Case Context: The Upper Tribunal Authority
Promulgated on 17 November 2025 by Upper Tribunal Judges Lindsley, Keith, and Blundell, Munir [2026] UKUT 81 (IAC) represents the leading judicial authority in England and Wales on solicitor supervision of generative AI tools.
While the underlying litigation arose in immigration proceedings, the Tribunal made it explicitly clear that the judgment was not an isolated finding on advocacy errors. At paragraph [37], the Tribunal observed:
"It would be easy to think that this is a case about the naive use of generative AI, but it is not merely about that; it is principally about supervision."
The Core Holdings That Govern Law Firm Operations
1. Supervisor Culpability Exceeds Drafter Error (Paragraph [38])
The Tribunal established an unambiguous standard of management accountability:
"A supervisor who fails to ensure that the work of a more junior fee-earner does not contain false cases or citations is likely to be more culpable than a lawyer who fails to ensure that his own work is free from such 'hallucinations'."
For conveyancing partners and COLPs, this single sentence shifts the regulatory risk. If a paralegal or junior conveyancer uses an unapproved generative tool to review a leasehold management pack or draft a Report on Title, and the supervising solicitor signs off without independently scrutinising the reasoning and sources, the regulator will hold the supervisor primarily responsible for the supervisory failure.
2. The "We Don't Use AI" Defence Is Dead (Paragraph [46])
In Munir, the firm's COLP sought to absolve management of responsibility by claiming there was no mechanism by which his staff could use AI, as the firm had not purchased or licensed AI software. The Tribunal firmly rejected this defence:
"That is to overlook the fact that anyone with access to Google has access to AI."
The Tribunal found the COLP had taken no proactive steps to train, warn, or govern staff regarding AI risks. As a result, the supervising solicitor and COLP was formally referred to the Solicitors Regulation Authority (SRA) at paragraph [56].
3. Permanent Loss of Legal Privilege & Confidentiality Breach (Paragraphs [21] & [60])
The Tribunal held that inputting client matter documentation into open-source or public consumer-grade LLMs (such as free tiers of ChatGPT) constitutes an irremediable breach of confidentiality under SRA Code of Conduct Paragraph 6.3. Because consumer platforms typically ingest user data to train future foundational models, the information enters the public domain, causing an irreversible waiver of legal professional privilege.
Operational Translation: The Conveyancing Supervision Protocol
How does a residential conveyancing practice protect its partners and COLP from the liabilities outlined in Munir? UtterConnection establishes three non-negotiable operational controls:
Protocol 1: Zero Public AI on Client Data (DLP Hard Boundaries)
Specify Microsoft Purview Data Loss Prevention (DLP) controls within the firm's Microsoft 365 tenant to restrict client property packs, TR1 forms, and financial documentation from being pasted into unapproved browser tools.
Protocol 2: The Marked Draft Standard
Mandate that no AI-assisted draft reaches a supervisor without an explicit tag, the stated prompt objective, and at least one identified open legal query. This forces active scrutiny and eliminates superficial rubber-stamping.
Protocol 3: Contemporaneous Supervisory Logging
Maintain an immutable, timestamped sign-off log inside the case management system recording which qualified practitioner checked the draft, what sources were verified, and when the advice was approved.
Conclusion: Oversight Is a Process, Not a Policy
Munir [2026] UKUT 81 made clear that having an AI policy sitting on an intranet page does not satisfy a solicitor's supervisory obligations. When a claim or regulatory audit occurs, the firm must be able to demonstrate that supervision was operational, active, and contemporaneous.
Is Your Practice Protected Against Munir Liabilities?
Evaluate your firm's supervisory controls, shadow AI exposure, and COLP risk profile in under five minutes.
Take the Free AI Reality Check