In partnership with
The regulatory story this week is about accountability rather than capability. The SRA has published updated supervision guidance following the Court of Appeal's decision in Mazur, and for the first time it folds AI-assisted work into the same supervision framework that governs any delegated task: an authorised person stays responsible, and AI output has to be checked by a human before anyone relies on it. Two further developments emphasise this point, a criminal investigation into a police officer accused of using AI to create evidence, and survey data suggesting most clients have no idea their lawyers are using AI at all.
AI in Practice
The SRA brings AI-assisted work inside its supervision rules
On 12 June 2026 the SRA published updated guidance on effective supervision, following the Court of Appeal's judgment in Mazur v Charles Russell Speechlys [2026] EWCA Civ 369. The court held in March that an unauthorised person (a paralegal, a legal executive, or anyone without a current practising certificate) may lawfully carry out the conduct of litigation provided an authorised person exercises proper direction, management, supervision and control throughout. That restored the delegation model most firms had operated for years, after a 2025 High Court ruling had thrown it into doubt. The revised guidance turns that principle into what firms must now be able to demonstrate, and it extends the same logic to AI.
The guidance treats AI as another form of delegated work. Where AI assists in producing legal work, an authorised individual keeps ultimate responsibility, and the output must be subject to appropriate human review, scrutiny and professional judgement before it is used. That is less of a new duty and more a restatement of an old one, but it is the first time the SRA has written it into its supervision guidance, and it arrives in the same month as the Cork v Smith admonishment (covered here on 29 May), in which a junior at Pinsent Masons relied on an AI-fabricated insolvency rule that nobody checked.
Anyone hoping for a detailed AI rulebook should lower their expectations. Gordon Exall, who has written the guidance up at length, points out that it gives only two short paragraphs to AI (Civil Litigation Brief). The SRA's position, in the words of its executive director Jonathan Peddie, is that "firms should take a risk-based approach tailored to their circumstances", supported by worked case studies rather than prescriptive rules (the trade press puts the expanded guidance at 24 pages, up from nine). For most firms the message is narrow and rather old-fashioned: supervise AI as you would supervise a capable but unreliable junior, and be able to show how you did it.
Takeaways
Act: Make AI use visible to supervisors. If your sign-off process does not record that AI-assisted work was checked against primary sources by a named, authorised person, change it so it does. The SRA's worked case studies are the closest thing to a template.
Watch: Whether the SRA expands its pool of case studies, as it has said it will, and whether it ever issues anything more specific on AI than these two paragraphs. The Law Society has been pressing for AI-specific guidance since last autumn.
Risk: After Mazur, supervision is the thing that makes delegated and AI-assisted work lawful, not a background virtue. Weak supervision of AI output is now squarely a conduct issue, not only a quality one.
On your radar
A police officer faces a criminal investigation over AI-generated evidence: Derbyshire Constabulary has referred one of its officers for criminal investigation into perverting the course of justice, after allegations that AI was used to create evidential material in a number of cases (first reported by the Financial Times on 12 June). The officer has been removed from frontline duties, no arrests have been made, and the force says it is working with the CPS on potentially affected cases. It is thought to be the first case of its kind in the UK criminal justice system, and it follows National Police Chiefs' Council advice that forces should not use AI to prepare court statements. Why it matters for UK lawyers: AI-fabricated material is no longer confined to invented citations in skeleton arguments, it is now alleged to have reached the evidence itself. For anyone conducting or defending litigation, the provenance of documents and statements produced by an opponent or a public authority is becoming a live question. Where evidence could have been AI-generated, ask how it was produced before you rely on it. (ITV News)
Nearly nine in ten lawyers use AI, but their clients have not been told: Clio's inaugural UK and Ireland Legal Insights Report, based on surveys of more than 500 legal professionals and 500 members of the public, finds AI use is now near-universal in firms, with 70% of users having adopted it in the past year. The figure that stands out is disclosure: 81% of firms say they tell clients about AI use at least occasionally, but only 7% of clients recall being told, and 79% of the public think lawyers should disclose it. The report also finds 17% of firms still have no AI policy. This is vendor research and should be read as such, but the sentiment is hard to dismiss. Why it matters for UK lawyers: disclosure is becoming a question of client trust as much as compliance, and the BSB has already said transparency is required where AI materially affects the service. Check whether your engagement terms or client-care letter say anything about AI, and if your firm is in the 17% with no policy, that is the more pressing gap. (Clio | Legal Futures)
Thomson Reuters brings agentic AI into Westlaw and Practical Law in the UK: Thomson Reuters has launched CoCounsel Legal in the UK, embedding agentic "Deep Research" capabilities directly into Westlaw and Practical Law rather than selling them as a separate product, according to trade reporting. The pitch is that legal research, document search and drafting assistance now sit inside tools many firms already license. Why it matters for UK lawyers: if your firm subscribes to Westlaw or Practical Law, capable AI features may be arriving in your existing stack without a fresh procurement decision, which is convenient and also a supervision question. Ask your library or knowledge team what is now switched on, and re-test research and citation-checking workflows before anyone relies on agentic output in a filing. (Non-Billable)
Perplexity becomes the fifth Big Tech firm to target legal: Perplexity is moving into legal in earnest, tied to its agentic platform Perplexity Computer, and will pitch the sector at a "Computer for Counsel" event in New York next week. It joins OpenAI, Anthropic, Microsoft and Palantir in targeting the legal world. Connectors to Box, DocuSign and the legal-research tool Midpage are already live, with LexisNexis, Clio and Ironclad said to be on the way, and the company is targeting both smaller firms and in-house teams. Its marketing leans hard on citations and accuracy, an explicit pitch at the hallucination problem. Why it matters for UK lawyers: this is an event and a roadmap rather than a shipped UK product, but Perplexity is already a popular consumer research tool, so some of your fee earners may already be putting legal questions to it without telling anyone. The accuracy statement does not remove the verification duty the SRA has just restated. Treat it like any other AI tool: know who is using it, on what, and how the output gets checked. (Artificial Lawyer)
Ad Break
In order to help cover the running costs of this newsletter, please check out the advert below. In line with my promises from the start, adverts will always be declared.
Become An AI Expert In Just 5 Minutes
If you’re a decision maker at your company, you need to be on the bleeding edge of, well, everything. But before you go signing up for seminars, conferences, lunch ‘n learns, and all that jazz, just know there’s a far better (and simpler) way: Subscribing to The Deep View.
This daily newsletter condenses everything you need to know about the latest and greatest AI developments into a 5-minute read. Squeeze it into your morning coffee break and before you know it, you’ll be an expert too.
Subscribe right here. It’s totally free, wildly informative, and trusted by 600,000+ readers at Google, Meta, Microsoft, and beyond.
For Review
The new SRA guidance on effective supervision 2: guidance and checklists on the use of artificial intelligence (Civil Litigation Brief)
Gordon Exall's walkthrough of the supervision guidance, with the candid observation that the SRA devotes only two paragraphs to AI, and his own set of practical checklists to fill the gap. The most useful companion to the lead story if you are turning the guidance into something your fee earners can actually follow (the full text sits behind Civil Litigation Brief membership).
Read or listen: Civil Litigation Brief
Legal AI: Battle lines (Law Gazette)
Sam Kenworthy's feature on the state of the legal AI market, surveying the contest between the big model providers, the specialist legal tech vendors and the firms building their own tools, and arriving at the same conclusion the courts keep reaching: the technology is useful, but it has to be checked. A good orientation piece if you want the commercial picture behind the regulatory noise.
Read or listen: Law Gazette
Practice Prompt
Try the below prompt to build a supervision and sign-off record for a piece of AI-assisted work, which fits a week in which the SRA has made supervision the thing that makes AI use lawful. Ensure you fill in context and constraints and other aspects marked with {}. Remember to adhere to the Golden Rules and do not upload confidential or privileged information to public tools.
You are a legal supervision assistant. Your task is to help a UK solicitor create a documented supervision and sign-off record for a piece of AI-assisted work, consistent with the SRA's effective supervision guidance and the principle in Mazur that an authorised person must exercise direction, management, supervision and control.
The work:
- Task: {e.g., "first-draft witness statement" / "legal research note on a limitation point" / "contract review and redline" / "letter before action"}
- Tool used: {e.g., "Claude for Legal" / "CoCounsel in Westlaw" / "Microsoft Legal Agent" / "ChatGPT (non-confidential inputs only)"}
- Who did the work: {e.g., "trainee solicitor" / "paralegal" / "fee earner, 2 years PQE"}
- Supervising authorised person: {name and role}
- Client matter type and sensitivity: {e.g., "contentious probate, privileged material" / "commercial contract, no personal data"}
Produce:
1. Supervision map
For this task, set out what "direction, management, supervision and control" requires in practice: what the supervisor should specify before the work starts, what they should check during it, and what they must review before sign-off. Be specific to the task type above, not generic.
2. Human review checklist
List the verification steps a human must complete before the AI-assisted output is relied on, including:
- Every legal authority, statutory provision and rule checked against the primary source
- Every factual assertion traced to a document on the file
- Any quotations confirmed accurate
- Any calculations independently checked
Flag which steps cannot be delegated to the same person who used the AI.
3. Disclosure decision
Advise whether the use of AI on this task should be disclosed to the client or the court, applying the position that disclosure is required where AI materially affects the nature or scope of the service. Give a short reasoned recommendation and a draft line of wording if disclosure is appropriate.
4. Record
Produce a short sign-off record (date, task, tool, fee earner, supervisor, checks completed, disclosure decision) that could be saved to the matter file as evidence of supervision.
Constraints:
- {Add firm-specific constraints, e.g., "we are SRA-regulated" / "all checks must be evidenced on the file" / "no client data may be entered into the tool"}
- Apply the regulatory framework of England and Wales throughout
- Do not invent SRA rules, case law or product features. Where the position is uncertain, say so and recommend what to confirm
- This is a supervision-planning aid, not legal adviceHow did we do?
Hit reply and tell me what you would like covered in future issues or any feedback. We read every email!
Thanks for reading,
Serhan, UK Legal AI Brief
Disclaimer
Guidance and news only. Not legal advice. Always use AI tools safely.
Recommended Newsletters
Below are a few newsletters that I recommend, for various reasons. Check them out!
