Misuse of AI has struck once again in the courtroom, and despite the vary stark warnings given by previous cases, solicitors are continuing to over rely on artificial intelligence to their detriment.
The latest law firm to experience this is Pinsent Masons, which has self-referred to the SRA following an engagement with a judge at within the Insolvency and Companies List in the High Court.
Background
It started when Pinsent Masons responded by letter to an enquiry the court had made regarding a rule that would allow its block transfer application. On 30 March 2026 Pinsent Masons wrote to ICC Judge Mullen informing him of a court rule which would supposedly allow the application – this supposedly being Rule 12.37(5) of the Insolvency (England and Wales) Rules 2016. This rule was quoted within the letter.
When ICC Judge Mullen looked up this rule and saw it was incorrect, he responded to Pinsent Masons stating that the rule referenced had been misquoted and asked where the specific wording quoted could be found. Pinsent Masons responded on 14th April 2026 stating that whilst the words do not appear within the rule, the wording was “a summary conclusion, drawn from reading provisions of Rule 12.37.”
However, as the first letter had clearly identified the rule as a direct quotation, this further explanation was also incorrect. Suspecting AI hallucination because of these incorrect explanations, the judge ordered a hearing and witness statements from the writer of the letters and the Partner involved.
The witness statements given by Pinsent Masons confirmed that AI had been used to produce the letters and that there had been a failure to supervise an un-named junior lawyer’s use of AI in producing the letters.
Pinsent Masons filed as evidence the transcripts of the chats that this junior lawyer had with the AI, running to some 59 pages. The Judge noted from these transcripts that the AI had begun hallucinating immediately, resulting in rules and legislation that were quoted incorrectly. The judge found that the junior lawyer asked the AI what the legislation said rather than looking it up themselves from a reliable source and that had they done so, they would have recognised the AI was producing incorrect information.
Further to this, when it came to the 14th April letter, the judged noted from the transcript that the AI recognised that the quote of Rule 12.37(5) had been incorrect and drafted a letter for the junior lawyer acknowledging the mistake. However, the lawyer asked the AI to remove any reference to an apology, bringing into question their judgment.
In his final remarks, the judge said that all three solicitors involved, including the junior lawyer, should have been aware of the dangers of using AI to conduct legal research. He stated that the junior lawyer had used AI to research a legal point and the AI invented the quotation. The junior associate did not check this and neither did the solicitors supervising them. Even once the mistake had been acknowledged by AI, the junior lawyer didn’t inform his supervisors.
The judge went on further to say that whilst the two more senior solicitors did not intend on misleading the court, there was a failure by both to supervise the junior lawyer and check the text of any letters before they were sent out.
However, the Judge, noting Pinsent Mason’s frank witness statements provided to the court, the accompanying evidence and the minimal loss suffered by the client, decided it was disproportionate to consider any contempt proceedings. He therefore decided that the public admonishment of the law firm and the referral to the SRA was the proportionate response.
Conclusion
So what does this case tell us?
It’s the first time that the internal chats between a junior solicitor and an AI bot have been quoted in a judgment and the revelations are stark, being that the lawyer essentially outsourced his critical thinking to an AI bot. This will have been an extremely embarrassing judgment to be named in for all solicitors involved, and it certainly raised some important points regarding AI use in general.
First, if you are ever asking AI a legal question, always check the data from which it’s derived to ensure it is not a hallucination.
Second, if you are using AI in your business, make sure all your staff are properly trained on using it in the correct way without over reliance. Check whether any documents you are working on may have been generated by AI and make sure you have robust processes in place to ensure proper human oversight at all times.
There’s no doubting that, in experienced hands, AI can be an excellent drafting aid, helping professionals produce first drafts, organise information, identify issues for further review and sense-check their thinking quickly and efficiently.
AI works best in the hands of experienced professionals who understand both its strengths and its limitations. It can assist with a task, but it cannot take responsibility for it. It can provide a useful starting point, but it does not provide definitive answers or replace proper research and verification.
Most importantly, AI cannot replace professional judgment, experience or critical thinking. The value of any AI-generated output depends entirely on the person reviewing it. An experienced professional will know when something looks wrong, when further investigation is needed and when a conclusion does not stand up to scrutiny.
The lesson from this case is not necessarily that AI should be avoided. Rather, it should be treated in the same way as any other tool – useful, efficient and capable of improving productivity, but only when accompanied by proper oversight, expertise and accountability. As this judgment demonstrates, the risks arise when people stop treating AI as an assistant and start treating it as an authority.
