The Upper Tribunal (UT) has referred a barrister to the Bar Standards Board (BSB) for investigation after he relied on a fabricated legal citation generated by ChatGPT.
Facts:
The case began on 14 March 2025, when a barrister, Mr. Rahman, drafted grounds of appeal for an immigration case, citing a non-existent case, Y (China), but using the actual citation of R (YH) v Secretary of State for the Home Department [2010] EWCA Civ 116. The citation he used was for a different, unrelated case. Despite this error, permission to appeal was granted on limited grounds.
At a hearing in June 2025, the Judge challenged Mr. Rahman on the false citation. Unable to provide a satisfactory explanation, Mr. Rahman was given a break. In a remarkable turn, he returned and claimed that "ChatGPT research" undertaken during the lunch break had confirmed that the fake case was genuine. When ordered to provide a copy of the judgement, he instead submitted a misleading nine-page printout from the Internet.
Over the next month, Mr. Rahman's story changed several times. He first blamed his actions on an "acute illness" and a recent hospital stay in Bangladesh, claiming that he had simply intended to cite a different case. However, in a later hearing in July 2025, he finally confessed that he had used
ChatGPT both to draft the original appeal grounds and the misleading printout, acknowledging that the case was fake. He argued he was a "victim" of the AI, claiming he did not know that the technology could produce misinformation.
The UT noted that this was not the first time Mr. Rahman had been referred to the BSB, as he was previously referred in January 2025 for concerns over his competence and for acting without a proper licence.
Decision:
The UT found that Mr. Rahman's actions demonstrated a clear lack of integrity. His shifting explanations and initial attempt to defend the AI-generated information showed a deliberate effort to mislead the Court. While the Court did not believe that he knowingly put false material before the Tribunal (and thus did not refer the matter for police investigation), it found his failure to verify the AI's output a serious breach of professional duties.
The UT's decision explicitly references the guidance in R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin), which warns that lawyers who fail to check AI-generated citations with reputable legal sources are likely to be referred to their regulator. The case concludes that taking unprofessional shortcuts that mislead the Court is never excusable, regardless of the lawyer's personal circumstances.
Implications:
This case serves as a stark warning about the risks of using unverified AI in legal practice. This case establishes the legal precedent that the UT expects lawyers to maintain professional standards and not rely solely on generative AI for legal research. As part of a lawyer's duty to the Court, honesty and integrity in professional conduct are of paramount importance. The use of AI tools such as ChatGPT for legal research is not an acceptable replacement for using reputable, verified legal databases such as Westlaw, EIN, or Bailii. The case also makes it clear that health or personal circumstances are not an excuse for providing incompetent work. An immigration lawyer's failure to adhere to these standards could lead to their referral to a professional regulator, a fine, or even disbarment.
This case serves as a cautionary tale for lay applicants and those representing themselves, individuals who might otherwise be tempted to use ChatGPT to save money on legal fees, especially in view of the fact that an immigration application is a critical, life-changing process. The Judge noted that the "immigration client group can be particularly vulnerable," and that this vulnerability is compounded when they are given false information.
Generative AI, as this case has proven, can fabricate legal facts and cases. A lay applicant using AI might well believe they have a strong legal argument based on a non-existent case, which might then lead to a considerable waste of the Court’s valuable time.
