AI in the Courtroom

In this blog series we have been examining the growing use of Artificial Intelligence (AI) within regulatory investigations and proceedings and the benefits and pitfalls of this emerging technology.
 

In this blog, we consider how AI is being used in courts and tribunals by both legal representatives and litigants in person. Over the last 12 months, there has been a developing body of judgments handed down where the use of AI has been referenced as creating an issue in the hearing. These cases typically involve written submissions that have been provided to the court, by both legal qualified representatives and by litigants in person. We summarise some of these cases within this blog and reflect on the lessons learnt.

R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin)
 

The case of R (Ayinde) v London Borough of Haringey was one of the first reported cases in the English courts to grapple with the issue of what to do when “AI goes wrong”. The case received a significant amount of press attention, beyond just the legal press, and sent a sobering message to members of the legal profession.

The underlying case was a judicial review of a decision by a local council not to assess the Claimant for emergency housing.  The parties reached agreement on the outcome of the case, and so the court’s task was limited to determining costs, including a wasted costs application made against the Claimant's solicitors and barrister, seeking an order that they compensate the Council for costs incurred by their improper conduct. During the course of the proceedings, concerns were identified in relation to the use of fake case references in the Claimant's pleadings.

The case was referred to a Divisional Court under the court’s Hamid jurisdiction. This jurisdiction relates to the court’s inherent power to regulate its own procedures and to enforce duties that lawyers owe to the court.

The President of the King's Bench Division listed a hearing to consider what steps the Court should take, which the President has said may include contempt of court proceedings if it can be shown that the relevant conduct "does or may amount to an interference with the due administration of justice in High Court proceedings."

Following a hearing on 23 May 2025, the High Court decided not to issue contempt proceedings but the judgment sends a clear message that would be equally relevant to regulatory proceedings: AI needs to be used responsibly; legal representatives using AI tools to conduct research (or relying on others using AI) need to ensure they  check the accuracy of the output; and that in supervisory positions need to put in place “practical and effective measures” to ensure that individuals – especially regulated professionals - understand and comply with their regulatory and ethical obligations if using AI.

A similar issue arose in proceedings before the Administrative Court, in the case of Bandla v SRA [2025] EWHC 1167 (Admin).  In that case, a statutory appeal of a decision of the Solicitors Disciplinary Tribunal, the Appellant, a former practising solicitor, included fake authorities in his Grounds of Appeal.  Whilst the Appellant denied that this arose from the use of AI, and whilst Mr Justice Fordham did not make a factual finding in this respect, he nonetheless struck out the Grounds of Appeal as an abuse of process, citing the need to "take decisive action to protect the integrity of [the Court's] processes against any citation of fake authority."

D (a child) (recusal)
 

This case involved a litigant in person using AI, rather than a legal representative.

The mother, who represented herself in a hearing, submitted a ‘lengthy’ skeleton argument citing authorities, some of which were non-existent while others were irrelevant. An earlier document, in which the mother set out why the district judge in her case should recuse himself, included citations which ‘did not exist at all’. In the appeal hearing, the mother ‘accepted that she has used artificial intelligence to assist her in preparing the document’.

Lord Justice Baker, with whom Lord Justice Cobb and Lord Justice Miles agreed, expressed sympathy for litigants in person, finding it ‘is entirely understandable’ that LiPs resort to AI for help. However, all parties, represented and unrepresented, ‘owe a duty to the court to ensure that cases cited in legal argument are genuine and provide authority for the proposition advanced’, he said.

The judge said: ‘I absolve the mother of any intention to mislead the court. Litigants in person are in a difficult position putting forward legal arguments’.

PSA v NMC and Tchampet 2026 EWHC 141
 

In this case, the legal representative for the Registrant presented written submissions that were partially generated by AI. In particular, he relied on three authorities, two of which did not contain the words he attributed to them and none of which supported the propositions which he said they supported. This was pointed out to him at the hearing. He immediately admitted what he had done and that the references were phantoms created by AI. He promised not to use AI to generate submissions in future and to check his references personally. It was not possible to refer him to a regulator because, although he qualified as a solicitor, he was struck off the roll of solicitors by the Solicitors Disciplinary Tribunal in 2011 and has since then acted as an unregulated representative. He was warned of the potential for him to be charged with contempt of court but it was decided that it was not in this case necessary to take any further steps at this time.

The proliferation of fake case citations
 

When cases, such as those that have been mentioned above, are reported on, judges have been warned not to cite the AI generated case in full as they risk inadvertently embedding such bogus ‘authorities’ as precedents, creating self-propagating misinformation.

Matthew Lee, a barrister at Doughty Street Chambers and founder of the Natural and Artificial Intelligence in Law blog, warned that there was a tension between transparency and protection, where judges instinctively want to set out fabricated authorities in full but risk unintended consequences.  Lee told Counsel magazine: ‘Well-intentioned judges often cite hallucinated cases and their erroneous legal principles in full within official judgments to show the extent of the problem to those reading. However, judges may be inadvertently exacerbating the issue because those AI-generated inaccuracies are being integrated into the established legal canon indirectly.’

New Guidance for Barristers from the Bar Standards Board
 

On 18 May 2026, the Bar Standards Board introduced new guidance[1] on the safe and responsible use of AI for barristers to enable them to use AI safely, whilst continuing to meet their professional obligations.

It also provides good practice principles to support barristers at all stages of adopting and using technology. Such as:

• Considering how AI use will impact their ability to fulfil the Core Duties and rules in the BSB Handbook
• Maintaining a basic level of technology and AI awareness
• Evaluating the risks, benefits and costs of new technologies before using them
• Ensuring IT systems and data governance are properly implemented
• Being transparent, upholding client confidentiality and protecting sensitive information
• Recognising that clients, solicitors and opposing parties may use AI tools and how this use may impact barristers’ responsibilities.

Conclusion
 

Whether you are a legal representative or lay individual preparing for a court hearing, the use of AI to help with those preparations can be extremely useful and time saving. The cases cited above, however, provide a useful reminder to anyone appearing in a court or tribunal that any material produced must be carefully reviewed and verified for accuracy and to avoid any risk of misleading the court and undermining the administration of justice. Judges have explicitly warned that "blind reliance" on AI will be met with severe penal sanctions. For individuals that are regulated and who fall down in relation to their professional duties to check the accuracy of AI-generated work, also risk a referral to their regulator. So the message is clear - use AI responsibly and check (and check again) any output to ensure that it is accurate and reliable, particularly when case citations are being used.