Must lawyers verify AI output before relying on it?

Australia

Asia-Pacific · verified 2026-09-04

Universal. Every Australian instrument imposes a personal duty to verify citations and evidence references. NCAT PD 7 and Victoria's SC GEN 25 add that verification must not be carried out with another AI tool. Western Australia's guidelines require a human who takes legal responsibility for the document.

• Practice Note SC Gen 23: Use of Generative Artificial Intelligence
• Generative Artificial Intelligence Practice Note (GPN-AI)
• Practice Direction: Use of Artificial Intelligence (PD-AI)
• Practice Note SC GEN 25: The Use of Artificial Intelligence by Court Users
• Guidelines concerning the use of Generative Artificial Intelligence in litigation in South Australian Courts
• The Use of Generative Artificial Intelligence: Guidelines for Responsible Use by Non-Lawyers
• Practice Direction Number 5 of 2025: Accuracy of References in Submissions
• NCAT Procedural Direction 7: Use of Generative Artificial Intelligence (Gen AI)
• Guidelines for the use of generative AI
• NCAT Procedural Direction 7 (official PDF, cl 16-17)
• Supreme Court of WA — Guidelines for the use of generative AI (official PDF, cl 7-8)
• Supreme Court of Queensland — PD 5 of 2025 (official PDF, cl 9-13)

Canada

Americas · verified 2026-09-04

Verification is required by court rule or directive in most jurisdictions: Federal Court human-in-the-loop verification; Alberta's tri-court notice (6 Oct 2023) requires exclusive reliance on authoritative sources; Saskatchewan GA-PD #12 (1 Jan 2026) requires independent verification and meaningful human oversight; Ontario SCJ practice directions require verification against trusted sources.

• Notice to the Parties and the Profession: The Use of Artificial Intelligence in Court Proceedings
• Ontario Rules of Civil Procedure Rule 4.06.1(2.1) and (2.2): Certification of authenticity of authorities (O. Reg. 384/24)
• Alberta Courts — tri-court notice on Large Language Models (PDF)
• Saskatchewan King's Bench GA-PD #12 (PDF)

Ireland

Europe · verified 2026-09-04

PD HC 142: every AI-assisted citation or proposition must be independently verified by a human, and checking with another AI tool is expressly insufficient. Reinforced by the Judicial Council guidelines for judges and by the Law Society guidance v4.

• Practice Direction HC 142: Responsible Use of Generative Artificial Intelligence in Court Documents
• Guidelines for Judges: The Use of Generative Artificial Intelligence
• Use of Generative AI: Guidance for the Legal Profession

Philippines

Asia-Pacific · verified 2026-09-04

Canon VI, Section 5 of the 2025 Code of Judicial Conduct and Accountability permits a judge to use AI tools only for research of relevant legal authorities and requires the sources always to be verified before citation. Canon III, Section 6 of the court-personnel code mirrors this.

• A.M. No. 25-04-04-SC: The 2025 Code of Judicial Conduct and Accountability
• A.M. No. 25-06-11-SC: The 2025 Code of Conduct and Accountability for Court Officials and Personnel
• A.M. No. 25-11-28-SC: Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary
• A.M. No. 25-04-04-SC, 2025 Code of Judicial Conduct and Accountability
• A.M. No. 25-06-11-SC, 2025 Code of Conduct for Court Officials and Personnel

Poland

Europe · verified 2026-09-04

New § 23e of the Zbiór Zasad Etyki Adwokackiej (adopted 12–13 June 2026) requires an advocate to personally assess and verify results produced by technological tools, so as not to harm the client by failing to evaluate output independently. KIRP recommendations impose the same duty on radcowie prawni.

• Amendments to the Collection of Rules of Advocates' Ethics and Dignity of the Profession introducing § 23e on technology and AI
• AI w pracy radcy prawnego — Rekomendacje dotyczące korzystania przez radców prawnych z narzędzi opartych na sztucznej inteligencji
• Naczelna Rada Adwokacka — amendments to the ethics code on AI

Qatar

Middle East · verified 2026-09-04

The same practice direction requires comprehensive verification and certification of the accuracy of all sources before submission when AI tools are used; parties and representatives remain responsible for fabricated or inaccurate material.

• Practice Direction No. 1 of 2026 and Practice Guidance on the Use of Artificial Intelligence
• QICDRC — Practice Direction No. 1 of 2026 (official)

Singapore

Asia-Pacific · verified 2026-09-04

The court guide makes the user responsible for all filed content including verifying citations, and warns that personal costs orders may follow the citation of non-existent authorities, a warning the High Court has since made good repeatedly. The Law Society advisory requires verification against primary sources.

• Registrar's Circular No. 1 of 2024: Guide on the Use of Generative AI Tools by Court Users
• Advisory on the Use of Publicly Available AI Tools
• Singapore Courts — Guide on the use of generative AI tools by court users (PDF)
• Tan Hai Peng Micheal v Tan Cheong Joo [2026] SGHC 49

Slovakia

Europe · verified 2026-09-04

Resolution 12/4/2025 Art. 3(2): the lawyer retains full responsibility for AI outputs used. Guidance Art. 3(3): relying on AI output without human review is impermissible; the lawyer must actively verify facts, literature and case citations, legislation and legal conclusions to avoid 'hallucinations'.

• Rules on the use of AI tools in advocacy practice (SAK Presidium Resolution 12/4/2025, as amended by 21/14/2025)
• SAK — Uznesenie P SAK č. 12/4/2025 as amended (Vestník SAK čiastka 87, PDF)

Spain

Europe · verified 2026-09-04

CGAE Interpretive Circular 3/2026, approved by the Pleno on 10 April 2026 and amended on 13 April, provides that signing or filing AI-assisted documents without verification is a serious infraction under article 125.u of the profession's statute; the signer is solely responsible and a tool error is no excuse.

• CGAE Interpretive Circular 3/2026: use of generative AI systems and the duty of verification
• CGAE Circular 3/2026 (ICAIB mirror)
• CGAE news release

United States

Americas · verified 2026-09-04

Rule 11(b) certification plus ABA Formal Opinion 512 (29 Jul 2024) require verification of AI output before filing. Federal courts have held that submitting fabricated authorities violates Rule 11 even where the underlying proposition is correct, and that counsel must personally read every citation.

• ABA Formal Opinion 512: Generative Artificial Intelligence Tools
• Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the USPTO
• Interim guidance on the use of artificial intelligence in the federal judiciary
• Policy Memorandum 25-40: Use of Generative Artificial Intelligence in EOIR Proceedings
• ABA Formal Opinion 512 (official PDF)

Austria

Europe · verified 2026-09-04

ÖRAK Leitfaden: because AI systems can produce fabricated information, 'die anwaltliche Sorgfaltspflicht erfordert daher eine ausnahmslose Überprüfung und Verifizierung aller KI-generierten Ergebnisse'; final responsibility for every legal statement remains with the lawyer, and staff must be instructed never to adopt AI-generated text, legal reasoning or sources unchecked.

• Künstliche Intelligenz (KI) in Anwaltskanzleien — Leitfaden und Checkliste für KI-Dienstleister
• ÖRAK — Künstliche Intelligenz (KI) in Anwaltskanzleien, Stand September 2025 (official PDF)

Belgium

Europe · verified 2026-09-04

OVB/Avocats.be guidelines: 'De advocaat controleert het resultaat, met inbegrip van vermelde bronnen' — the lawyer must verify that the legislation, doctrine and case law cited in the output actually exist, and must remain critical of the reasoning even where each step cannot be traced to a source.

• Richtlijnen voor advocaten voor het gebruik van Artificiële Intelligentie (OVB/Avocats.be guidelines on AI for lawyers)
• OVB/OBFG — Richtlijnen voor advocaten Gebruik Artificiële Intelligentie (official PDF)

Brazil

Americas · verified 2026-09-04

OAB Recommendation 001/2024 of 11 November 2024 makes the lawyer responsible for verifying AI output, and the Sao Paulo Bar's ethics tribunal ruled on 16 April 2026 that all AI output must be fully reviewed before filing, citing article 77 of the civil procedure code. Courts enforce this through bad-faith fines.

• Recomendacao no 001/2024: Responsible Use of Generative AI in Legal Practice
• OAB/SP Ethics and Discipline Tribunal ruling on partners' duty to supervise AI use (proc. 25.0886.2025.014989-0)
• OAB Recomendacao 001/2024 (official PDF)
• OAB/SP ethics ruling (Migalhas)

Czechia

Europe · verified 2026-09-04

ČAK's 2023 Stanovisko treats AI output as unpredictable and potentially inaccurate, and states that using AI leaves the lawyer's statutory and professional responsibility wholly untouched. The Constitutional Court reinforced this in I. ÚS 3004/25, holding counsel fully answerable for hallucinated arguments.

• Stanovisko k užívání umělé inteligence (AI) při poskytování právních služeb
• Ústavní soud, usnesení sp. zn. I. ÚS 3004/25 — procedural fine for a constitutional complaint citing non-existent decisions
• ČAK — Stanovisko k užívání umělé inteligence (AI) při poskytování právních služeb (PDF)
• Ústavní soud — I. ÚS 3004/25 (NALUS)

Estonia

Europe · verified 2026-09-04

The Bar's AI guide requires advocates to evaluate AI output critically, verify its accuracy and correct errors, on the principle that 'the attorney always provides legal services, not artificial intelligence'. Advocates must keep their expertise and independent judgment rather than let AI displace critical thinking. Non-binding.

• AI guide for advocates [Tehisaru juhend advokaatidele]
• CCBE Guide on the use of generative AI by lawyers
• Eesti Advokatuur — AI guide for advocates (3 Dec 2024)

European Union

Europe · verified 2026-09-04

CCBE Guide on the use of generative AI by lawyers (2 Oct 2025) makes lawyers fully responsible for AI-assisted output and requires checking for invented citations. CEPEJ(2025)18Final requires judicial users to verify output before reliance. Neither is binding.

• CCBE Guide on the use of generative AI by lawyers
• Guidelines on the use of generative AI by courts and tribunals (CEPEJ(2025)18Final)
• Artificial Intelligence Ethics Charter
• CCBE guide (official PDF)

Finland

Europe · verified 2026-09-04

Data Security Manual B 05.2, Annex 5 point 2: whether advice rests wholly or partly on AI output, the advocate is personally liable; tools produce hard-to-spot errors and 'fabricated references' [sepitettyjä viittauksia], so the advocate must always verify correctness before use. Recommendatory in character.

• B 05.2 Data Security Manual [Tietoturvaopas], Annex 5 'AI guide for the advocate' [Tekoälyopas asianajajalle]
• CCBE Guide on the use of generative AI by lawyers
• Suomen Asianajajat — B 05.2 Tietoturvaopas (PDF, in force 1 Dec 2025)

France

Europe · verified 2026-09-04

Two official texts impose it. The Conseil d'État charter (11 Dec 2025) requires personnel to 'assurer systématiquement une vérification humaine' and never to assume AI output is true or complete. The CNB deontological guide (17 Mar 2026) makes the lawyer keep 'la maîtrise intellectuelle' of the work. Administrative courts have applied the duty (TA Orléans, 29 Dec 2025).

• Charte d'utilisation de l'intelligence artificielle au sein de la juridiction administrative
• Guide pratique — La déontologie et l'intelligence artificielle
• Guide pratique — Utilisation des systèmes d'intelligence artificielle générative
• Conseil d'État — Charte d'utilisation de l'IA (official PDF)
• CNB — Le CNB adopte un guide sur la déontologie et l'IA (official)

Germany

Europe · verified 2026-09-04

BRAK KI-Leitfaden §2.1 grounds the duty in § 43 sentence 1 BRAO (gewissenhafte Berufsausübung) and the principle of personal performance (§ 613 BGB): AI may only support, never replace, and 'eine eigenverantwortliche Überprüfung und Endkontrolle der KI-Ergebnisse' is required in every case. KG Berlin (20 Nov 2025, 17 WF 144/25) held counsel must check whether citations come from a 'fantasising' AI.

• Hinweise zum Einsatz von künstlicher Intelligenz (KI)
• BRAK — Hinweise zum Einsatz von künstlicher Intelligenz, Stand 12/2024 (official PDF)
• BRAK — KG ermahnt Anwältin wegen erfundener Urteile im Schriftsatz (official)

Hong Kong SAR

Asia-Pacific · verified 2026-09-04

Both instruments impose it as professional practice rather than rule: the Law Society's Circular 25-824 requires solicitors to verify AI-generated content, including citations, before use; the Judiciary's guidelines make judicial officers responsible for all output used and require verification of any permitted administrative use.

• Circular 25-824: Generative AI in Legal Practice: Risks and Tips
• Guidelines on the Use of Generative AI for Judges, Judicial Officers and Support Staff
• Law Society of Hong Kong — Circular 25-824 (official PDF)

Indonesia

Asia-Pacific · verified 2026-09-04

PERADI's guideline requires that every AI output pass through the advocate's own intellectual examination and filtration, with full professional accountability remaining with the advocate.

• Guideline on the Use of Artificial Intelligence for Indonesian Advocates [Pedoman Penggunaan Kecerdasan Buatan]
• PERADI — AI guideline announcement (6 Aug 2026)

International bodies

International bodies · verified 2026-09-04

A consistent soft-law duty to check output: CEPEJ(2025)18 requires outputs to be verified before reliance; the CCBE Guide of 2 October 2025 makes lawyers responsible for checking invented citations; SVAMC Guideline 4 and Ciarb 8.3 require parties and arbitrators to verify AI output; AAA-ICDR tells arbitrators to cross-reference primary sources.

• CCBE Guide on the use of generative AI by lawyers
• Guidelines on the use of generative AI by courts and tribunals (CEPEJ(2025)18Final)
• SVAMC Guidelines on the Use of Artificial Intelligence in Arbitration (First Edition)
• Ciarb Guideline on the Use of AI in Arbitration (2025)
• AAA-ICDR Guidance on Arbitrators' Use of AI Tools
• CCBE Guide (PDF)
• CEPEJ(2025)18Final

Israel

Middle East · verified 2026-09-04

At/60/24 requires lawyers to check AI output individually and corroborate it against external authority. Procedure 01-26 binds judges and staff to verify every factual finding, quotation, statutory and case-law reference against authoritative sources. Supreme Court decisions since Feb 2025 have made the duty enforceable through costs.

• Preliminary Opinion on the Use of Artificial Intelligence (AI) in Lawyers' Work, decision At/60/24
• Ethical Code for the Use of Artificial Intelligence Tools in the Judicial Authority (Courts Administration Procedure No. 01-26)
• Israel Bar opinion At/60/24 (official page)
• CTech, courts AI ethics code and verification requirement (20 Apr 2026)

Japan

Asia-Pacific · verified 2026-09-04

The JFBA AI Strategy Working Group notice of September 2025, Points of Attention on the Use of Generative AI in Attorney Practice, states that outputs may cite fictitious case law and must be checked against primary sources, and that the attorney bears final responsibility. The text is members-only.

• Points of Attention on the Use of Generative AI in Attorney Practice: Five Points for Appropriate Use
• Seminar page quoting the JFBA notice
• Practitioner commentary on the JFBA notice

Luxembourg

Europe · verified 2026-09-04

No Luxembourg instrument. The CCBE guide on the use of generative AI by lawyers (2 Oct 2025), which the Luxembourg Bar subscribes to as a CCBE member, requires output — including invented citations — to be checked.

• CCBE Guide on the use of generative AI by lawyers

Malaysia

Asia-Pacific · verified 2026-09-04

Bar Council Circular 242/2025 requires all generative-AI output to be independently verified against traditional legal databases, and says that where verification is not possible the content must not be relied on. It is professional guidance, not a binding rule.

• Circular No. 242/2025: Updates on the Use of Generative Artificial Intelligence in Legal Practice
• Circular No. 342/2023: Risks and Precautions in Using Generative Artificial Intelligence in the Legal Profession
• Malaysian Bar — Circular No. 242/2025 with annexed advisory (official PDF)
• Bar Council — Circular No. 342/2023 (official PDF)

Netherlands

Europe · verified 2026-09-04

NOvA recommendation Deskundigheid 3: 'Verifieer altijd de output — controleer citaten, jurisprudentie en feiten altijd handmatig vóór gebruik' and use only tools that cite sources. Enforced in Raad van Discipline 's-Hertogenbosch, 27 Jul 2026 (ECLI:NL:TADRSHE:2026:93) as an aspect of deskundigheid and integriteit under art. 10a Advocatenwet.

• Aanbevelingen AI in de advocatuur
• NOvA — Aanbevelingen AI in de advocatuur (official)
• ECLI:NL:TADRSHE:2026:93 (tuchtrecht.overheid.nl, official)

New Zealand

Asia-Pacific · verified 2026-09-04

Central to both instruments. The court guidelines require accuracy checking before reliance, warning that chatbots invent cases, citations and quotes and will confirm false information as accurate. The Law Society guidance requires all output to be verified before reliance.

• Guidelines for use of generative artificial intelligence in Courts and Tribunals
• Generative AI guidance for lawyers
• Courts of New Zealand — GenAI guidelines, ensure accountability and accuracy (PDF)

Nigeria

Africa · verified 2026-09-04

NBA-SLP AI Guidelines s.7.1 (28 Aug 2024) require outputs to be thoroughly reviewed before reliance in client work, verifying factual accuracy, legal soundness and bias. Non-binding; no court rule.

• Guidelines for the Use of Artificial Intelligence in the Legal Profession in Nigeria
• NBA-SLP AI Guidelines (official text, archived)
• NBA launch announcement (official)

Norway

Europe · verified 2026-09-04

Supreme Court guidance section 19: 'Det er kjent at KI kan hallusinere og finne opp blant annet rettskilder. Kilder og andre opplysninger KI oppgir, må kvalitetssikres.' The lawyer is personally responsible for all information used, including legal material, and must ensure central sources are not overlooked.

• Veiledning for advokater, section 19 'Bruk av KI-verktøy'
• Veiledning for bruk av kunstig intelligens i advokatvirksomhet
• Norges Høyesterett — Veiledning for advokater, s. 54 (PDF, July 2026)

South Korea

Asia-Pacific · verified 2026-09-04

The AI Guidebook for Judges, announced by the National Court Administration on 24 February 2026 and distributed in March 2026, provides checklists addressing hallucination, bias and privacy when judges use commercial AI. There is no lawyer-facing rule.

• AI Guidebook for Judges
• Dailian on the AI Guidebook for Judges

Sweden

Europe · verified 2026-09-04

Bar guidance (12 June 2025) s.3: GenAI can cite cases or statutes that do not exist, so a lawyer must critically review output before it enters advice — verifying sources, checking statistics and confirming the applicability of cited law. Advisory liability applies in full regardless of GenAI use (VRGA 5.3, 6.2).

• Vägledning om användning av generativ AI i advokatverksamhet
• Advokatsamfundet — Vägledning om användning av generativ AI i advokatverksamhet (PDF)

Switzerland

Europe · verified 2026-09-04

SAV Wegleitung §2.2: it is 'von grösster Wichtigkeit' to check the output independently and critically. It warns that an AI cannot verify its own output, and names hallucinations, missing/outdated training data and sycophancy — an AI adapting its answer to the user's view — as the error sources. §2.3: a lawyer cannot plead that 'the AI made a mistake'.

• SAV-Wegleitung für den Umgang mit künstlicher Intelligenz
• SAV — Wegleitung für den Umgang mit künstlicher Intelligenz (official PDF)

Taiwan

Asia-Pacific · verified 2026-09-04

Judicial Yuan Reference Guidelines for Use of AI Systems, issued by order of 22 April 2026, article 8: users must recognise hallucination risk and cross-check AI output against official documents, case files and legal databases before use. The guidelines apply to court personnel, not to lawyers.

• Reference Guidelines for the Use of Artificial Intelligence Systems by the Judicial Yuan and Subordinate Agencies
• Guideline text (LawPlayer mirror)

Türkiye

Middle East · verified 2026-09-04

TBB Guide: a lawyer may not submit any legislative provision, precedent, citation or fact obtained from an AI tool to a court, arbitrator, mediator or administrative body without confirming it against official and reliable sources. Submitting hallucinated citations may be a grave breach of the duty of professional care.

• Recommendation Guide on the Use of Artificial Intelligence for Lawyers [Avukatlar İçin Yapay Zekâ Kullanımı Tavsiye Rehberi]
• TBB Recommendation Guide on AI Use for Lawyers (official PDF)
• TBB announcement of the Guide (28 Aug 2026)

United Arab Emirates

Middle East · verified 2026-09-04

DIFC PGN 2/2023 requires AI-generated material to be verified for accuracy and reliability against independent sources before submission. In ADGM, Arabyads Holding v Alam (18 Dec 2025) held that verifying AI research, including that cited authorities exist, is a practitioner obligation.

• DIFC Courts PGN No. 2 of 2023 (official)
• Arabyads Holding Ltd v Gulrez Alam, costs judgment (ADGM, official PDF)

United Kingdom

Europe · verified 2026-09-04

SRA warning notice on misuse of AI (17 Aug 2026) and judiciary guidance require verification of AI output; BSB guidance (18 May 2026) requires barristers to verify output before relying on it or presenting it to a court; Ayinde v Haringey (2025) states the professional duty.

• Misuse of AI: Warning notice
• Artificial Intelligence (AI): Judicial Guidance
• Guidance on the use of Artificial Intelligence and Other Technologies

Vietnam

Asia-Pacific · verified 2026-09-04

Non-binding and journal-level. A Supreme People's Court journal article warns that ChatGPT does not access official case databases directly and that outputs must be independently verified. It is authored commentary, not an instrument.

• Tap chi Toa an nhan dan — using ChatGPT and the court virtual assistant to research judgments (9 Oct 2025)

Ecuador

Americas · verified 2026-09-04

A bill to reform the Organic Code of the Judicial Function, presented on 13 July 2026, would bar judges from using AI information without verifying accuracy, relevance and currency, and bar lawyers from filing AI-generated content without verifying it where it could mislead. It is at an initial stage.

• Bill reforming the Organic Code of the Judicial Function to regulate AI use in judicial proceedings
• Infobae on the reform bill

Argentina

Americas · verified 2026-09-04

The Rosario Civil and Commercial Court of Appeals held in August 2025 that delegating research to AI and filing unverified citations is extremely risky and even reckless; a General Roca appeals court found 26 non-existent citations. The Rio Negro acordada of 1 October 2025 makes verification mandatory for lawyers provincially.

• Camara de Apelacion en lo Civil y Comercial de Rosario, Sala II, Giacomino c/ Monserrat: AI-fabricated citations
• Acordada 15/2024 and October 2025 extension: Protocol of Good Practice for Generative AI in the Judiciary of Rio Negro
• Guide for the use of Artificial Intelligence for Lawyers (Guia para el uso de Inteligencia Artificial para Abogados)
• Rosario ruling (Abogados.com.ar)
• Rio Negro acordada extends the protocol to lawyers

Chile

Americas · verified 2026-09-04

Supreme Court Rol 23.322-2025 of 22 April 2026 imposed a one-month suspension and a 5 UTM fine for AI-fabricated doctrinal citations under articles 531 and 542 of the Organic Code of Courts; the second civil court of Concepcion fined a lawyer 1 UTM in February 2026. For judges, article 17 of the ethics code requires independent verification.

• Supreme Court Rol 23.322-2025: suspension and fine for AI-fabricated doctrinal citations
• Acta 244-2025: Judicial Ethics Code (article 17 on responsible use of technology and AI)
• Supreme Court Rol 23.322-2025 (Poder Judicial news)
• Concepcion fine (CNN Chile)

China

Asia-Pacific · verified 2026-09-04

Beijing Tongzhou District People's Court, selected for the People's Court Case Database in December 2025: a lawyer filed two AI-fabricated precedents; the court wrote criticism into the judgment and stated that litigants submitting unverified AI-generated cases bear consequences, from criticism to fines under Civil Procedure Law article 114 in serious cases.

• Beijing Tongzhou District People's Court: counsel criticised for filing AI-fabricated precedents
• China News Service report
• Ministry of Justice legal-publicity repost

Colombia

Americas · verified 2026-09-04

The Supreme Court's civil cassation chamber fined a lawyer 15 minimum monthly wages for a review petition citing ten non-existent rulings generated by AI, and the Council of State held in June 2026 that all AI output must be humanly verified and lawyers remain fully responsible. For judges, Acuerdo PCSJA24-12243 article 9 is binding.

• STC17832-2025: annulment of a judgment based on non-existent jurisprudential citations
• Council of State, Third Section: parameters for AI use in proceedings and disciplinary referral
• Acuerdo PCSJA24-12243: Guidelines for the respectful, responsible, safe and ethical use of AI in the Judicial Branch
• Supreme Court fine (El Tiempo)
• Council of State rules (Infobae)

Denmark

Europe · verified 2026-09-04

No express rule, but Advokatnævnet held in December 2025 that citing two non-existent judgments (U.2006.2042V and U.2008.1166H) showed a lack of due care and breached good lawyer conduct [god advokatskik], rejecting the 'typing error' explanation and imposing a DKK 10,000 fine.

• Djøfbladet — Danish lawyer fined for invented judgments (May 2026)

India

Asia-Pacific · verified 2026-09-04

Pooja Ramesh Singh v J&K Bank (2 July 2026) holds that Bar and Bench owe a zero-tolerance duty to verify AI-generated authority and that citing unverified AI judgments is advocate misconduct; Gadiya (2 September 2026) extends this to quasi-judicial officers. Kerala and Gujarat High Court policies impose verification on judicial officers.

• Pooja Ramesh Singh v Jammu and Kashmir Bank Ltd, 2026 INSC 668
• Vijay Ghanshyam Gadiya v Union of India, 2026 INSC 947
• Policy Regarding Use of Artificial Intelligence Tools in District Judiciary
• Policy on the Use of Artificial Intelligence in Judicial and Court Administration
• Draft Regulations for Use of Artificial Intelligence in Courts, 2026
• Pooja Ramesh Singh v J&K Bank (Indian Kanoon)
• Gujarat High Court AI policy (official PDF)

Italy

Europe · verified 2026-09-04

No statutory or bar rule names a verification duty; the courts have set it. Tribunale di Firenze (14 March 2025) censured the failure to verify ChatGPT citations; Tribunale di Siracusa (20 February 2026) treated unverified generative-AI output as gross negligence; the Cassazione declared a petition inadmissible for probable hallucination.

• Tribunale di Firenze, Sezione Imprese, order of 14 March 2025 (ChatGPT-invented Cassazione citations)
• Tribunale di Siracusa, Sez. II civ., judgment no. 338/2026 (AI-hallucinated citations, Art. 96 c.p.c.)
• Il Sole 24 Ore NT+ Diritto on Tribunale di Siracusa
• Litis.it on Tribunale di Firenze

Kenya

Africa · verified 2026-09-04

No rule, but the courts have set the position. Mwamuye J (Milimani High Court, Mar 2026) held computer-generated documents or outputs of artificial intelligence cannot substitute for human-drawn documents; Chigiti J (16 Apr 2026) set aside a judgment founded on AI-generated pleadings. Supreme Court Justice Lenaola warned in May 2025 that every authority in one submission was fictitious.

• Draft Judiciary Artificial Intelligence Policy and draft Practice Directions on AI use in courts
• High Court of Kenya rulings on AI-generated pleadings: Savla v Commission on Administrative Justice and the April 2026 set-aside
• Milimani court strikes out AI-generated application (Capital FM, 11 Mar 2026)
• High Court declares AI-drafted pleadings illegal (HapaKenya, 19 Apr 2026)
• Justice Lenaola warns lawyers over fake AI references (Capital News, May 2025)

Pakistan

Asia-Pacific · verified 2026-09-04

Ishfaq Ahmed v Mushtaq Ahmed (Supreme Court, April 2025): AI output including large-language-model legal references may be fabricated, and human oversight is essential to verify AI-generated content before reliance. The holding applies to judges; there is no lawyer-facing rule.

• Ishfaq Ahmed v Mushtaq Ahmed, 2025 SCP 112 (PLD 2025 SC 582)
• Sindh High Court Legal Research Cell digest, 1-15 April 2025

Portugal

Europe · verified 2026-09-04

The Guimaraes Court of Appeal on 10 February 2026 called citing six non-existent appellate and Supreme Court rulings absolutely inadmissible, stressed higher standards for legal professionals and suggested referral to the bar. For judges, CSM Recommendation 6 requires complete, critical personal review of any AI draft.

• Tribunal da Relacao de Guimaraes, 10 Feb 2026: six non-existent appellate rulings generated by AI
• CSM Recommendations on the use of artificial intelligence in judicial activity (Recomendacoes no uso de IA)
• CNN Portugal on the Guimaraes decision
• CSM Recommendations (official PDF)

South Africa

Africa · verified 2026-09-04

The duty is judge-made. Mavundla v MEC (KZN), Northbound Processing and Parker v Forsyth establish that practitioners are personally responsible for checking every authority. The LSSA AI Committee's revised draft ethics guidelines (1 July 2026) state practitioners are fully responsible for AI-assisted work but are not adopted.

• Revised Ethics Guidelines for Legal Practitioners in South Africa on the Use of Generative AI

Afghanistan

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Albania

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Algeria

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Armenia

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Azerbaijan

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Bahrain

Middle East · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Bangladesh

Asia-Pacific · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Belarus

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Bolivia

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Bosnia and Herzegovina

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Botswana

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Brunei

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Bulgaria

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Cambodia

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Cameroon

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Costa Rica

Americas · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Côte d'Ivoire

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Croatia

Europe · verified 2026-09-04

No Croatian rule or reported case; the Croatian Bar republished the CCBE guide, which sets a verification expectation.

• CCBE Guide on the use of generative AI by lawyers

Cuba

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Cyprus

Europe · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Dominican Republic

Americas · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Egypt

Africa · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026. No duty to verify AI-generated citations is imposed by any Egyptian court, the Ministry of Justice or the Egyptian Bar Association.

El Salvador

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Ethiopia

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Georgia

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Ghana

Africa · verified 2026-09-04

no rule, guidance or reported case found as of 4 September 2026

Greece

Europe · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026; Greek legal commentary discusses only the United States Mata v Avianca case.

Guatemala

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Honduras

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Iceland

Europe · verified 2026-09-04

No rule, bar guidance or reported case imposing an explicit duty to verify AI output or citations was found as of 4 September 2026. The Bar Association's site carries no AI material.

• Lögmannafélag Íslands

Iran

Middle East · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Iraq

Middle East · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Jamaica

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Jordan

Middle East · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Kazakhstan

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Kuwait

Middle East · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026. Neither the Court of Cassation nor the Ministry of Justice has imposed a duty to verify AI-generated citations.

Kyrgyzstan

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Laos

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Latvia

Europe · verified 2026-09-04

No rule or guidance imposing an explicit duty to verify AI output or citations found. The Collegium's Ethics Code (2019) predates generative AI and does not mention it.

• LZAK — internal normative acts index

Lebanon

Middle East · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Lithuania

Europe · verified 2026-09-04

No rule or bar guidance imposing an explicit duty to verify AI output or citations was published as of 4 September 2026. The Bar's September 2026 training covers 'safe, responsible and professional' use but is not a normative document.

• Lietuvos advokatūra — AI training announcement (1 Sept 2026)

Macao

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Malta

Europe · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026; practitioners reportedly look to United Kingdom Bar Council guidance.

• Legal 500 Malta AI chapter (secondary)

Mauritius

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Mexico

Americas · verified 2026-09-04

No express duty to verify AI output before filing has been imposed on parties or counsel. The 2025 collegiate-court criteria impose human supervision on judges, not a filing-verification duty on litigants.

Moldova

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Mongolia

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Montenegro

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Morocco

Africa · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026. Neither the Cour de cassation nor the Ministry of Justice has imposed a duty to verify AI-generated citations.

Mozambique

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Myanmar

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Namibia

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Nepal

Asia-Pacific · verified 2026-09-04

No rule for lawyers; the Judiciary ICT Policy addresses only the reliability of court-used systems.

• Judiciary Information and Communication Technology Policy, 2083 (2026)

Nicaragua

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

North Macedonia

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Oman

Middle East · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Panama

Americas · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Paraguay

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Peru

Americas · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Puerto Rico

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Russia

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Rwanda

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Saudi Arabia

Middle East · verified 2026-09-04

No legal-practice-specific verification duty. SDAIA's Generative AI Guidelines recommend human-in-the-loop review and fact-checking of generative output for all users, but are non-binding and not addressed to litigation.

• SDAIA — Generative AI Guidelines for the Public (official PDF)

Senegal

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Serbia

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Slovenia

Europe · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026; the CCBE guide is the available reference.

• CCBE Guide on the use of generative AI by lawyers

Sri Lanka

Asia-Pacific · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Tanzania

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Trinidad and Tobago

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Tunisia

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Uganda

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Ukraine

Europe · researched 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Uruguay

Americas · verified 2026-09-04

No rule, guidance or reported case found as of 4 September 2026.

Uzbekistan

Asia-Pacific · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Venezuela

Americas · researched 2026-09-04

No rule, guidance or reported case imposing an AI-specific verification duty found as of 4 September 2026.

Zambia

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Zimbabwe

Africa · researched 2026-09-04

No rule, guidance or reported case imposing a duty to verify AI-generated citations before filing or reliance was found as of 4 September 2026.

Hungary

Europe · verified 2026-09-04

Not researched — no MÜK or court statement on a duty to verify AI output could be confirmed. The MÜK homepage was checked on 4 September 2026 and surfaced no AI item, but the Szakmai álláspontok and Szabályzattár sections could not be searched.

• Magyar Ügyvédi Kamara

Romania

Europe · verified 2026-09-04

Not researched — no UNBR or court statement on a duty to verify AI output could be confirmed; unbr.ro returned HTTP 403 on 4 September 2026.

Thailand

Asia-Pacific · verified 2026-09-04

Not substantiated. No Courts of Justice or Lawyers Council instrument imposing a verification duty was located.