A year ago today I sent the welcome email for this newsletter. At that point, almost nothing had been said specifically about AI in legal practice in England and Wales. There was judicial guidance written for judges, a Bar Council note, a High Court warning in Ayinde, and a great deal of implication. One year on there is a warning notice, a draft competence standard, a final report from the Civil Justice Council, guidance from four regulators, and a growing body of case law.
It is a patchwork, and spread across a dozen websites. This issue puts it on one page: body by body, what they have said, how much it binds you, and what is still an open question. It is a snapshot as at 17 November 2026, and I will link back to it as things change.
The key
Binding: legislation in force, or a judgment which binds the court you are in.
Regulatory weight: not a new rule, but your regulator has said it will have regard to it, and ignoring it puts you at risk of disciplinary action.
Guidance: explains existing duties, or sets out good practice. Persuasive, not enforceable in itself.
Proposal: consultations, recommendations and announcements. Nothing changes until someone acts on them.
If you only have time for one line, it is this: no regulator has created a separate rulebook for AI, and every one of them has said your existing duties apply in full. What has changed this year is how clearly they have said so, and how often the courts have shown what happens when those duties are not met.
What it has said: the SRA's "Misuse of AI" warning notice of 17 August 2026 applies to every firm and individual it regulates. You remain accountable for your work "whether or not AI has been used in its preparation"; where false citations reach the court, "reliance on an output of AI would not be a suitable defence"; and uploading to an open tool means privilege "may be permanently waived and unable to be recovered" (SRA). Its effective supervision guidance, updated on 12 June, says an authorised individual "retains ultimate responsibility for any legal services delivered with AI assistance" (SRA). The AML sectoral risk assessment, updated on 4 September, now flags AI-enabled impersonation, including deepfakes, at onboarding (SRA). And on 8 October it opened a consultation proposing to write AI into the Statement of Solicitor Competence: checking "the outputs from technology including AI, by reference to primary resources", and a new supervision competence covering "clear oversight of the work being done, including the use of technology and AI" (SRA).
How binding: the warning notice carries regulatory weight: "If you fail to have proper regard to this warning notice, you are at risk of disciplinary action." The supervision guidance is guidance, but the warning notice leans on it. The AML risk assessment is one firms must have regard to under regulations 18 and 18A of the Money Laundering Regulations 2017. The competence changes are a proposal: the consultation closes at 23.59 on 3 December, with final versions expected in April 2027 and effect from September 2027.
Still open: whether AI can itself amount to "conducting litigation". The Law Society asked the SRA for guidance on this in November 2025. The Court of Appeal's judgment in Mazur does not mention AI, the SRA's statement on it does not either, and as far as this writer can find the SRA has not answered the question. The SRA said when it published the warning notice that a number of investigations into AI misuse were ongoing. No outcomes have been published.
Law Society
What it has said: on 21 October it published a practice note on using AI in legal practice. Using AI within a firm's secure systems, where data is not used for training, "is not likely, of itself, to require informed consent", and the expected approach is transparency. Using confidential client information, including to train AI, "is likely to require specific client consent". Firms billing by the hour should not bill for the time a task would have taken had AI not been used (Law Society, Legal Futures). Its 6 October report on agentic AI in the justice system put its position in one line from vice president Brett Dixon: "Legal judgement cannot be delegated" (Law Society Gazette).
How binding: guidance from the representative body, not regulation. It is the kind of document a court or the SRA may look to when asking what a reasonable solicitor would have done, and this author would treat the consent and billing points accordingly.
Still open: the AI and Mazur question above, which is the Law Society's own outstanding request.
Bar Council and Bar Standards Board
What they have said: the Bar Council's ethics note on generative AI was last reviewed in November 2025 (Bar Council Ethics). The BSB's guidance, valid from 18 May 2026, says free, general-purpose tools "will usually be too high risk to use for legal work or client-specific tasks", and that you "must be transparent about your use of AI with your client if it materially impacts the nature or scope of your legal service". Disclosure to the court is not generally required, but you must answer honestly if the court or the BSB asks (BSB, PDF).
How binding: guidance, but guidance which "explains how existing duties and rules under the BSB Handbook apply", and those duties are binding. The Bar Council note is guidance from the representative body.
Still open: very little for the Bar specifically. Barristers face the same open questions as everyone else, set out under the CJC and the courts below.
CILEx Regulation and the Council for Licensed Conveyancers
What they have said: CILEx Regulation published AI guidance on 1 October 2026: AI-generated output "must be reviewed, verified and approved by an appropriately competent individual", and "responsibility cannot be delegated" (CILEx Regulation). The CLC has published AI and technology guidance built on 11 principles, including transparency about technology use and human review of technology-supported outputs (CLC).
How binding: guidance for their own regulated communities. Worth a look if you supervise CILEx members or work with licensed conveyancers on the other side of a transaction.
Still open: the CLC is one of the four regulators in the Legal Services AI Growth Lab (below), alongside the SRA, the LSB and the ICO.
Judiciary
What it has said: the AI judicial guidance of 31 October 2025 is still the current version. It tells judges that where a chatbot appears to have been used it is appropriate to ask what checks for accuracy were done, and that they may need to ask lawyers to confirm they have independently verified what is before the court (Judiciary). On 9 October a position statement on the judiciary's own use of AI described "a cautious but positive approach", with "every AI output used in judicial work" to be "reviewed, verified and approved" by the responsible judge, and only approved tools to be used (Judiciary). In April the then Chancellor, Sir Colin Birss, said privilege "would not seem to attach" to material put into public AI systems (Judiciary). He became Master of the Rolls on 1 November.
How binding: guidance, written for judges, not for you. It still tells you what the judge in front of you has been told to ask. The speech is not law, and he framed his points as live issues.
Still open: Sir Colin Birss proposed in October a limit of about five pages for a litigant in person to state their case ("I propose to do it"), but no formal step has followed (Legal Futures). The position statement also gives a reasonable benchmark for approving your own firm's tools: UK data residency, enterprise data protection and web access disabled.
Civil Justice Council and the rule committees
What they have said: the CJC's final report on AI and court documents, published on 29 October, makes 15 recommendations (CJC final report, PDF). There should be no new declaration for statements of case, skeleton arguments or advocacy (Rec 4), no duty to name the tools used (Rec 14) and no permission regime (Rec 15). But a witness of fact should have to declare that AI used in preparing their statement "has not generated or influenced its substantive content" (Rec 6), whether or not the party is represented, and experts should make a similar declaration (Rec 9). The Civil Procedure Rule Committee should consolidate the case law on AI use and sanctions in a plain-language practice direction (Rec 3), and guidance for litigants in person should be considered "as a matter of urgency" (Rec 2). On the criminal side, the Criminal Procedure Rule Committee's minutes for 24 July record a joint AI working group across the procedure rule committees, and the formulation of "potential requirements for parties to disclose such use of artificial intelligence as may have been made in preparation for trial" (CrimPRC minutes, PDF).
How binding: proposals. Nothing in the CJC report is in force, and it sets no timetable. The Online Procedure Rules, by contrast, are binding: in force since 7 September 2026 and so far reaching possession claims only (legislation.gov.uk). The Online Procedure Rule Committee's voluntary code on responsible AI is still in preparation.
Still open: when the CPRC acts, and the final wording of the witness declaration. That is the change most likely to alter day-to-day litigation practice, and the line between AI which tidies a transcript and AI which "influences" evidence is not yet drawn.
Courts: the cases that matter
What they have said:
Ayinde [2025] EWHC 1383 (Admin), Divisional Court: where false citations reach the court, referral to the regulator is likely to be appropriate (para 29). The SRA now cites it (National Archives).
Mazur [2026] EWCA Civ 369, Court of Appeal, 31 March 2026: unauthorised staff may carry out litigation tasks under the supervision of an authorised individual who retains responsibility. The judgment does not mention AI (National Archives).
UK v SSHD (Munir) [2026] UKUT 81 (IAC): uploading confidential documents to an open AI tool is to "breach client confidentiality and waive legal privilege" (para 60), an observation in a decision about fake citations (National Archives).
Cork v Smith [2026] EWHC 1199 (Ch): AI invented the wording of an insolvency rule, a junior solicitor sent it to the court unchecked, and the supervisor did not know AI had been used. Public admonishment and a self-report (National Archives).
R v FGD [2026] EWCA Crim 918: a witness used AI to rehearse cross-examination. A retrial was ordered because the exchange was fully recorded, but "the coaching of witnesses in their evidence, by any means, is not permitted" (National Archives).
Hancox v Sutherland [2026] EAT 139: "Simply submitting the product of generative AI to a court or tribunal, or placing the onus to check a document onto an opponent, is not acceptable" (para 33) (GOV.UK).
Cornwall Council v M [2026] EWFC 306: an expert's kinship assessment was "AI generated from her notes". The judge found "there has not been adequate human oversight" and placed no significant weight on it (National Archives).
How binding: Mazur (Court of Appeal) and FGD (Court of Appeal, Criminal Division) are binding on the courts below. Ayinde is a Divisional Court decision which the regulator has adopted. The others are first-instance or tribunal decisions, persuasive outside their own jurisdictions, but they show how judges are actually responding.
Still open: no later judgment has yet considered Munir on privilege, so the point has not been tested in the civil courts. No case has yet dealt with AI and the conduct of litigation.
Information Commissioner's Office
What it has said: under SI 2026/425, in force since 12 May 2026, the Commissioner "must prepare" a statutory code of practice on AI and automated decision-making (legislation.gov.uk). The new rules on automated decision-making in UK GDPR (Articles 22A to 22D, from the Data (Use and Access) Act 2025) have been in force since 5 February 2026 (legislation.gov.uk). On 8 October the ICO opened a call for evidence on agentic AI, which will inform future guidance and the statutory code (ICO). The Upper Tribunal in Munir said that uploading client documents to an open AI tool should be referred to the ICO.
How binding: the ADM rules are binding law. The duty to produce the code binds the Commissioner, but there is no code yet: no draft has been published and no date given. The ICO's draft ADM guidance (consultation closed 29 May) is not yet final.
Still open: the agentic AI call for evidence closes this Friday, 20 November. If your firm is piloting agents which act across email and case files, that is the place to say what you need. The final ADM guidance is expected "winter 2026".
Government and Parliament
What they have said: there is no AI bill. The King's Speech in May promised a Regulating for Growth Bill with powers to relax rules temporarily to test AI products, but it has not yet been introduced. On copyright, the government dropped its preferred opt-out model in March and "no longer has a preferred option". The Ministry of Justice's Legal Services AI Growth Lab, with the SRA, the LSB, the ICO and the CLC, closed to applications on 27 September. It is advisory only: it "cannot change legislation or create exemptions from the law" (GOV.UK). The Law Commission is consulting on bringing software and AI systems within product liability law, closing 14 January 2027 (Law Commission). And the UK Jurisdiction Taskforce's non-statutory legal statement of 7 July concluded that existing private law can deal with AI harms without a new liability regime (HSF Kramer).
How binding: all proposals, apart from the UKJT statement, which is persuasive guidance on the existing law.
Still open: the government's response to the Joint Committee on Human Rights, which called in September for a wide-ranging AI bill, was due by 14 November. None had been published by 17 November. The Growth Lab's first cohort of around a dozen participants has not been announced. The government says it will launch its strategy for AI in the justice system at the Global Business and Law Summit on 1 December.
Legal Services Board
A short one. The LSB's business plan for 2026/27 names AI as a risk focus, and a September speech noted that standards for consumer-facing legal AI tools "were almost entirely absent" (LSB). It has issued nothing that binds practitioners on AI. The bigger question it raises is about the chatbots people use instead of a lawyer, which no legal regulator covers.
Ad Break
Adverts help cover the running costs of this newsletter. This one is chosen and placed automatically by our newsletter platform, so it is not a product I have tried or endorse. If it catches your interest, clicking through costs you nothing and helps keep this newsletter going.
Can’t Buy the Next AI IPO Yet? Watch 3 Stocks.
Wall Street is preparing for what could become one of the biggest AI IPOs in history. But ordinary investors may have to wait until shares begin publicly trading to participate directly.
Meanwhile, three publicly traded companies could offer exposure to the AI infrastructure and investment boom surrounding Anthropic. Two are familiar names. The third is a lesser-known company that many investors haven't connected to the story.
A new, free briefing reveals all three tickers, potential entry prices, and a bonus fourth company holding thousands of patents tied to the AI race.
See the three stocks before the next wave of attention arrives. Free briefing, no credit card required.
What to watch in year two
3 December 2026: the SRA competence consultation closes. Final wording April 2027, in effect September 2027.
1 December 2026: the government's justice AI strategy, due at the Global Business and Law Summit.
Winter 2026: the ICO's final ADM guidance, then a draft of the statutory AI code (some commentators think 2027 is more realistic).
No date: the CPRC's response to the CJC, starting with the witness declaration and the new practice direction.
14 January 2027: the Law Commission product liability consultation closes.
No date: an SRA answer on AI and the conduct of litigation, and the first outcomes of its AI investigations.
No date: the joint procedure rule committees' work on disclosing AI use in criminal trial preparation.
What to do about it
The regulators have told you what they expect. The Golden Rules, 2026 edition turns it into ten things to do, updated last week with everything above. If you read one thing alongside this issue, make it that.
The whole year on one page
Every entry links to its source. The table is also available below as a free one-page PDF to keep or share.
Body | What | Force | Date |
|---|---|---|---|
SRA | Regulatory weight | 17 Aug 2026 | |
SRA | Guidance | 12 Jun 2026 | |
SRA | Must have regard (MLRs 18, 18A) | 4 Sep 2026 | |
SRA | Proposal (closes 3 Dec) | 8 Oct 2026 | |
Law Society | Guidance | 21 Oct 2026 | |
Bar Council | Guidance | Nov 2025 | |
BSB | Guidance | 18 May 2026 | |
CILEx Regulation | Guidance | 1 Oct 2026 | |
CLC | Guidance | 2026 | |
Judiciary | Guidance (for judges) | 31 Oct 2025 | |
Judiciary | Guidance (for judges) | 9 Oct 2026 | |
CJC | Proposal | 29 Oct 2026 | |
OPRC | Binding | 7 Sep 2026 | |
Court of Appeal | Binding | 31 Mar 2026 | |
Court of Appeal | Binding | 4 Jun 2026 | |
Upper Tribunal | Persuasive | 17 Nov 2025 | |
EAT | Persuasive | 17 Sep 2026 | |
Parliament | Binding | 5 Feb 2026 | |
Parliament | Binding (code not yet made) | 12 May 2026 | |
ICO | Proposal (closes 20 Nov) | 8 Oct 2026 | |
Government | Proposal (not introduced) | 13 May 2026 | |
MoJ | Advisory | 8 Jun 2026 | |
Law Commission | Proposal (closes 14 Jan 2027) | 8 Oct 2026 | |
UKJT | Guidance | 7 Jul 2026 |

How did we do?
Is there a body I have missed, or a status you think I have wrong? Hit reply. I will keep this page up to date through year two.
Thanks for reading,
Serhan, UK Legal AI Brief
Disclaimer
Guidance and news only. Not legal advice. A snapshot as at 17 November 2026. Always use AI tools safely and in line with your regulator's guidance and your firm's policy.

