Rebutting Section 72 “Danger to the Community” After a Grave Index Offence: Risk Must Be So Low It Can Be Discounted, and Devaseelan Requires a Real Starting-Point Analysis

Case: KD v Secretary of State for the Home Department
Citation: [2026] EWCA Civ 349 (CA, Civ Div)
Date: 23 March 2026

1. Introduction

This appeal concerned the interaction between (i) the Refugee Convention’s protection against refoulement, (ii) the statutory exclusion mechanism for serious offenders under section 72 of the Nationality, Immigration and Asylum Act 2002 (“NIAA 2002”), and (iii) the procedural discipline required when a claimant re-litigates an asylum narrative previously rejected after a full fact-finding hearing.

The respondent, KD, a Turkish national, entered the UK clandestinely in 2001 and claimed asylum based on Kurdish/Alevi identity and asserted involvement with DHKP-C. In 2004, an adjudicator (Judge Peart) dismissed the claim in emphatic terms, finding the account fraudulent. KD nonetheless remained unlawfully and, in 2005, murdered his wife; he was later convicted and sentenced to life imprisonment (minimum 12 years). After release, KD made a further asylum claim (again based on political persecution) and advanced a new Articles 2 and 3 ECHR claim based on an alleged blood feud arising from the murder.

The First-tier Tribunal (FtT) allowed the asylum appeal, finding KD had rebutted the section 72 presumption and accepting renewed political-risk evidence (including a late-produced arrest warrant). It also accepted that a blood feud existed but left the human-rights analysis on that ground incomplete. The Upper Tribunal (UT) dismissed the Secretary of State’s appeal. The Secretary of State appealed to the Court of Appeal.

The core issues were:

• Section 72: whether KD rebutted the presumption that, having committed a particularly serious crime, he constituted a danger to the community.
• Devaseelan discipline: whether the FtT lawfully departed from the 2004 adverse credibility findings when KD made essentially the same political-persecution claim.
• Documentary reliability: whether the FtT lawfully treated the 2001 arrest warrant as reliable, particularly given its late production.
• Delay: whether the UT’s 12-month delay rendered its decision unsafe.

2. Summary of the Judgment

The Court of Appeal (Arnold LJ and Dove LJ concurring) allowed the Secretary of State’s appeal. It held that:

• The UT’s delay and brevity were “entirely unacceptable” and deprived the FtT decision of any supportive appellate endorsement, but delay alone did not make the UT decision unsafe on the facts.
• The FtT’s conclusion that KD had rebutted the section 72 presumption was perverse: given the gravity of the index offence and the acknowledged continuing risk to intimate partners, the evidence did not support displacement of the presumption.
• The FtT materially misapplied Devaseelan (Second Appeals – ECHR – Extra-Territorial Effect) Sri Lanka) [2002] UKIAT 00702; [2003] Imm AR 1 and failed to treat the 2004 determination as the true “starting point”; it effectively re-tried the same claim without the mandated circumspection.
• The asylum and political-opinion-based Articles 2/3 claims were dismissed; however, because the FtT left the blood-feud human-rights claim undetermined (and because no findings were preserved), the matter was remitted to the UT for expedited determination on the blood-feud ground (including sufficiency of protection and internal relocation).

3. Analysis

3.1 Precedents Cited

(A) Delay and safety of the decision

The court applied the principle that delay is not itself dispositive; the question is whether it renders the decision unsafe such that it would be unjust to let it stand:

• SS (Sri Lanka) v SSHD [2018] EWCA Civ 1391; [2018] Imm AR 1348 – delay warrants intervention only if it compromises safety/unfairness.
• Bank St Petersburg PJSC v Arkhangelsky [2020] EWCA Civ 408; [2020] 4 WLR 55 – delay can corrode confidence and may undermine safety depending on context.

The Court of Appeal strongly criticised the UT’s 12-month delay and the “remarkably slight” reasons. Nonetheless, it did not allow the appeal on this ground because the appellate court could address the underlying legal errors in the FtT decision directly.

(B) Section 72 “danger to the community” and risk assessment

The court placed the section 72 question within the Refugee Convention framework, especially Article 33(2), and relied on:

• EN (Serbia) v Secretary of State for the Home Department [2009] EWCA Civ 630; [2010] QB 633 – “danger to the community” requires a real danger; where there is a real risk of repetition of a particularly serious crime, danger is likely established.
• MA (Pakistan) v Secretary of State for the Home Department [2014] EWCA Civ 163 – tribunals are not bound by probation assessments; even a quantified 17% reoffending risk over two years is not insignificant in deportation contexts.
• Miftari v SSHD [2005] EWCA Civ 481 – the perversity threshold: a decision no reasonable decision-maker could reach on the evidence.

These authorities informed two key moves in the reasoning: (i) “risk” is not to be treated as negligible merely because labelled “low”, and (ii) seriousness of the feared harm is integral to whether a real but “low likelihood” risk still constitutes danger.

(C) Devaseelan discipline and second-appeal structure

The court treated Devaseelan as central to fairness and consistency in repeat asylum litigation:

• Devaseelan (Second Appeals – ECHR – Extra-Territorial Effect) Sri Lanka) [2002] UKIAT 00702; [2003] Imm AR 1 – provides structured guidance on how to treat earlier determinations.
• Djebbar v SSHD [2004] EWCA Civ 804; [2004] Imm AR 497 – approves Devaseelan for “sensible flexibility and desirable consistency”.
• BK (Afghanistan) v SSHD [2019] EWCA Civ 1358; [2019] 4 WLR 111 – summarises the Devaseelan guidelines (starting point; circumspection with personal evidence not previously adduced; caution; avoiding re-litigation absent materially different facts; “very good reason” requirement).

The Court of Appeal’s critique was not that the FtT failed to cite Devaseelan, but that it treated “starting point” as a formula rather than an analytic method.

(D) Documentary reliability in asylum cases

• Tanveer Ahmed [2002] UKIAT 00439; [2002] Imm AR 318 – the claimant must show that a document can be relied upon, assessed “in the round”.

Although the Court of Appeal treated Ground 4 as “subsumed” within Devaseelan errors, the arrest warrant featured as an archetypal Tanveer Ahmed problem: late production without adequate explanation demanded a reliability analysis, not a mere reference to corroboration.

(E) Country guidance and returnee risk (contextual, not decisive)

The FtT referenced country guidance including IA (Turkey) 2003 UKAIT 00034 and IK (Returnees – Records – IFA Turkey) CG 2004 UKIAT 00312, and the earlier 2004 determination referred to MS (Turkey) GBTS Info at Borders, Turkey [2004] UK IAT 00192, CE Turkey CG [2004] UK IAT 00233, and AG Turkey CG [2004] UK IAT 00168. The Court of Appeal’s intervention did not turn on the content of this guidance, but on the unlawfulness of the FtT’s fact-finding foundation (dangerousness and Devaseelan misuse).

3.2 Legal Reasoning

(A) Ground 1: delay—condemnation without dispositive effect

The court accepted that the UT’s delay and thin reasoning undermined confidence. It highlighted a concrete indicator of potential misrecollection: the UT’s statement that the presenting officer “could not say very much” was difficult to reconcile with extensive preparatory notes (including a reference to MA (Pakistan)). Even so, applying SS (Sri Lanka) v SSHD and Bank St Petersburg PJSC v Arkhangelsky, the court held that the appeal should succeed (or fail) on the underlying legal merits rather than delay alone.

(B) Ground 2: section 72 perversity—risk must be effectively discountable in a case of extreme harm

The court accepted that risk assessments allow latitude and require appellate restraint. But it found the FtT’s reasoning crossed the line into perversity because it failed to engage adequately with:

• The magnitude of the index offence and feared repetition: the type of harm contemplated was the “most extreme form of domestic violence”.
• The continuing real risk: the evidence still described KD as a “medium risk of serious harm to intimate partners”.
• The community’s composition: “The community consists of all its members, including those with whom a person is on intimate terms.”
• “Low likelihood” is not “insignificant”: echoing MA (Pakistan), a low likelihood can still be a real likelihood (i.e., not fanciful), particularly where consequences are grave.

The court distilled an important practical standard: in a case where both the index offence and the potential harm on repetition are exceptionally grave, rebutting section 72 requires evidence that the likelihood of future serious offending is so low that it can effectively be discounted. Because KD could not reach that level on the evidence, the court substituted the conclusion that the section 72 presumption was not displaced, excluding KD from Refugee Convention protection.

(C) Grounds 3 and 4: Devaseelan is a method, not a mantra

The Court of Appeal’s most jurisprudentially significant contribution lay in its clarification of what it means to treat an earlier determination as the “starting point”. It held that this requires the later tribunal to:

• Identify the overlapping issues and what exactly was authoritatively decided earlier;
• Engage with why the earlier tribunal rejected the account (especially where credibility findings were comprehensive);
• Apply the relevant Devaseelan guidelines (including circumspection and “very good reason” inquiries), not merely cite them.

Applying that approach, the court held the FtT committed multiple errors, including:

• Re-litigation of an identical historical question: KD’s asserted DHKP-C involvement in Turkey between 1994 and 2001 was the same issue determined in 2004.
• Illusory “new evidence”: AD and AHD did not provide materially better, circumstantial evidence of KD’s actions in 1994–2001; they were not better placed than KD (and the late wife) who were disbelieved for strong reasons.
• Failure to weigh prior credibility demolition: Judge Peart’s findings were not a peripheral “view”; they were detailed, based on oral evidence, and included an explicit fraud finding.
• Lack of “greatest circumspection” (guideline 4): the FtT accepted late personal evidence without explaining how it overcame the earlier contradictions, inconsistencies, and improbabilities.
• No “very good reason” analysis (guideline 7): AD had been in the UK long before 2004, yet the FtT did not address why he did not testify then.
• Documentary reliability not confronted: the 2001 arrest warrant was not produced in 2004, no satisfactory reason was given, and the FtT treated it as corroborative without the necessary Devaseelan/Tanveer Ahmed scepticism.

The UT’s error was to treat these as matters of evaluative discretion, and to minimise the document point as “immaterial”. The Court of Appeal rejected that framing: where a late document is used to rehabilitate a previously disbelieved narrative, the antecedent question is reliability, not merely materiality.

(D) Remedy: dismissal of political-opinion protection; remittal of blood-feud Articles 2/3

Having found (i) section 72 not displaced and (ii) Devaseelan misapplied, the court dismissed asylum and political-opinion-based human-rights claims. However, it remitted the blood-feud claim to the UT because the FtT had not determined internal relocation and sufficiency of protection. Notably, the court ordered that no findings be preserved, because the blood-feud fact question and the protection/relocation questions were “closely connected”, and because the wider FtT findings (based on the same witnesses) were set aside.

3.3 Impact

• Higher practical bar to rebut section 72 in grave-violence cases: while section 72 is always rebuttable, this judgment signals that where the index offence and potential recurrence consequences are exceptionally severe, a tribunal must explain why risk is so low it can effectively be ignored; labels like “low likelihood” will not suffice if the risk is real.
• “Community” includes intimate partners: the judgment squarely rejects any implicit narrowing of “community” that would downplay risk to domestic partners; this matters in deportation litigation involving domestic abuse histories.
• Devaseelan compliance tightened: tribunals are reminded that Devaseelan is an analytic framework. A “starting point” must be located by identifying the earlier issue, the earlier reasoning, and whether the new material truly changes the factual landscape. This is likely to be cited where repeat claims attempt to re-run discredited narratives with supportive family evidence.
• Late documents and credibility rehabilitation: the judgment reinforces that a late-produced document (especially one that “should” have existed at the earlier hearing) requires explicit reliability reasoning; tribunals cannot treat such material as mere corroboration without confronting why it was absent and how it can be trusted.
• Institutional message on UT delay: although not outcome-determinative, the court’s criticism is unusually direct and may be deployed in future cases to argue that a delayed, skeletal UT decision provides little appellate “buttress” to an FtT determination under challenge.

4. Complex Concepts Simplified

• Refoulement: returning someone to a country where they face threats to life or freedom for a Convention reason. The Refugee Convention generally prohibits it (Article 33(1)).
• Article 33(2) Refugee Convention (exception): even a refugee can lose the benefit of non-refoulement if there are reasonable grounds to regard them as a danger to the community (including after conviction for a particularly serious crime).
• Section 72 NIAA 2002: a statutory mechanism that presumes certain serious offenders (here, plainly including a murderer) both committed a “particularly serious crime” and “constitute a danger to the community”. The presumption is rebuttable, but the burden is on the claimant.
• Perversity (in public law/appeals): not “a mistake”, but a conclusion that no reasonable tribunal could reach on the evidence.
• Devaseelan “starting point”: when someone brings a second appeal on similar facts, the first determination is the baseline. The new tribunal must avoid simply re-trying old issues unless there are materially different facts, and it must treat previously available personal evidence with great caution.
• Tanveer Ahmed approach to documents: documents are not presumed reliable; the claimant must show they can be relied upon, looking at everything together (timing, provenance, consistency, credibility context).
• Sufficiency of protection / internal relocation (blood feud cases): even if a threat exists in one area, the question is whether the home state can protect the person, or whether they can safely and reasonably live elsewhere in the country.

5. Conclusion

KD reinforces two linked principles in deportation-asylum litigation involving serious offenders and repeat claims. First, where the index offence and potential harm are of extreme gravity, rebutting the section 72 presumption requires demonstrably discountable future risk; tribunals must not treat a “low likelihood” as negligible where consequences are severe. Second, Devaseelan is not satisfied by ritual incantation: a later tribunal must perform a disciplined starting-point analysis, identify whether the later case is materially different, and justify any departure from earlier credibility findings—especially where late documents are used to revive a previously rejected narrative without adequate explanation. The political-opinion protection claims failed; the blood-feud Articles 2/3 claim survives only to the extent it must be freshly determined on remittal.