In a newly reported Hamid choice of the Higher Tribunal, a barrister has been referred to the Bar Requirements Board for investigation following his use of a false quotation generated by ChatGPT. The case is MS v Secretary of State for the Dwelling Division (Skilled Conduct: AI Generated Paperwork) Bangladesh [2025] UKUT 305 (IAC).
Background
The barrister, Mr Muhammad Mujeebur Rahman, drafted grounds of attraction for the appellant on 14 March 2025, during which he referred to the case of Y (China), which doesn’t exist, with a quotation which was really for the case of R (YH) v Secretary of State for the Dwelling Division [2010] EWCA Civ 116.
Permission to attraction was granted on restricted grounds. On the error of legislation listening to on 20 June 2025 the decide requested Mr Rahman to take him to the related paragraph of Y (China), noting that the quotation was for a unique case about recent claims, reasonably than delay which was the difficulty on this attraction. Mr Rahman mentioned he didn’t want to depend on YH (Iraq) and after a few unsuccessful makes an attempt to quote different instances the tribunal took a break, offering Mr Rahman with a duplicate of R (Ayinde) v London Borough of Haringey, Al-Haroun v Qatar Nationwide Financial institution QPSC [2025] EWHC 1383 (Admin) and asking him to think about his place.
I do wonder if he really learn Ayinde throughout that break, as a result of reasonably remarkably when Mr Rahman returned, he informed the tribunal that “he had undertaken ChatGPT analysis in the course of the lunch break and the quotation for Y (China) was right, and it was a choice made by Tablet and Sullivan LJJ and Sir Paul Kennedy.” The tribunal then directed him to both present a duplicate of the choice by 24 June, or else to clarify what had occurred.
After the listening to, Mr Rahman handed the tribunal clerk “9 stapled pages which weren’t a judgment of the Courtroom of Attraction however an web print out with deceptive statements together with references to the fictional Y (China) case with the quotation for YH (Iraq).”
In compliance with the tribunal’s course, on 24 June 2025 Mr Rahman wrote and said that:
he had the truth is meant to quote YH (Iraq) (and specifically paragraph 24 of that call which he cites as saying all elements in an applicant’s favour have to be taken under consideration) and apologising for his failure to quote the complete and proper identify of the case. He blamed this on having suffered from “acute sickness” earlier than drafting the grounds; and on having been on a go to to Bangladesh between tenth and 18th June 2025, and the truth that he had been hospitalised in Bangladesh as a consequence of diabetes, ldl cholesterol issues and hypertension. He additionally argued that we should always not penalise him for this error as he has 5 relations (spouse and 4 kids) relying on him.
The tribunal then listed the Hamid listening to for 23 July 2025. Mr Rahman offered the tribunal with a letter of the identical date acknowledging with regards to Ayinde that it’s a breach {of professional} duties to depend on citations obtained by means of AI with out checking their veracity by means of respected authorized sources.
He accepted that he had used ChatGPT to draft the grounds of attraction and in addition to create the doc that he had handed to the tribunal clerk following the listening to on 20 June 2025. He offered varied excuses regarding his well being for why he had completed this and argued “that he was misled by the search engine and is thus additionally a sufferer”.
Mr Rahman accepted that YH (Iraq) was not related to the attraction. He apologised and mentioned that he shouldn’t be referred to the Bar Requirements Board as he would “act with integrity sooner or later and is unwell and anxious as to how he’ll help his household.”
There have been additionally points surrounding who was the consultant on document with the tribunal. Mr Rahman mentioned that he was instructed by Lextel Solicitors who have been on document with the Higher Tribunal, nevertheless the tribunal didn’t have any document of them for this case.
The tribunal had Mr Rahman of Lexminders’ Chambers Restricted as being on document, nevertheless Mr Rahman suggested the tribunal that this firm had solely existed between February 2023 and June 2024 when it was dissolved and since then he had operated as a self-employed barrister. The Higher Tribunal famous that within the First-tier Tribunal’s attraction, Mr Rahman had accomplished a bit 84 discover stating that he was showing on a direct entry foundation.
Referral to the Bar Requirements Board
The Higher Tribunal mentioned that they’d already referred Mr Rahman to the Bar Requirements Board in January this yr for considerations in one other attraction that he was conducting litigation with out being authorised to take action and that he lacked primary skilled competence. The tribunal dismissed his makes an attempt to clarify his behaviour and mentioned that if he had been unwell then he ought to have knowledgeable the tribunal and various counsel instructed if there was time.
The tribunal mentioned that:
Mr Rahman due to this fact has moved from an acceptance of the usage of ChatGPT however with a defence of the analysis and a defence of the pretend case of Y (China) on the day of the Panel error of legislation listening to; to a declare that it was a regrettable oversight and he did the truth is wish to depend on an irrelevant however real case in his letter of twenty fourth June 2025; to an acceptance earlier than us that he used ChatGPT to help in formulating the unique grounds and in manufacturing of the doc he handed to the Panel on twentieth June 2024, and that the case of Y (China) is pretend. We discover due to this fact that Mr Rahman has immediately tried to mislead the Tribunal by means of reliance on Y (China), and has solely made a full admission of this truth in his third rationalization to the Higher Tribunal. He has not due to this fact acted with integrity and honesty in coping with this difficulty, in addition to having tried to mislead the Tribunal within the grounds by means of the usage of an AI generated pretend authority.
The tribunal discovered that Mr Rahman “didn’t know that AI giant language fashions, and ChatGPT specifically, have been able to producing false authorities. It follows that this isn’t a case the place it could be applicable to refer the matter for police investigation or to provoke contempt proceedings.” Nevertheless a referral to the Bar Requirements Board was deemed “most undoubtedly applicable”. The choice ended with a sign that the tribunal would really like this and the sooner referral of Mr Rahman thought-about shortly by the Bar Requirements Board due to the seriousness of the considerations raised.
Headnote
The headnote states:
1. AI giant language fashions akin to ChatGPT can produce misinformation together with fabricated judgments full with false citations.
2. The Divisional Courtroom has offered steering within the case of R (Ayinde) v London Borough of Haringey, Al-Haroun v Qatar Nationwide Financial institution QPSC [2025] EWHC 1383 (Admin) that the consequence of utilizing AI giant language fashions in a means which leads to false authorities being cited is more likely to be referral to knowledgeable regulator, such because the BSB or SRA, as it’s a lawyer’s skilled accountability to make sure that checks on the accuracy of quotation of authority or quotations are carried out utilizing respected sources of authorized data. The place there may be proof of the deliberate inserting of false materials earlier than the Courtroom police investigation or contempt proceedings might also be applicable.
3. Taking unprofessional short-cuts which is able to very probably mislead the Tribunal isn’t excusable.
Conclusion
The tribunal famous on the outset of the choice that “the immigration shopper group could be notably weak”. That is after all right, and why it’s typically tough to seek out any sympathy for attorneys in these conditions, regardless of the stress and issue of the job. It definitely appears that what occurred right here was {that a} shopper paid for work to be carried out of their attraction which was then as a substitute outsourced to an AI device.
I personally suppose, given the by now well-known dangers of errors, hallucinations in addition to varied biases, that anybody who’s utilizing generative AI in authorized work needs to be upfront about it on the outset. That means individuals (whether or not that be purchasers or instructing solicitors) could make an knowledgeable choice about whether or not or how they’re completely happy for AI for use of their case.
