MinLaw vs SRA vs Indian Courts: How AI Rules for Lawyers Compare Across Jurisdictions
What each regime actually says, what is settled law versus still in draft, and what a lawyer working across Singapore, England and Wales, and India needs to know.
No two of these jurisdictions regulate AI in legal practice the same way. Singapore has a coordinated pair of documents: a sector-wide guide from the Ministry of Law and a formal court circular for filings. England and Wales relies on several separate bodies — the Solicitors Regulation Authority, the representative Bar Council, and the judiciary — each restating existing professional duties rather than writing new rules. India has no finalised rule for lawyers at all yet: the bar regulator has been directed by the Supreme Court to convene an expert committee and prescribe disciplinary norms but has not yet done so, while the Supreme Court’s own draft regulations for AI in courts, still out for comment, would be the first jurisdiction of the three to make AI disclosure mandatory.
Here is what each regime actually says, what is settled law versus still in draft, and what a lawyer working across any of these jurisdictions needs to know.
England and Wales: Guidance Layered on Existing Duties
There is no single AI rule for solicitors or barristers in England and Wales. Instead, several bodies have each issued their own non-binding guidance, all pointing back to the same underlying professional conduct rules.
The Solicitors Regulation Authority’s most recent statement is its compliance tips for solicitors regarding the use of AI and technology, last updated 9 February 2026. It introduces no new rules. It applies the existing SRA Standards and Regulations — principally the competence obligations at paragraphs 3.2 and 3.3, the confidentiality obligation at paragraph 6.3, and the duty not to mislead at paragraph 1.4 of the Code of Conduct for Solicitors — to AI specifically. The guidance expects firms to have leadership and oversight, risk assessments, documented policies, training, and ongoing monitoring in place, with the Compliance Officer for Legal Practice responsible for regulatory compliance when new technology is introduced.
The Bar Council, which represents barristers rather than regulating them, issued updated guidance on generative AI on 25 November 2025. It is expressly not “guidance” for the purposes of the Bar Standards Board Handbook, but states that its principles reflect current professional expectations in light of recent High Court judgments. The judiciary’s own Artificial Intelligence Judicial Guidance, first issued in December 2023 and refreshed since, was most recently updated in October 2025; it is aimed at judges, tribunal members, and court staff rather than practitioners.
The Civil Justice Council ran a consultation titled Use of AI for Preparing Court Documents, opened in February 2026 and closed on 14 April 2026, asking whether a formal disclosure requirement should be introduced for AI-assisted filings. Its final report is expected later in 2026, and its stated direction of travel is to maintain the current position for AI use in professional legal drafting, introduce proportionate transparency for expert evidence, and treat litigants in person as a distinct problem. As of this writing, England and Wales has no mandatory disclosure rule, and the CJC is not presently minded to propose one for pleadings and skeleton arguments.
None of this guidance requires a solicitor to tell a client — still less the court — that AI was used, unless the firm’s own policy or the client relationship calls for it. The SRA’s position is that it should always be clear to clients where they are interfacing with AI, but this is expressed as an expectation tied to the existing duty not to mislead, not a standalone disclosure rule with a prescribed mechanism.
What is not in doubt is where responsibility sits when AI goes wrong. In Ayinde v Haringey LBC and Al-Haroun v Qatar National Bank, decided together by the Divisional Court on 6 June 2025, Dame Victoria Sharp P held that a language model such as ChatGPT is not capable of conducting reliable legal research, warned that unverified reliance on AI output could amount to contempt of court, and referred the lawyers involved to their regulators while declining to bring contempt proceedings. Since then, a wasted costs order followed fabricated citations in Ndaryiyumvire v Birmingham City University, and in March 2026 Bournemouth Family Court publicly named an unregistered barrister after all the advocates agreed that four of the citations or propositions in her skeleton argument were erroneous.
Singapore: A Coordinated Two-Part Framework
Singapore is the most procedurally developed of the three. It operates on two documents that work together rather than in tension.
The Ministry of Law’s Guide for Using Generative AI in the Legal Sector was published on 6 March 2026, following a public consultation held in September 2025. It is non-binding but detailed, and it applies broadly — to lawyers in private practice, in-house counsel, paralegals, legal technologists, alternative legal service providers, law students, and anyone providing GenAI tools into the legal sector. Its central principle is that legal professionals remain ultimately accountable for all work product regardless of how it was produced, and it sets out expectations on professional ethics, confidentiality, and transparency across common use cases including legal research, contract review, and drafting.
Separately, the Singapore Courts issued a Guide on the Use of Generative AI Tools by Court Users, which took effect on 1 October 2024 and applies across the Supreme Court, the State Courts (including the Small Claims Tribunals, the Employment Claims Tribunals, and the Community Disputes Resolution Tribunals), and the Family Justice Courts. This guide takes a neutral, permissive stance: it does not prohibit using generative AI to prepare court documents, but it places full responsibility on the court user — whether a lawyer, a litigant in person, or a witness — to ensure that anything submitted is independently verified, accurate, true, and appropriate. It does not impose a rule of pre-emptive disclosure. Unless the court specifically asks, a party does not have to declare that AI was used, because responsibility for the content rests with the user regardless. It does, however, require that any AI-provided legal citation be verified against authoritative sources, specifically naming the Singapore Courts judgments pages and Singapore Statutes Online, and it prohibits using generative AI to create or tamper with evidence.
The consequences of skipping verification have already played out. In Tan Hai Peng Micheal v Tan Cheong Joo, decided on 6 March 2026, Justice S. Mohan of the Singapore High Court ordered two lawyers to pay S$5,000 each in personal costs after written submissions cited two authorities that did not exist — one combining a real case name with a citation belonging to a different case, the other fabricated outright.
India: A Regulatory Gap for Lawyers, and a Fast-Moving Draft for Courts
India currently has the least settled framework of the three, and also the most ambitious document in draft form.
For lawyers specifically, there is no formal Bar Council of India guidance on AI use, whether on disclosure, verification duties, or professional discipline for filing hallucinated material. That gap became a live issue in July 2026, when the Supreme Court set aside NCLT and NCLAT orders in the Essel Infraprojects insolvency matter that had relied on six defective precedents — three of which did not exist and three of which were real citations carrying invented passages or a wrong case title. The Court held that a decision materially founded on non-existent authority cannot be sustained, and directed the Bar Council of India to constitute a committee of experts and prescribe guiding norms, with disciplinary consequences for breach. As of this writing, the Bar Council has not yet issued those rules.
Ahead of the bar regulator, the judiciary itself has moved first. On 3 June 2026, the Supreme Court’s AI Committee released draft Regulations for Use of Artificial Intelligence in Courts, 2026, open for public comment until 20 June 2026. This is a draft, not yet in force. If adopted in its current form, it would be the most prescriptive of the three regimes covered here. Draft Regulation 43(3) would require any party or legal representative using AI in the preparation or submission of pleadings, documents, or evidence to disclose that fact at the time of filing, and would empower courts to demand details of the AI system used, the extent of its involvement, and the verification steps taken. The draft is explicit that AI hallucinations are treated as advisory output requiring human verification, and that a party cannot point to AI as a defence if a filing turns out to be false, fabricated, or misleading. The draft regulations are built around five stated principles: human primacy, transparency, accountability, data protection, and judicial independence, and would apply across the Supreme Court, all High Courts, and every subordinate court, tribunal, and statutory adjudicatory body once brought into force.
Side by Side
| England and Wales | Singapore | India | |
|---|---|---|---|
| Lead instrument for lawyers | SRA compliance tips (solicitors), Bar Council guidance (barristers) | MinLaw Guide for Using Generative AI in the Legal Sector | No finalised Bar Council of India guidance; Supreme Court has directed one be written |
| Status | Non-binding, restates existing Code of Conduct duties | Non-binding sector guide | Not yet issued |
| Court-specific instrument | None yet; Civil Justice Council consultation closed April 2026, final report due later in 2026 | Guide on the Use of Generative AI Tools by Court Users, in force since 1 October 2024 | Draft Regulations for Use of AI in Courts, 2026, not yet in force |
| Mandatory disclosure to court | No | No, unless the court specifically asks | Proposed under draft Regulation 43(3), not yet in force |
| Verification duty | Yes, tied to existing competence duty | Yes, express in the court guide | Yes, express in the draft regulations |
| ”AI did it” as a defence | No | No | No, expressly excluded in the draft |
| Reported sanctions case | Ayinde v Haringey LBC; Al-Haroun v QNB (June 2025) | High Court personal costs order, S$5,000 each, two lawyers (2026) | Supreme Court set aside tribunal orders over six defective precedents, three non-existent (July 2026) |
What Is the Same Everywhere
Strip away the institutional differences and three principles hold across all three jurisdictions. First, using AI does not transfer or reduce professional responsibility; the lawyer who files the work answers for it, not the tool or its provider. Second, verification against a primary source — a real case report, a real statute, a real judgment — is treated as non-negotiable, whether that duty is expressed through an existing competence obligation, as in England and Wales, or spelled out explicitly, as in Singapore’s court guide and India’s draft regulations. Third, none of the three regimes accepts “the AI got it wrong” as an excuse once something fabricated reaches a court.
What Is Genuinely Different
The real divergence is architectural, not principled. England and Wales relies on multiple existing regulators each restating old rules for a new technology, with no unified instrument and no mandatory disclosure rule yet in force. Singapore built two coordinated instruments from the outset — one for the profession and one specifically for court filings — and deliberately chose not to require pre-emptive disclosure, on the reasoning that responsibility already sits with the filer regardless. India is the outlier on both ends: it has the biggest regulatory gap for lawyers’ day-to-day professional conduct, and simultaneously the most prescriptive draft rule for court filings — a mandatory disclosure requirement that neither England and Wales nor Singapore currently imposes.
Practical Takeaway
For a lawyer working across any of these jurisdictions, the safest working assumption is the strictest one available today: verify every AI-generated citation against a primary source before filing, keep a record of what was checked, and be prepared to disclose AI use even where it is not yet formally required — since the regulatory direction of travel in all three jurisdictions points toward more disclosure and more documented verification, not less.
Frequently Asked Questions
Does Singapore require lawyers to disclose AI use in court filings? No, not pre-emptively. The Singapore Courts’ Guide on the Use of Generative AI Tools by Court Users, in force since 1 October 2024, does not require disclosure unless the court specifically asks, because responsibility for the content rests with the filer regardless of disclosure.
Does the SRA require solicitors to tell clients they used AI? The SRA expects it to always be clear to clients where they are interfacing with AI, but this is an expectation tied to the existing duty not to mislead clients, not a standalone rule with a specific mandated disclosure mechanism.
Will India require lawyers to disclose AI use in court filings? It is proposed, not yet law. Draft Regulation 43(3) of the Supreme Court’s Draft Regulations for Use of Artificial Intelligence in Courts, 2026, would require disclosure at the time of filing. The draft was open for public comment as of mid-2026 and had not yet been finalised or brought into force.
Which jurisdiction has the strictest AI rules for lawyers? On paper, India’s draft court regulations would be the strictest if adopted, given the proposed mandatory disclosure requirement. In practice, all three jurisdictions already treat unverified AI-generated citations as a serious professional failure, regardless of whether a formal disclosure rule exists.
Has the Bar Council of India issued AI rules for lawyers? Not as of mid-2026. The Supreme Court directed the Bar Council of India to constitute an expert committee and prescribe norms, with disciplinary consequences for breach, after a July 2026 ruling that set aside tribunal orders resting on non-existent precedents, but the Bar Council had not yet issued those norms.
For a deeper look at the SRA’s position specifically, see SRA AI Compliance Tips, Explained. For the mechanics of hallucination and the case law behind it across all three jurisdictions, see What Is AI Hallucination in Law? The AI Bar’s Legal Research with AI module covers jurisdiction-aware verification workflows built for practitioners in the UK, India, Singapore, and common law jurisdictions worldwide.