Firms Urged to Heed Warning Over AI Use

In the space of just a few weeks, the Law Society Gazette has reported on two further sets of law firms and their staff being referred to the Solicitors Regulation Authority (SRA) after fake legal authorities, suspected to have been generated by AI, were placed before the court. In one case, a small West Midlands firm in a civil appeal application. In the other, international firm Pinsent Masons and three of its solicitors, following what Insolvency and Companies Court Judge Mullen described as “astonishing” AI failures that twice misled the court.    

That two such cases have emerged in quick succession, involving firms of very different sizes, should leave no doubt that judges across England and Wales are hardening their position on the misuse of generative AI in the preparation of court documents.

Submission of AI-generated Authorities — AML Legal

In the first case, His Honour Judge Grimshaw, who handed down the decision in Rodney v Gee’z Micro Bar & Pitstop, said the conduct of Kossar Qureshi, director of West Midlands firm AML Legal, and Mahmood Hussain, a consultant at the firm, was serious enough that the regulator needed to consider whether to take further action. HHJ Grimshaw made it clear that warnings from the bench were no longer a sufficient response to the growing problem of hallucinated case references being put before the court.

It was shown that West Midlands firm, AML Legal, submitted three documents in support of a client’s application for permission to appeal in a civil dispute. The documents contained incorrect citations of legal authorities, leading the court to suspect they had been produced with the assistance of generative AI. One of the documents had been signed with a statement of truth in Hussain’s name.

Hussain told the court that the documents had been prepared with the help of a paralegal. However, he accepted he was responsible for the conduct of the case and should have qualified the legal research before allowing it to be filed. Qureshi, as director, had signed off the email that filed the bundle but tried to argue that, as the firm’s principal, her name was automatically added through the firm’s case management system.

However, the judge did not accept this explanation, saying, “given that most of the documents filed do not in fact have Ms Qureshi’s name on them, nor is she identified as the principal of the firm on the original skeleton argument.”

Qureshi accepted the seriousness of the concerns raised and confirmed that she has reviewed the firm’s internal procedures to ensure safeguards were in place.

While the judge accepted Hussain was genuinely remorseful, he declined his plea not to be referred to the regulator.

In stressing the seriousness of this growing issue, the judge said,

‘Misleading material was placed before the court in terms of mis-cited legal authorities, said to support propositions that they simply did not, when even the most simple of checks would have shown that not to be the case, or shown it to be the case, i.e. that they were incorrect,’

HHJ Grimshaw added,

‘That is inexcusable on the part of a professionally qualified lawyer. The fact that one document was endorsed with a statement of truth signed by that same solicitor makes that criticism all the more valid, in my judgment.’

Source: The original Law Gazette report can be read here: ‘Judge refers solicitors to SRA over ‘inexcusable’ submission of AI-generated authorities’

A Second Case — Pinsent Masons and the “Cavalier Attitude” to Accuracy

In the latest example, the Law Society Gazette has reported that international firm Pinsent Masons had self—referred to the Solicitors Regulation Authority (SRA) after Insolvency and Companies Court Judge Mullen found the firm had misled the court on two separate occasions during an otherwise routine block transfer insolvency application in Cork & Anor v Smith.

The firm cited an insolvency rule that did not exist, with purported statutory wording attached. When the judge queried it, the firm sent a follow-up letter, also drafted with the assistance of AI, which the judge found again misapplied the law and “muddied the waters.”

The judge was particularly critical of the junior solicitor responsible for the research (referred to as LA), who, in his words, seemed to have “almost entirely outsourced the thinking process” to an AI program. The supervising solicitors were also held to account for failing to supervise LA properly.

While it was acknowledged that solicitors and junior legal teams are under considerable pressure at times, the judge added, “None of that excuses a failure to check the accuracy of the material that was placed before the court.”

Pinsent Masons has apologised unreservedly and confirmed it is “taking steps to strengthen our processes and oversight to ensure this does not happen again.”

Source: The original Law Gazette report can be read here: Pinsents refers itself to SRA over AI failures

Reliance on Generative AI is Part of a Growing Issue for Law Firms

These latest reports are not one-offs and follow a series of rulings that highlight a growing issue within the legal industry. It is not just because of misleading materials presented to the court, but also warnings over data protection when loading information into open-source AI.

In our previous article, ‘Open-Source AI and Client Data: What Your Firm Must Do Now’, we covered the judgment published by the Upper Tribunal Immigration and Asylum Chamber that highlighted concerns about the use of open-source artificial intelligence tools in legal practice, warning lawyers against putting client documents into ChatGPT.

Our earlier guidance note set out the dangers of uploading client documents to open-source AI platforms such as ChatGPT, the practical difference between open and closed-source models, and the steps every firm should take to bring its AI use policy up to scratch.

The wider message from that judgment, that supervisors are personally responsible for checking the work of those they oversee, has now been reinforced by HHJ Grimshaw in unambiguous terms.

While AI use is not being outright discouraged, firms should heed these warnings if their teams rely on generative AI. If AI is used for the preparation of documents placed before a court, and the authorities cited have not been independently verified, or there are concerns that data protection rules have been breached, the courts are increasingly willing to take a robust approach.

It is now more likely that the court will refer the matter onward to the regulator rather than waive the issue with a slap on the wrist and a warning from the bench.

Using Generative AI? What Your Firm Should Do Now

For firms that have already reviewed their AI use policy and briefed their staff in line with our earlier guidance, Rodney v Gee’z should serve as confirmation that the right course has been taken. For firms that have not, the position should now be treated as urgent.

The core lessons remain the same as those set out in our earlier note:

• Any document containing legal authorities or case citations must be independently verified before filing. As the judge noted, the simplest checks would have caught the errors in this case.

• A solicitor who delegates legal research to a paralegal or junior fee-earner remains personally responsible for the accuracy of the output. Use of a paralegal is not a defence.
• Statements of truth on documents containing AI-generated content carry significant personal regulatory risk. The judge in Rodney v Gee’z specifically cited the statement of truth as making the criticism more valid.
• Open-source AI tools should not be used with client data in any circumstances, regardless of any qualifying language in current firm policies. Closed-source enterprise tools, supported by appropriate data processing agreements, remain the only suitable option for tasks involving client information.
• If past uploads to open-source tools are known to have occurred, consider reporting obligations to clients, the SRA, and the ICO.
• Supervisors who fail to ensure that fee-earners under their oversight are aware of AI risks may be treated as more culpable than the fee-earner who makes the error itself.

For a fuller explanation of the policy wording firms should review, the practical steps for disabling data sharing on consumer AI tools, and our recommended approach to all-staff briefings, please refer to our earlier note: Open-Source AI and Client Data: What Your Firm Must Do Now.

Further Compliance Support

If your firm requires assistance reviewing, drafting, or updating your AI use policy, assessing your current tool landscape, or preparing staff communications and training, please contact your RACO compliance manager.

To ask any questions about the Regulation and Compliance Office and how we can support your Law Firm, please call us on 0203 985 8550, complete the online enquiry form below, or email us info@regulationandcomplianceoffice.co.uk