R v Kyei [2025] EWCA Crim 1497: Social Media Posts by a Juror’s Partner and the Threshold for Jury Irregularity

1. Introduction

1.1. Overview of the case

This Court of Appeal (Criminal Division) decision concerns applications for leave to appeal against convictions arising out of a gang‑related shooting and subsequent destruction of a getaway car. The applicants were:

Courtney Ellis – convicted of the murder of Michael Small and the attempted murder of Serge Mukendi, and of perverting the course of justice.
Kyei (the first applicant) – convicted of perverting the course of justice by participating in the collection and destruction of the Mini Cooper used in the shooting.

Although the underlying criminal conduct was serious and complex, the appeal turned almost entirely on a jury irregularity allegation arising after the verdicts, namely:

Two Instagram comments posted by the jury forewoman’s partner (“Person Y”) on a public post about one applicant’s music project;
One of those comments was posted while the jury were still deliberating; both were mocking and plainly hostile;
The forewoman (“Person X”) was alleged to have improperly discussed her views and/or the jury’s deliberations with Person Y, and to have been influenced in turn by information he found online about the applicants.

The central question for the Court of Appeal was whether these events, and the subsequent police investigation, gave rise to an arguable jury irregularity rendering the convictions unsafe. The case therefore sits squarely in the modern line of authorities dealing with:

Juror conduct in the digital age (social media, smart devices, WhatsApp groups); and
The high threshold that must be met before an appellate court will interfere with a jury’s verdict on the basis of alleged extraneous influence or apparent bias.

1.2. Procedural posture

The applicants sought:

Leave to appeal against conviction, relying chiefly on alleged jury irregularity.
Leave to adduce fresh evidence under section 23 of the Criminal Appeal Act 1968 (“CAA 1968”) relating to a juror’s conduct and the Instagram material.

A single judge:

Referred the jury irregularity ground to the full court, in light of the seriousness of the allegations; but
Refused leave on a separate ground concerning the admission and use at trial of certain WhatsApp/Instagram messages (the “K‑Koke” ground).

The full court therefore had to decide:

Whether any further inquiry or evidence was required in relation to the alleged irregularity;
Whether an irregularity had in fact been established; and
Whether, if there were an irregularity, it might have had a material impact on the verdicts such that they were unsafe.

1.3. The key issues

The judgment raises and clarifies several important issues:

Social media activity by a juror’s partner: When, if ever, will hostile or predictive posts commenting on a defendant’s likely conviction by a juror’s partner undermine the safety of the verdict?
Threshold for jury irregularity: What level of evidence is necessary before the Court of Appeal will infer that deliberations were improperly disclosed or influenced by extraneous material?
Appearance of bias vs. proven irregularity: How does the objective “fair‑minded and informed observer” test operate where the primary concern is not direct proof of contamination, but a perceived risk?
Scope of post‑trial investigations and disclosure: What investigatory steps and disclosure are required when jury integrity is questioned, particularly where digital communications are inaccessible owing to refusal of PIN codes?
Limits of appellate intervention: How far can the Court go on the basis of inference and suspicion, in the absence of direct evidence that other jurors were influenced?

2. Summary of the Judgment

2.1. Factual background in brief

The prosecution case (summarised only briefly in the judgment) was that:

On 5 July 2019, Michael Small was shot dead at close range in Harrow Road, North London, by a masked gunman who escaped in a Mini Cooper on false plates.
Shortly afterwards, the same Mini Cooper was used in an attempted shooting of Serge Mukendi, in which the firearm failed to discharge.
The driver of the Mini was alleged to be Ellis, making him an accomplice to murder and attempted murder.
The Mini Cooper was quickly dumped, and the next evening collected by tow truck and taken to a salvage yard for destruction. Police later recovered parts from which DNA linked Kyei to the vehicle.
Both Ellis and Kyei were convicted of perverting the course of justice by arranging the collection and destruction of the Mini, knowing it had been used in serious crime.

2.2. The Instagram comments and investigation

Key events relating to the jury issue were:

Jury selection and trial: Jury sworn in October 2023; trial ran into December; deliberations began on 14 December 2023 but were adjourned over Christmas; resumed 16 January 2024.
Verdicts: Unanimous guilty verdicts returned on 19 January 2024; juror 6 served as forewoman (Person X).
Instagram comments by Person Y (forewoman’s partner):
- Comment 1: “They are going jail mate ain't no project coming out [smiling emoji]” – now accepted as posted on 16 January 2024, during deliberations.
- Comment 2: “Enjoy the showers [smiling emoji]” – forensic analysis showed it was posted at 00:02 on 20 January 2024, i.e. after verdicts.
Discovery: On 20 January 2024, Kyei noticed Comment 2; on reviewing the profile of Person Y, he saw photographs of Person Y with the jury forewoman (Person X) and their children, thereby linking the account to a juror.
Police investigation: Both X and Y were arrested, gave “no comment” interviews, refused device PIN codes and access to Instagram accounts. Other jurors were interviewed and/or gave statements; none reported any misconduct, outside influence, or breach of oath.
Forensic digital work: Examination of Kyei’s phone confirmed Comment 2 as post‑verdict; Comment 1 had been deleted and its time of posting could not forensically be pinned down, though it was ultimately accepted as 16 January.
Outcome of investigation: Police concluded there was no evidential basis to prosecute X or Y, and that there was no realistic prospect of uncovering evidence of improper communication between them about deliberations.

2.3. The applicants’ case

The applicants argued, in essence, that:

The forewoman must have discussed the case – and, more seriously, the developing deliberations and her views on guilt – with Person Y, enabling him to make the pre‑verdict comment predicting imprisonment.
Person Y had plainly carried out internet research revealing gang affiliations and other negative material about the applicants, which would have prejudiced him and, via discussion, prejudiced Person X.
The refusal of both X and Y to provide PINs for their devices strongly supported an inference that those devices contained incriminating communications.
The situation created an appearance of bias and a real risk that extraneous information had reached the jury, undermining the integrity of the verdicts.
The police investigation was, at best, “light touch” and, at worst, inadequate, especially in not compelling access under section 49 of the Regulation of Investigatory Powers Act 2000 (RIPA 2000).

2.4. The Crown’s case

The Crown responded that:

Comment 2 was post‑verdict and irrelevant to deliberations.
Comment 1 was made by Person Y, not by a juror, and required no special inside information to be written – it was of the kind routinely found on social media when serious charges are publicly known.
There was no evidence that Person X had ever disclosed deliberations, votes, or arguments to Person Y, or that he had fed any online research back to her.
All other jurors who provided statements confirmed:
- They had no concerns about X’s conduct;
- They were not influenced or coerced;
- They decided the case solely on the evidence.
The investigation and disclosure process, including the use of independent disclosure counsel, was sufficient and proportionate.

2.5. The Court’s decision

The Court of Appeal held that:

There was no sufficient evidence of any jury irregularity that could render the convictions unsafe.
Even assuming the first Instagram comment was made during deliberations, the court could not properly infer that it resulted from prohibited communications between Person X and Person Y.
It was unsurprising, and not in itself damaging, if Person X had told her partner which case she was trying (even if contrary to directions), but there was no evidence she disclosed protected aspects of deliberations.
The unanimous evidence of the other jurors supported the view that deliberations remained confidential and unimpaired.
The integrity of the trial process and jury’s deliberations remained intact; the convictions were safe.
The applications for leave to appeal against conviction were therefore refused.
The separate “K‑Koke” ground (relating to cross‑examination and bad character material) was held not to be before the court; in any event, the single judge’s rejection of that ground was endorsed.

3. Detailed Analysis

3.1. Legal and doctrinal context

3.1.1. Jury irregularity and “unsafe” convictions

Under section 2 of the Criminal Appeal Act 1968, the Court of Appeal must allow an appeal against conviction if the conviction is “unsafe”. Allegations of jury irregularity fall into that framework: the question is not simply whether an irregularity occurred, but whether it “may have” affected the verdict such that the conviction cannot be regarded as safe.

In modern case law, especially following authorities such as R v Mirza and R v Connor and Rollock, the courts:

Maintain a strong presumption that jurors obey their oath and directions;
Are slow to investigate or disturb jury deliberations, primarily due to section 8 of the Contempt of Court Act 1981 (which protects deliberation secrecy) and the need to maintain the finality of jury verdicts;
Will intervene only where there is clear evidence that extraneous material or improper influence may have affected the verdict.

The present case was clearly decided against that doctrinal background, even though the judgment text as provided contains no explicit case citations. The Court repeatedly emphasised the need for evidence, not speculation, before concluding that a verdict is unsafe.

3.1.2. Fresh evidence: section 23 CAA 1968

The applicants sought to rely on the Instagram comments and the product of the police investigation as fresh evidence. Section 23 CAA 1968 allows the Court of Appeal to receive such evidence if:

It appears capable of belief;
It would have been admissible at trial; and
There is a reasonable explanation for why it was not called at trial; and
It may afford a ground for allowing the appeal.

Those requirements are applied flexibly but with an eye to relevance and potential impact. Here, the Court was willing to consider the Instagram material and the juror statements but ultimately concluded that, even taken at their highest, they did not cast doubt on the safety of the convictions.

3.2. Precedents and authorities (express and implicit)

The supplied judgment extract does not record any specific case authorities. Nonetheless, the reasoning reflects – and effectively re‑applies – several established principles from earlier appellate decisions. These can be grouped as follows:

3.2.1. The Mirza / Connor line: protecting deliberations but addressing irregularities

In R v Mirza [2004] and R v Connor and Rollock [2004], the House of Lords emphasised:

Deliberations are ordinarily inviolate: the court will not inquire into the reasoning process within the jury room.
However, irregularities involving extraneous material or external influence can justify investigation and may vitiate verdicts.

The present judgment is firmly in line with this structure:

The court accepts that social media posts by a juror’s partner, especially during deliberations, are serious enough to merit police investigation and disclosure.
But it insists on a boundary: questions about how jurors reasoned, why they voted as they did, and internal pressures within the jury are not to be explored absent clear evidence of improper external influence.

3.2.2. The appearance of bias test: Porter v Magill and its progeny

Counsel for the applicants framed part of their case using the language of “appearance of bias” and the fair‑minded and informed observer. This invokes the Porter v Magill [2002] 2 AC 357 test:

Would the fair‑minded and informed observer, having considered the facts, conclude that there was a real possibility that the tribunal was biased?

In the jury context, that test must be applied with sensitivity to:

The presumption that jurors follow their oath and directions;
The actual evidence (not speculation) about what jurors did and knew.

Here, while not citing Porter expressly, the Court effectively rejected the assertion that a fair‑minded observer would see a real risk of bias, because:

There was no proof that the forewoman shared deliberations or votes with Person Y;
Other jurors independently confirmed the integrity of the process;
The Instagram post, though crude and ill‑advised, could be explained without assuming contamination of the jury.

3.2.3. Jurors and the digital age

There is a growing body of case law and contempt proceedings dealing with jurors’ use of the internet and social media (for example, Attorney General v Fraill, R v Dallas, and related cases). These cases typically involve:

Jurors themselves researching a defendant or case online;
Communicating with defendants or witnesses; or
Posting about the case on social media during trial.

The present case is distinctive in that it was not the juror, but her partner, who posted. The judgment implicitly extends the existing framework to this scenario:

Posts by a juror’s partner are capable of raising serious concerns;
They can justify arrests and investigation; but
They do not automatically invalidate the trial – the key question remains: did the jury’s decision‑making actually become tainted by extraneous material?

3.3. The Court’s legal reasoning

3.3.1. No renewal – and no merit – in the “K‑Koke” ground

Before addressing the core irregularity ground, the court dealt with an attempted resurrection of the “K‑Koke” ground (relating to cross‑examination and alleged straying beyond a bad character ruling). The court held:

The ground had not been renewed and was therefore not properly before the court.
Even if it had been, the single judge’s reasoning – that the trial judge acted within his discretion and any concerns were resolvable by summing‑up – was “unassailable”.

This underlines an important procedural point: in leave applications, precision as to which grounds are renewed matters, and the Court will not lightly reopen grounds already refused without clear and specific renewal.

3.3.2. Evidence and inference: the Instagram posts

The Court then turned to the Instagram comments. Its reasoning can be distilled as follows:

Timing of posts:
- “Enjoy the showers” (Comment 2) was definitively posted after verdicts and thus irrelevant to any question of influence during deliberations.
- “They are going jail mate...” (Comment 1) was accepted as posted on 16 January 2024, when deliberations were ongoing.
Source of information for Comment 1:
- The Court accepted it was “likely” that Person X told her partner which case she was trying. That, however, was:
  - Unsurprising in human terms;
  - Not in itself damaging, provided she did not reveal protected aspects of deliberations.
- Person Y could also have known much about the case from publicly available information – press reports, the nature of the charges, and their seriousness – and formed a view that conviction was likely.
- The content of the post did not require knowledge of the confidential deliberative process; it was a crude prediction consistent with the gravity of the charges.
No evidence of forbidden disclosure:
- Person X, via her prepared statement, expressly denied disclosing:
  - Statements made by jurors;
  - Opinions expressed;
  - Arguments advanced; or
  - Votes cast.
- No independent evidence contradicted this. Importantly:
  - Other jurors did not observe or suspect any improper disclosure or external influence;
  - None said their own decision‑making had been affected or pressured.
Refusal to disclose PINs and inference:
- Defence counsel invited the Court to draw an adverse inference from X and Y’s refusal to provide device PINs, suggesting it indicated incriminating content.
- The judgment implicitly rejects this invitation, emphasising that without actual content, such an inference remains speculative; appellate intervention cannot rest on speculation alone.

In short, the Court accepted that:

Some level of communication between X and Y about the existence of the trial was probable;
But there was no evidential basis to find that the substance of deliberations or confidential juror views were shared;
Nor that any online research by Y fed back into the jury room.

3.3.3. The weight placed on evidence from other jurors

The Court placed considerable emphasis on the responses received from other jurors. From seven written statements and broader inquiries, the salient points were that:

No juror reported:
- Being approached by any third party;
- Being encouraged to disclose deliberations;
- Noticing any attempt to influence their verdicts.
Each juror confirmed:
- They had been true to their oath;
- Their decision was their own and based solely on the evidence.
Juror A mentioned X acknowledging she had discussed general frustrations with her partner over Christmas (“I said the same thing to [Person Y]”), but nothing that went to votes or specific views.
Juror B noted that X sometimes retained her smartwatch, but neither:
- Saw her using it during deliberations; nor
- Observed any sharing of deliberations with outsiders.
Juror C perceived that X was well-prepared and structured in deliberations – a neutral or positive observation, not evidence of improper conduct.

Taken together, these accounts led the Court to conclude that:

“All the evidence points to a conclusion that the integrity of the trial process, and in particular the deliberations of the jury, remained intact…”

This is a strong reaffirmation of the principle that fellow jurors’ evidence about what did not happen in the jury room – no improper external influence, no pressure, no sharing of deliberations – carries real weight in assessing alleged irregularities.

3.3.4. Adequacy of the police investigation and disclosure

Defence counsel criticised the investigation as “light touch” or even “entirely inadequate”, arguing that:

No effective attempts were made to obtain communications between X and Y;
Section 49 RIPA 2000 powers might have been used to compel PIN disclosure;
The inability to see those communications prevented a full and fair assessment.

The Court’s response is notable in several ways:

It held that the prosecution had investigated thoroughly and that all relevant and potentially helpful material had been disclosed.
Independent disclosure counsel had reviewed juror statements to strike a balance between:
- Protecting juror anonymity and the confidentiality of deliberations; and
- Ensuring transparency and fairness to the applicants.
The Court expressly found that the applicants had been given disclosure of all relevant material to which they were entitled.

Importantly, the judgment does not:

Second‑guess the choice not to deploy RIPA section 49 powers; or
Hold that failure to do so rendered the investigation inadequate.

Instead, the Court focuses on the available evidence and concludes that, even without access to device contents, what was uncovered did not support the assertions of irregularity. The standard is not perfection in investigation, but sufficiency for the appellate court to assess whether the convictions remain safe.

3.3.5. Appearance of bias vs. need for evidence

Counsel for the applicants urged the Court to concentrate on the appearance of bias, arguing that:

The combined facts – a forewoman whose partner posts pre‑verdict that the defendants are going to jail, hostility in the tone of posts, refusal to disclose PINs – would lead a fair‑minded observer to conclude there was a real risk of bias.
The Court should not require proof of actual contamination in order to find the verdict unsafe.

The Court’s answer is implicit but clear:

Appearance of bias must be assessed on the basis of established facts, not stacked speculation.
The only safe findings were:
- X likely told Y which case she was on;
- Y formed his own negative views and posted crude comments;
- No deliberation content, juror votes, or specific arguments were shown to have been shared.
In that context, a fair‑minded and informed observer, knowing all this, would not see a real possibility that the jury’s decision was tainted.

Accordingly, while “appearance of bias” remains a vital safeguard, it cannot be invoked to compensate for the absence of evidence that anything improper, beyond humanly unsurprising references to the existence of the trial, actually occurred.

3.4. Key legal principles and clarifications emerging from the case

Although the Court did not claim to be creating new law, the judgment provides clear and practical guidance on several points.

3.4.1. Social media posts by jurors’ family members

The judgment effectively establishes that:

Hostile or mocking social media posts by a juror’s partner may justify investigation but do not, by themselves, amount to a jury irregularity.
What matters is whether those posts:
- Demonstrate that deliberation content or juror votes were disclosed; or
- Show that extraneous, prejudicial information was fed back into the jury room.
A merely predictive or hostile comment (“they are going to jail”) that could have been made on the basis of public information does not suffice, absent more.

3.4.2. Limited consequences of a juror breaching “do not discuss” directions

The Court accepts it is “likely” that Person X said something to Person Y about the case. Strictly, this may have breached the trial judge’s directions, but the Court treats that in isolation as:

Unfortunate and technically improper; but
Not automatically fatal to the verdict;
Not sufficient evidence, without more, of a material irregularity.

This is of practical importance: it reaffirms that not every breach of a direction about external discussion will render a conviction unsafe. The key is whether the breach has a realistic potential to influence the jury’s assessment of the evidence.

3.4.3. Use of other jurors’ testimony in post‑trial inquiries

The judgment reinforces that:

Other jurors’ statements, carefully curated to avoid intruding into the content of deliberations, are a legitimate and valuable source of evidence in assessing alleged irregularities.
Where those jurors uniformly attest to:
- Independent decision‑making;
- Absence of external influence;
- Absence of coercion or improper pressure;
the Court will place great weight on such evidence.

3.4.4. Threshold for appellate intervention

The Court’s approach underscores a key point: suspicion, inference, and unease are not enough. For an appellate court to find a conviction unsafe on jury‑irregularity grounds, there must be:

Some tangible evidence of:
- Extraneous material entering the jury’s deliberations; or
- Improper pressure or influence on jurors.
Or, at least, such compelling circumstantial evidence that the only reasonable inference is contamination of the deliberative process.

In this case, the Court found that threshold plainly not met.

3.5. Likely impact on future cases

3.5.1. For criminal appeals based on social media activity

Future appellants alleging jury irregularity arising from social media posts by jurors’ relatives, friends, or partners will face a demanding standard. This case suggests:

Evidence that a juror’s relative posted hostile comments during a trial will usually trigger investigation, but
An appeal is unlikely to succeed unless it can be shown that:
- The juror disclosed deliberations or confidential juror views; or
- The juror was influenced by information from those posts or external research.

3.5.2. For trial judges and directions to jurors

The judgment will likely reinforce the need for:

Clear, repeated directions that jurors must not:
- Discuss the case with anyone outside the jury except the court staff/judge if needed;
- Allow others to research or discuss the case on their behalf;
- Use smart devices (including watches) for communication during deliberations.
Firm control of devices when juries retire – as occurred here, where phones were removed, though smartwatches presented a minor enforcement challenge.

However, the decision also reassures judges that:

Not every breach, particularly if minor and not clearly prejudicial, compels a discharge of the jury or a stay of proceedings.

3.5.3. For police and prosecutors: scope of post‑verdict investigations

This case is a concrete example of best practice in several respects:

Swift referral by the trial judge to the Registrar of Criminal Appeals and the police once the issue arose;
Arrest and interview of the juror and her partner;
Forensic digital examination (even though constrained by lack of PIN access);
Systematic contact with other jurors and collation of statements;
Use of independent disclosure counsel to manage sensitive material.

The Court’s endorsement of these steps will likely be used as a benchmark in future investigations into alleged jury irregularity in the digital context.

3.5.4. For defendants and defence practitioners

Defence representatives should draw several practical lessons:

Digital evidence (screenshots, timestamps, metadata) must be carefully preserved and, where possible, validated forensically.
Allegations of jury irregularity should be framed with:
- Specific assertions about what irregularity is alleged (e.g. disclosure of votes, receipt of extraneous information), and
- Concrete evidence or at least strongly corroborative circumstances.
Arguments based solely on refusal to provide PINs or obvious hostility in social media comments are unlikely to succeed absent more direct evidence of contamination.

4. Complex Concepts Simplified

4.1. “Jury irregularity”

A “jury irregularity” is anything that happens in relation to the jury which:

Should not have happened; and
Might affect the fairness of the trial or the safety of the verdict.

Examples include:

Jurors researching the defendant online;
Jurors being approached by outsiders trying to influence them;
Jurors themselves posting about the case in public;
Jurors discussing the case improperly with non‑jurors.

Not every irregularity will make a conviction unsafe; what matters is whether it could realistically have affected the jury’s decision.

4.2. “Unsafe conviction”

The Court of Appeal must quash a conviction if it thinks it is “unsafe”. This does not mean:

That there was any misconduct by the trial judge or lawyers; or
That the jury definitely got it wrong.

It means that, taking everything into account:

There is a serious doubt about the reliability or fairness of the conviction.

In jury irregularity cases, the question becomes: is there a real possibility the jury’s decision was influenced by something it should not have been?

4.3. “Appearance of bias” and the fair‑minded observer

The “appearance of bias” test asks:

Would a fair‑minded and properly informed observer think there was a real possibility that the decision‑maker (here, the jury) was biased?

This is an objective test. It requires:

Knowledge of all the relevant, proven facts; and
A calm, reasonable view – not mere suspicion or anxiety.

In this case, the Court essentially held that such an observer, knowing what the investigation had found (and not found), would not see a real possibility that the jury was biased or improperly influenced.

4.4. Fresh evidence on appeal – section 23 CAA 1968

“Fresh evidence” is material that was not used at the original trial but is later put before the Court of Appeal. The Court can receive it if:

It is credible (capable of belief);
It could have been used at trial;
There is a good reason it was not used before; and
It might realistically have led to a different verdict.

Here, the Instagram posts and police investigation were treated as fresh evidence. The Court accepted they were credible and properly before it, but concluded they did not cast sufficient doubt on the safety of the convictions.

4.5. Section 49 RIPA 2000 (PIN disclosure)

Section 49 of the Regulation of Investigatory Powers Act 2000 allows certain authorities, in some circumstances, to issue a notice requiring a person to disclose:

A PIN code or password; or
Otherwise provide access to encrypted information.

Failure to comply with a valid section 49 notice can be a criminal offence. In this case:

Defence counsel suggested such a notice might have been used to require X and Y to disclose their device PINs.
We are not told whether that power was in fact considered or used.
The Court did not treat any failure to use it as undermining the validity of the investigation.

5. Conclusion

The decision in R v Kyei [2025] EWCA Crim 1497 is an important modern authority on jury integrity in the digital age. It demonstrates that:

Social media activity by jurors’ partners, even if hostile and ill‑timed, does not automatically render convictions unsafe;
What matters is evidence of actual or contamination of the deliberative process – such as disclosure of votes, opinions, or extraneous prejudicial material reaching the jury;
The Court of Appeal will place substantial weight on statements from other jurors attesting to the independence and purity of their decision‑making;
The threshold for setting aside a jury’s verdict on irregularity grounds remains high: speculation, suspicion, and unease, even when understandable, are insufficient.

At the same time, the case underscores the seriousness with which such allegations are treated:

Arrest and investigation of a juror and partner;
Digital forensic work and examination of defendants’ devices;
Careful, controlled inquiries of other jurors under the supervision of independent disclosure counsel.

Ultimately, the Court found that the integrity of the deliberations in this case remained intact and that there was no arguable basis for concluding that the verdicts were unsafe. The judgment thus provides both reassurance about the robustness of the jury system and a clear framework for analysing alleged irregularities arising from the increasingly complex and intrusive realities of social media and digital communication.