Pinsent Masons refers itself to SRA over AI failures

The deployment of artificial intelligence in professional services has hit a significant regulatory stumbling block. Pinsent Masons, a prominent international law firm, has self-referred to the Solicitors Regulation Authority (SRA) following a highly embarrassing incident where AI-generated errors were submitted to the court not once, but twice.

The case, Cork & Anor v Smith, overseen by Insolvency and Companies Court (ICC) Judge Mullen, serves as a stark warning. It is one of the first instances of a major firm reporting itself to a regulator over AI failures, and it provides a direct read-across for any RICS or ICE-regulated consultancy deploying AI on client deliverables.

The Anatomy of the Error

The incident began when a junior lawyer, referred to in court as 'LA', used an AI program to draft a response to a query from the judge regarding a block transfer application. The initial response cited a rule concerning the court's power that the judge could not verify. Recognising it as an AI hallucination, the judge demanded an explanation.

Astonishingly, the follow-up letter, also drafted using AI by the junior lawyer, misapplied the law again and attempted to "muddy the waters" rather than providing a candid explanation. The court heard that the AI had even warned the junior lawyer that it was not "fully confident" in its interpretation of the law, a warning that was ignored.

Judge Mullen's assessment was scathing. He noted that the junior lawyer seemed to have "almost entirely outsourced the thinking process" to the AI. Furthermore, he highlighted a critical failure in supervision by the senior staff, solicitor Samantha Poulton and partner Steven Cottee, who approved the flawed drafts.

The Failure of Supervision

The crux of the issue lies not just in the use of AI, but in the lack of robust oversight.

❝

As Judge Mullen articulated, "Ms Poulton, Mr Cottee and, I will assume, LA, will have found this incident very embarrassing. Ms Poulton and Mr Cottee are experienced solicitors undertaking technical work in a reputable firm, no doubt under considerable pressure at times. I have no doubt that LA was working under pressure too. None of that excuses a failure to check the accuracy of the material that was placed before the court."

The senior lawyers were unaware that AI had been used to draft the responses, revealing a significant gap in internal communication and quality control processes. The judge emphasised that it was incumbent upon senior staff to ensure that junior team members understood the necessity of verifying AI-generated content and being candid about its use.

What is particularly striking about this case is the speed at which the junior lawyer moved from drafting to submission. Both Cottee and Poulson checked the first draft letter within approximately two hours, confirming it was acceptable without conducting any substantive verification of the claims. This rushed approval process, coupled with the absence of any inquiry into how the content was generated, represents a systemic failure of professional diligence.

The Illusion of Competence

This case perfectly illustrates the danger of treating Large Language Models (LLMs) as authoritative sources of truth rather than sophisticated text generators. AI can produce highly convincing, well-structured arguments that are entirely factually incorrect.

❝

Judge Mullen addressed this directly: "AI has the potential to be wholly unreliable. AI may, of course, provide a jumping-off point for research and legal reasoning, but it does not, at least at present, do away with the need for proper research and thought on the part of a legal professional, even a very junior legal professional."

For project delivery professionals, the parallels are obvious. Whether drafting a legal argument, a technical specification, or a cost estimate, the reliance on AI without rigorous human verification is a recipe for professional negligence. The case serves as a cautionary tale for any firm deploying AI-assisted workflows on client-facing deliverables. The confidence with which AI generates plausible-sounding content can lull teams into a false sense of security, leading them to skip critical verification steps.

Strategic Implications for Project Delivery

We must view this incident as a critical governance wake-up call. If a top-tier law firm can stumble so publicly, engineering and construction consultancies are equally vulnerable. The integration of AI into workflows must be accompanied by stringent quality assurance protocols and clear policies regarding its acceptable use.

Firms must establish frameworks that dictate when and how AI can be utilised, mandating human oversight for any output that impacts client deliverables or regulatory compliance. The assumption that senior staff can simply skim-read AI-generated documents is dangerously flawed; verification must involve checking the underlying facts and calculations, not just the grammar and formatting.

Takeaway

• Supervision is non-negotiable: Senior professionals cannot abdicate their responsibility to verify work simply because it was generated by a machine; rigorous human oversight is mandatory.

• Establish clear AI usage policies: Firms must implement explicit guidelines detailing when AI can be used, requiring junior staff to be transparent about its application in their work.

• Beware the illusion of competence: AI models excel at producing confident-sounding text that may be factually incorrect; all outputs must be treated as drafts requiring independent verification.

• Governance must evolve: Existing quality assurance processes must be updated to specifically address the unique risks associated with AI-generated content, ensuring regulatory compliance is maintained.

Learn the lessons

Are your internal governance structures robust enough to prevent an AI-generated hallucination from reaching your clients? Subscribe to the Project Flux newsletter to understand how to build resilient AI quality control processes for project delivery.