than an absence of jurisdiction.

iii) Henry Boot Construction (UK) Ltd v Malmaison Hotel (Manchester) Ltd [2001] QB 388

Cited for the policy foundation of the 1996 Act: finality, speed, and minimal court intervention (linked to s.1(a)). This policy underpins the “one-tier” permission model, whereby the first-instance judge acts as the sole gatekeeper for appeals in specified arbitration-related decisions.

iv) Cetelem SA v Roust Holdings [2005] EWCA Civ 618

A key authority on the meaning of “decision under this section” in a similar leave-restriction provision (s.44(7)). Clarke LJ distinguished between:

• decisions that are wrong on law/fact but made within the court’s statutory power (the judge “should not” have made them), which remain “under the section” and are therefore caught by the permission bar; and
• decisions that are outside the statutory power (the judge “could not” have made them), which are not fairly treated as “under” the section and can escape the bar.

K1 v B (No. 2) uses Cetelem as the conceptual comparator but holds that K1’s case falls on the “should not” side: interpreting the scope of s.68(2)(g) is part of the merits of a s.68 application, not an instance where the High Court made an order it lacked power to make.

v) Sumukan Ltd v Commonwealth Secretariat [2007] EWCA Civ 243

Considered a different leave restriction (s.69(6)) and the special statutory notion that parties may “exclude” s.69. Waller LJ treated the existence of an exclusion agreement as a threshold issue: if parties have excluded s.69, the court’s jurisdiction under that section is not engaged at all.

In K1 v B (No. 2), the Court distinguishes Sumukan: s.68 does not contain an analogous “exclusion agreement” gateway. Whether a case falls within s.68(2) and whether there is substantial injustice are the substantive merits, not a separate jurisdictional precondition akin to party exclusion of the court.

vi) National Iranian Oil Company v Crescent Petroleum Co International Ltd [2023] EWCA Civ 826

Highly influential in the Court’s approach to the policy and breadth of the leave restriction (there, s.67(4)). Males LJ emphasized:

• the policy of having the first-instance court as the sole gatekeeper, avoiding delay/expense and leaving awards “in limbo”;
• that decisions “part of the process” and “within the compass” of s.67/s.68 are caught by the leave restriction; and
• that s.73 (loss of right to object) is ancillary and not itself a jurisdiction question.

K1 v B (No. 2) applies this logic: deciding whether s.68(2)(g) is engaged is “overseeing the arbitration process” and is within s.68’s compass—therefore caught by s.68(4).

vii) ASM Shipping Ltd v TTM Ltd [2007] 1 Lloyd's Rep 16

Cited via NIOC for Longmore LJ’s proposition that refusing relief for whatever reason is still “a decision under” s.68, just as granting relief would be. This supports treating threshold or dispositive reasons within the s.68 framework as still falling under the statutory permission restriction.

viii) North Range Shipping Ltd v Seatrans Shipping Corp [2002] EWCA Civ 405; CGU International Insurance Plc v Astrazeneca Insurance Co Ltd [2006] EWCA Civ 1340

These authorities identify narrow exceptions to the leave restriction, including where a purported decision is not a decision at all or where there has been a fair trial infringement. The Court notes these exceptions but finds none applicable.

ix) McDonald v Rose [2019] EWC Civ 4

Cited regarding the procedural function of “time runs from” wording in consent orders when judgments are handed down before permission-to-appeal is determined. It supports the Court’s conclusion that the consent order was only about timing, not about creating any entitlement to seek permission from an appellate court.

x) Osler v Osler [2024] EWCA Civ 516

Used to reinforce the principle that statutory limitations on appellate routes cannot be bypassed even where a higher court has (purportedly) given leave: a fortiori, they cannot be bypassed by consent order or case management language.

xi) A1, A2, A3 v P [2025] EWHC 3372

Provides procedural context: the s.67 challenge ultimately succeeded for two parties but failed for K1 (there “A2”), leaving K1 as the sole applicant in this renewed permission application. It does not affect the jurisdictional conclusion under s.68(4).

xii) Department of Economic Policy and Development of the City of Moscow v Bankers Trust Company [2004] EWCA Civ 314

The Court relied on this for principles governing confidentiality/privacy orders. It accepted that publication of an anonymised judgment was a proportionate way to reflect public interest while protecting arbitral confidentiality, especially given that the Court lacked jurisdiction to entertain the substantive appeal.

B) Legal Reasoning

i) Statutory structure: “the court” and the leave restriction

The starting point is the 1996 Act’s deliberate limitation of appeals in arbitration matters. Section 68(4) requires leave for any appeal “from a decision of the court under this section”; and s.105 defines “the court” as the High Court or County Court. Thus, where s.68(4) applies, the Court of Appeal cannot act as an alternative permission granter.

ii) Consent order cannot vary statute (and s.68 is mandatory)

K1 argued that wording in the consent order (approved after the High Court judgment) effectively preserved a right to apply to the Court of Appeal for permission. The Court rejected this on several layers:

• The language was about when time starts, not about whether a right exists.
• No application was made to vary the statutory scheme; and even if it had been, the judge could not confer appellate jurisdiction contrary to statute.
• Section 68 is mandatory (by s.4(1) and Schedule 1) and operates “notwithstanding any agreement to the contrary”: party agreement (and therefore consent orders) cannot create appellate permission jurisdiction.

iii) The key classification: jurisdictional “could not” vs merits “should not”

K1’s main argument sought to recharacterise the High Court’s refusal (that s.68(2)(g) was not engaged) as a “jurisdiction” decision, and therefore not a “decision under” s.68, thereby escaping s.68(4).

The Court rejected this as a category error. Unlike Cetelem SA v Roust Holdings [2005] EWCA Civ 618 (where the statutory power was expressly limited to certain types of order in urgent cases), and unlike Sumukan Ltd v Commonwealth Secretariat [2007] EWCA Civ 243 (where the statute itself treats party agreement as excluding the court’s jurisdiction), s.68 is structured as a merits-based remedy: the applicant must establish (1) one of the closed list of serious irregularities in s.68(2), and (2) substantial injustice.

Deciding whether the pleaded facts and legal characterisation satisfy s.68(2)(g) is therefore not an anterior inquiry into whether the court’s supervisory jurisdiction exists; it is the court doing the s.68 job—“overseeing the arbitration process”.

iv) Workability and policy: preventing erosion of the one-tier scheme

The Court stressed that K1’s approach would be “artificial and unworkable” because it would allow many s.68 disputes to be reframed as “jurisdictional” simply because they involve interpretation of a sub-paragraph (or “substantial injustice”) on assumed facts. That would gut s.68(4)’s finality function, contrary to:

• the statutory policy in Henry Boot Construction (UK) Ltd v Malmaison Hotel (Manchester) Ltd [2001] QB 388;
• the DAC’s stated intent that leave “may not be obtained from the Court of Appeal”; and
• National Iranian Oil Company v Crescent Petroleum Co International Ltd [2023] EWCA Civ 826, warning that even permission applications cause delay/expense and keep awards in limbo.

Accordingly, the High Court’s refusal remained a “decision under” s.68 and the Court of Appeal had no permission jurisdiction.

C) Impact

The judgment strengthens and clarifies the “one-tier gatekeeper” model for arbitration challenges under the 1996 Act:

• No appellate permission route: where s.68(4) applies, an applicant cannot ask the Court of Appeal for permission simply by reframing the High Court’s merits analysis as a “jurisdiction” determination.
• Interpretation disputes are still “under” s.68: arguments that a pleaded complaint does not fall within s.68(2) (including s.68(2)(g)) are treated as part of the s.68 merits, not a gateway issue.
• Consent/case management cannot circumvent statute: parties and judges cannot, by consent orders, create an appellate route barred by the Act, particularly for mandatory provisions.
• Practical consequence: parties must treat the first-instance permission decision as effectively final except in the narrow exceptional categories (e.g., a decision made without jurisdiction in the Cetelem sense; no decision at all; fair trial violations).

More broadly, the reasoning discourages tactical “jurisdictionalisation” of merits points and promotes rapid finality—key to London-seated arbitration’s efficiency and to the Act’s minimal-intervention philosophy.