As of 2026-10-07, 114 official documents govern AI use by lawyers, judges or litigants in other jurisdictions. 22 require disclosure or certification of AI use, 57 impose an explicit verification duty, and 25 prohibit some use outright. The incident tracker records 2 incidents in this jurisdiction.
Afghanistan category map →Albania category map →Armenia category map →Argentina category map →Austria category map →Azerbaijan category map →Bosnia and Herzegovina category map →Bangladesh category map →Belgium category map →Bulgaria category map →Brunei category map →Bolivia category map →Botswana category map →Belarus category map →Switzerland category map →Côte d'Ivoire category map →Chile category map →Cameroon category map →China category map →Colombia category map →Costa Rica category map →Cuba category map →Cyprus category map →Czechia category map →Germany category map →Denmark category map →Dominican Republic category map →Algeria category map →Ecuador category map →Estonia category map →Egypt category map →Spain category map →Ethiopia category map →Finland category map →France category map →Georgia category map →Ghana category map →Greece category map →Guatemala category map →Honduras category map →Croatia category map →Hungary category map →Indonesia category map →Iraq category map →Iran category map →Iceland category map →Italy category map →Jamaica category map →Jordan category map →Japan category map →Kenya category map →Kyrgyzstan category map →Cambodia category map →South Korea category map →Kazakhstan category map →Laos category map →Lebanon category map →Sri Lanka category map →Lithuania category map →Luxembourg category map →Latvia category map →Morocco category map →Moldova category map →Montenegro category map →North Macedonia category map →Myanmar category map →Mongolia category map →Macao category map →Malta category map →Mauritius category map →Mexico category map →Malaysia category map →Mozambique category map →Namibia category map →Nigeria category map →Nicaragua category map →Netherlands category map →Norway category map →Nepal category map →Oman category map →Panama category map →Peru category map →Philippines category map →Pakistan category map →Poland category map →Puerto Rico category map →Portugal category map →Paraguay category map →Romania category map →Serbia category map →Russia category map →Rwanda category map →Sweden category map →Slovenia category map →Slovakia category map →Senegal category map →El Salvador category map →Thailand category map →Tunisia category map →Türkiye category map →Trinidad and Tobago category map →Taiwan category map →Tanzania category map →Ukraine category map →Uganda category map →Uruguay category map →Uzbekistan category map →Venezuela category map →Vietnam category map →Zambia category map →Zimbabwe category map →
Mahkamah Agung
A Supreme Court of Indonesia letter of 2 September 2026 circulating draft judgment-writing guidance and templates for consultation. Its annex contains what appears to be the Indonesian judiciary's first written restriction on AI: draft templates must not be uploaded to AI document-processing services.
Policy · Prohibition · Confidentiality
2026-09-02In force
Lietuvos advokatūra
A voluntary full-day practical course run by the Lithuanian Bar's Information Technology and Artificial Intelligence Committee on 9 September 2026. It is the Bar's first substantive public AI initiative for advocates and stands in place of any published guidance document, which Lithuania still lacks.
Policy · Competence · Permissive
2026-09-01In force
Supreme Court of Japan
The Supreme Court of Japan announced on 28 August 2026 a budget request for the 2027 financial year to build and test a generative-AI system on records of settled civil cases. AI would draft record summaries and issue lists to reduce judges' workload, with leak prevention for personal data; AI is excluded from judges' decision-making.
Policy · Prohibition · Confidentiality
2026-08-28Proposed
Türkiye Barolar Birliği (TBB)
A 44-page principle-based, technology-neutral guide covering lawyers, trainee lawyers and law-office staff. It classifies AI use into risk tiers, sets duties of human oversight, verification, professional secrecy, data minimisation and client information, and adds practical annexes: vendor criteria, model office-policy clauses and client-engagement wording. Advisory, reviewed periodically by TBB.
Guidance · Verification duty · Confidentiality · Client consent · Competence · Supervision · Record-keeping · Prohibition · Disclosure
2026-08-28In force
Prime Minister of Vietnam
Vietnam's current national AI strategy, approved by the Prime Minister on 28 August 2026 and running to 2030 with a vision to 2045. It supersedes Decision No. 127/QD-TTg. It is a development strategy rather than a regulatory instrument and does not address artificial intelligence in legal practice.
Policy · Permissive
2026-08-28In force
PERADI
Indonesia's first bar AI guideline, issued by PERADI on 6 August 2026. It uses permitted, supervised and prohibited risk tiers, treats AI as augmentation rather than a replacement for human reasoning, and requires every AI output to pass through the advocate's own intellectual review before use.
Guidance · Verification duty · Confidentiality · Supervision · Prohibition · Competence
2026-08-06In force
Poland
Poland's AI systems act implements the EU AI Act from August 2026. It creates the Komisja Rozwoju i Bezpieczeństwa Sztucznej Inteligencji as the market-supervision body and single point of contact, names the minister for informatisation as notifying authority, sets cooperation and complaints functions, and phases in key institutional chapters from October 2026.
Statute · Risk classification · Record-keeping · Permissive
2026-07-27In force
Germany
Germany's KI-MIG implements the EU AI Act nationally from 29 July 2026. It designates the Bundesnetzagentur as the default market-surveillance authority and single point of contact, creates an independent AI market-surveillance chamber for listed high-risk systems including justice-related uses, establishes complaint and coordination functions, and requires at least one AI regulatory sandbox.
Statute · Risk classification · Record-keeping · Permissive
2026-07-22In force
Parliament (GR)
Greek implementing act for the EU AI Act, listed by the Special Secretariat for AI and Data Governance. It designates the Hellenic Data Protection Authority as principal market-surveillance authority and EETT as notifying authority, and amends the 2022 emerging-technologies law that already required public-sector algorithmic impact assessments and AI registers. No justice-specific provisions were identified.
Statute · Risk classification
2026-07-20In force
Asamblea Nacional
Bill amending the Organic Code of the Judicial Function so that judges and judicial officials may not use AI-generated information without first verifying its accuracy, relevance and currency; lawyers may not file AI-generated content without verifying it where it could mislead; and incorporating unverified AI content becomes a serious disciplinary offence for judicial personnel.
Statute · Verification duty
2026-07-13Proposed
NAIO
A pre-drafting public consultation run through Malaysia's Unified Public Consultation portal on a proposed AI Governance Bill, open from 10 July to 1 August 2026. No bill had been introduced in the Malaysian Parliament as at 4 September 2026, and justice or legal-sector uses are not separately addressed.
Consultation · Risk classification
2026-07-10Consultation
Norwegian Supreme Court
The Supreme Court of Norway's guidance for lawyers appearing before it, which since April 2025 contains a dedicated section 19 on the use of AI tools. It was added after the Court received its first brief citing statutes and preparatory works that did not exist, and is cross-referenced from every pleading-drafting section of the guidance.
Practice direction / note · Verification duty · Competence
2026-07-01In force
Supreme Court of Nepal
Nepal's judiciary-wide ICT policy approved by the Supreme Court full bench in mid-2026. Strategy 5.2 adopts AI as an innovation tool: AI-based systems will support drafting, research and case management, but AI shall not be used in a way that directly intrudes on judicial decisions, and third-party AI must be assessed for reliability and bias before use.
Policy · Prohibition · Permissive · Competence
2026-06-24In force
Comite de Etica FJ
Twenty-three-principle protocol drafted by the Judiciary Council, National Court, Prosecutor General and Public Defender: no AI may issue, replace or determine judicial decisions; public tools are barred for case data and only institutional tools allowed; independent verification is mandatory; every procedural document must declare when and how AI was used. It awaits formal adoption.
Policy · Prohibition · Verification duty · Disclosure · Confidentiality
2026-06-16Proposed
NRA
A resolution of the Polish Bar Council adopted at its 12–13 June 2026 plenary in Rzeszów amending the advocates' ethics code. It adds § 23e regulating the use of technological tools, including AI, and amends § 19(6). The rules bind Polish advocates and are enforceable through the disciplinary system.
Ethics opinion · Verification duty · Confidentiality · Competence · Supervision · Disclosure
2026-06-13In force
Consejo de Estado
In an extraordinary unification appeal citing rulings the Court's reporting office could not locate, the Council of State held that AI may only assist with searching and organising information, that all AI output requires human verification, that adjudication is non-delegable, and that AI-detection tools are too unreliable to carry judicial weight. It referred the lawyer for discipline.
Judgment (guidance) · Verification duty
2026-06-09In force
Andmekaitse Inspektsioon
The Estonian DPA's standing guidance area on artificial intelligence, explaining the EU AI Act's risk-tiered structure, the interaction between AI and personal-data protection, individuals' rights, and practical duties for organisations adopting AI. It also tracks the Digital Omnibus and AI Omnibus legislative timetable.
Guidance · Risk classification · Competence · Record-keeping
2026-06-01In force
NJPMC
National guidelines approved by the NJPMC at its 57th meeting on 29 April 2026, developed by the National Judicial Automation Committee after consultation with all High Courts. AI is to assist, not replace, judicial decision-making; permitted uses are case management, legal research, predictive analytics and document processing, with bias, privacy and training safeguards.
Guidance · Prohibition · Confidentiality · Competence
2026-04-29In force
MoCDTI
Ghana's first national AI strategy, launched 24 April 2026 by the Ministry of Communication, Digital Technology and Innovations. It targets education, healthcare, agriculture, finance and public service delivery, and proposes an independent Responsible AI Authority within the first year. Justice, courts and legal services are not named as target sectors.
Policy · Permissive
2026-04-24In force
Corte Suprema
The Supreme Court suspended a lawyer for one month and fined her 5 UTM under articles 531 and 542 of the Organic Code of Courts for a cassation brief citing a non-existent 2023 treatise and a non-existent 2021 work, produced with AI; the unintentional-error defence was rejected as a breach of procedural good faith.
Judgment (guidance) · Verification duty
2026-04-22In force
Judicial Yuan
Judicial Yuan guidelines issued by order of 22 April 2026 for courts and judicial agencies using AI systems. Chapter 2 governs user conduct: users must understand that generative AI may hallucinate and must cross-check outputs against official documents, case files and legal databases before use; AI may only assist adjudication and deployments must incorporate risk management.
Guidance · Verification duty · Confidentiality · Prohibition
2026-04-22In force
CGAE
Interpretive circular under Article 23 of the Right of Defence Act construing the Spanish code of conduct. Generative-AI drafting is lawful, but filing AI-assisted documents with unverified errors breaches diligence and responsible technology use and is a serious infraction under Article 125.u EGAE; the signer is solely responsible and tool failure proves a breach of supervision.
Ethics opinion · Verification duty · Supervision · Competence · Record-keeping
2026-04-13In force
CSM
Recommendations proposed by the CSM's AI Monitoring Group for judges' use of AI, citing AI Act Annex III point 8(a). Ten principles include effective human control, non-substitution, exclusive judicial responsibility and transparency; only systems supplied by the case-management platform or the CSM may be used; every AI-assisted decision must carry an express reference to AI use.
Guidance · Disclosure · Verification duty · Prohibition · Confidentiality · Supervision
2026-04-08In force
TSJ NavarraNavarre
Counsel's October 2025 appeal quoted passages attributed to the Constitutional Court, Supreme Court and two regional high courts that did not match any real ruling. The chamber opened Article 247 LEC bad-faith proceedings in February 2026 but archived them without fine after the lawyer apologised and withdrew the citations, while warning about AI hallucination and the non-delegable verification duty.
Judgment (guidance) · Verification duty
2026-04-04In force
JPDP
A regulatory guideline under the Personal Data Protection Act 2010 (Act 709) as amended in 2024. Part E governs AI and generative AI in profiling and automated decision-making, requiring human oversight, barring decisions made on a sole automated factor, and requiring staff training and appointed reviewers.
Regulation · Supervision · Client consent · Competence · Risk classification
2026-04-01In force
CNB
The first explicit French deontological framework for generative AI in law firms, adopted by the CNB's general assembly on 17 March 2026. It permits AI on condition that the lawyer keeps intellectual control and responsibility, and applies professional secrecy, GDPR compliance, competence, prudence, independence, client information and balanced fee-setting to AI-assisted work, with recent case-law examples.
Guidance · Verification duty · Confidentiality · Competence · Client consent · Permissive
2026-03-17In force
ECRAI / Egypt NCAI
Egypt's umbrella AI governance guide, published March 2026 by the Egyptian Center for Responsible AI. It defines a four-tier risk classification with a mandatory dual-check conformity model for high-risk systems, drawing on UNESCO, OECD, ISO/IEC and the G7 Hiroshima Process. It does not name justice or the courts.
Guidance · Risk classification · Disclosure · Competence
2026-03-15In force
High Court of Kenya
Two 2026 High Court rulings in the same litigation set Kenya's de facto position: AI-generated pleadings are procedurally deficient. Mwamuye J held that computer-generated documents or AI outputs cannot substitute for human-drawn documents; Chigiti J then set aside the resulting judgment, reasoning that the Civil Procedure Rules do not authorise AI-drafted pleadings and that their use confers an unfair advantage.
Judgment (guidance) · Verification duty · Prohibition
2026-03-06In force
TBMM AI Research Commission
The parliamentary AI research commission's March 2026 report, the first comprehensive institutional statement of Türkiye's AI regulation debate. It recommends a framework Turkish AI law, a Turkish AI Authority, regulatory sandboxes, a text-and-data-mining exception in copyright law, modernisation of the data protection regime, and ratification of the Council of Europe AI Framework Convention.
Report · Risk classification
2026-03-01In force
TSJ CanariasCanary Islands
First effective Spanish sanction for AI hallucinations. On appeal from the Audiencia Provincial de Santa Cruz de Tenerife, counsel cited 48 Supreme Court rulings and a CGPJ report that did not exist, sourced from ChatGPT without checking CENDOJ. The chamber fined the lawyer about EUR 420-450 for procedural bad faith under Article 247 LEC.
Judgment (guidance) · Verification duty
2026-02-28In force
Korea National Court Administration
Guidebook prepared by a National Court Administration task force of eight judges and distributed to all Korean judges in March 2026. It sets standards and about twenty graded worked examples for using commercial AI in trial and administrative work, with checklists on hallucination, data bias, personal-data breaches and security, and prompt-writing guidance.
Guidance · Verification duty · Confidentiality · Competence
2026-02-24In force
Trib. SiracusaSicily
In a contract dispute counsel cited four Cassazione decisions whose numbers or content did not exist, apparently generated by AI without checking the CED database. The court held that uncritical use of generative AI is gross negligence rather than a mitigating factor and applied Article 96(3)-(4) c.p.c., ordering EUR 30,200 in fees, damages and a court-fund payment.
Judgment (guidance) · Verification duty
2026-02-20In force
Kenyan Senate
Kenya's horizontal AI bill, published 19 February 2026. It creates an Office of the Artificial Intelligence Commissioner, a four-tier risk classification set by regulations, obligations for high-risk systems including human oversight and human rights impact assessments, a public register of high-risk systems, and regulatory sandboxes. Justice and court uses are not named.
Statute · Risk classification · Record-keeping · Disclosure · Client consent · Prohibition
2026-02-19Proposed
AEPD
Guidance for controllers and processors deploying AI agents that plan and execute multi-step tasks with personal data. It identifies agent-specific risks such as autonomy, tool access and chained actions, and directs risk assessment and DPIAs, human oversight, data minimisation and clear processor arrangements. Not legal-sector specific but relied on by Spanish bars for AI use in firms.
Guidance · Risk classification · Supervision · Confidentiality
2026-02-18In force
Supreme Court of the Philippines
The Supreme Court of the Philippines' judiciary-wide AI framework, adopted 18 February 2026 and announced on 19 March 2026. It binds justices and judges, court personnel, court users including members of the Bar and the public, and AI vendors, and creates a permanent Committee on Human-Centered Augmented Intelligence.
Court rule · Disclosure · Verification duty · Prohibition · Confidentiality · Supervision · Record-keeping · Competence
2026-02-18In force
TRG
In a criminal appeal concerning illegal weapons possession the defence cited excerpts from six Relacao and Supreme Court rulings that do not exist. The court called this absolutely inadmissible, explained AI hallucination, held legal professionals to a higher standard, ordered 5 UC in costs and suggested reporting the lawyer to the Ordem dos Advogados.
Judgment (guidance) · Verification duty
2026-02-10In force
CGAE
CGAE white paper finding that about 60 per cent of Spanish lawyers already use AI, mapping practical uses and risks and recommending mandatory digital-competence training, informed human supervision and transparency for every AI tool, and protection of professional ethics and client confidentiality. It is non-binding and precedes the binding Circular 3/2026.
Report · Competence · Supervision · Confidentiality
2026-01-30In force
CGPJ
Binding instruction to Spanish judges and magistrates on AI in jurisdictional work: effective human control, no autonomous decision-making or substitution of judicial reasoning, no AI content without personal critical validation, no unapproved tools except study with open sources and no judicial data, full judicial responsibility, CGPJ supervision of personal-data processing and CGPJ-provided training.
Regulation · Prohibition · Verification duty · Confidentiality · Supervision · Permissive
2026-01-30In force
Korea National Assembly
Korea's horizontal AI statute, promulgated 21 January 2025 and in force from 22 January 2026 with its Enforcement Decree. It defines high-impact AI by sector, imposes transparency duties including notice that generative outputs are AI-generated, and imposes risk-management, explanation and human-oversight duties on high-impact operators. Justice uses are not separately enumerated.
Statute · Risk classification · Disclosure
2026-01-22In force
Taiwan Legislative Yuan
Taiwan's framework AI statute, passed 23 December 2025 and promulgated 14 January 2026. It sets principles for AI development, requires the Ministry of Digital Affairs to establish an internationally aligned risk-classification framework with sectoral tiered rules, requires warnings on high-risk AI products, and obliges government bodies to conduct risk assessments before using AI.
Statute · Risk classification · Disclosure
2026-01-14In force
Conseil d'État
A deontological charter published on 11 December 2025 binding every member, magistrate and staff member of the French administrative jurisdiction — Conseil d'État, cours administratives d'appel, tribunaux administratifs, Cour nationale du droit d'asile and Tribunal du stationnement payant. It sets seven principles for generative AI, excludes high-risk AI systems, and covers external chatbots, which are not blocked but tightly conditioned.
Policy · Verification duty · Confidentiality · Competence · Prohibition · Supervision
2025-12-11In force
Tongzhou District CourtBeijing
A plaintiff's lawyer filed supplementary submissions citing two reference cases generated by an AI model and never verified. The judge found the cited case numbers belonged to unrelated real cases and wrote a criticism into the judgment without a fine. Entered in the People's Court Case Database in December 2025, it sets graduated consequences for unverified AI citations.
Judgment (guidance) · Verification duty
2025-12-01In force
Ústavní soud
The Czech Constitutional Court fined counsel Mgr. Pavel Kehl CZK 25,000 over a constitutional complaint citing decisions that did not exist or whose holdings were grossly misdescribed. The court identified anglicisms, sentence fragments and inconsistent citation formatting as indicators of AI drafting, and held counsel fully responsible for hallucinated arguments.
Judgment (guidance) · Verification duty
2025-12-01In force
Finnish Bar Association
The Finnish Bar's information-security manual, recommendatory in character but functioning as commentary on the binding B 05.1 guideline. Its Annex 5, first added in April 2024 and substantially updated in November 2025, is Finland's principal profession-facing generative-AI guidance for advocates, in force from 1 December 2025.
Guidance · Verification duty · Confidentiality · Competence · Supervision · Record-keeping · Risk classification
2025-12-01In force
NOvA
The Dutch Bar's recommendations on responsible AI use, published at the end of 2025 and organised around the five statutory core values of the profession. They are notable in Europe for recommending that lawyers ask the client's permission before using AI on a file, and for requiring manual verification of every citation and fact.
Guidance · Verification duty · Confidentiality · Client consent · Competence · Supervision · Permissive
2025-12-01In force
KVKK
A 64-page KVKK guide, published November 2025, answering fifteen questions on how Law No. 6698 applies to generative AI. It covers the model lifecycle, controller/processor allocation, the general processing principles, lawful bases, cross-border transfers, transparency, data-subject rights, security, and practical guidance for everyday users and for children's use of AI tools.
Guidance · Client consent · Confidentiality · Record-keeping · Competence
2025-11-24In force
Drzavni zbor
Slovenia's AI Act implementing statute designating five market-surveillance authorities: the Information Commissioner for prohibited practices and high-risk systems in biometrics, criminal justice, migration and democratic processes; AKOS as single contact point and sandbox operator; the Bank of Slovenia; the Insurance Supervision Agency; and the Market Inspectorate. It also sets procedures, offences and fines.
Statute · Risk classification
2025-11-21In force
Corte Suprema
Chile's first Judicial Ethics Code, adopted by auto acordado and published on 19 November 2025. Article 17 permits AI as a support tool but declares judicial decisions always human and non-delegable, requires judges to verify any AI result independently, forbids automatic incorporation of AI text into rulings, and requires control over data shared with external platforms.
Guidance · Verification duty · Prohibition · Confidentiality · Supervision
2025-11-19In force
Supreme Court of the Philippines
The Philippines' new judicial conduct code, promulgated 11 November 2025, is one of the first anywhere to write artificial-intelligence limits into a code of judicial ethics rather than into guidance. Canon VI addresses competence, diligence and the ethical use of technology and artificial intelligence by judges.
Court rule · Verification duty · Prohibition · Competence
2025-11-11In force
CSJ Sala Civil
The Supreme Court annulled a Sincelejo Superior Tribunal ruling that quoted passages from two cassation decisions which do not exist, treating reliance on unverifiable sources as a via de hecho breaching due process, and ordered the Judicial Council and the Rodrigo Lara Bonilla Judicial School to disseminate the decision to all judges.
Judgment (guidance) · Verification duty
2025-11-05In force
ICAMMadrid
Madrid Bar guide keeping human professional judgment at the centre of AI-assisted practice: verify outputs, protect client data, inform clients about AI use, bill time transparently, train lawyers, and vet vendors on no-training clauses, EU data residency and Article 28 GDPR agreements. It includes an AI glossary, an EU AI Act compliance timeline and infographics.
Guidance · Verification duty · Confidentiality · Disclosure · Competence
2025-10-31In force
Supreme Court of the Philippines
The companion conduct code for Philippine court officials and personnel, promulgated on 28 October 2025. Canon III, Section 6 mirrors the judicial code's AI provision, confining AI use to research of legal authorities and requiring the verification of sources before they are cited in any court document.
Court rule · Verification duty · Competence
2025-10-28In force
Congreso
Government bill classifying AI systems as unacceptable, high, limited or no-evident risk, with obligations for high-risk deployers, a Technical Advisory Council under the Ministry of Science, and enforcement through the future Personal Data Protection Agency. Approved by the Chamber of Deputies and before the Senate's Future Challenges Committee since October 2025.
Statute · Risk classification
2025-10-24Proposed
CSJN
Under Acordada 10/2025 the Argentine Supreme Court implemented automatic generation of plain-language summaries by AI for every judgment published from the LEX100 system used by federal and national courts, with more than 110,000 summaries produced by October 2025. The Court also uses AI to classify incoming appeals. Human-review rules were not stated.
Court rule · Permissive
2025-10-15In force
CNF
Letter from CNF President Francesco Greco to all local bar presidents noting that Article 13 of Law 132/2025 obliges professionals to give clients an informativa on the AI systems they use, and circulating a CNF model text lawyers may adopt in dealings with clients and assisted parties, to be disseminated to all members.
Guidance · Disclosure
2025-10-13In force
Malta (MDIA Act)
Maltese subsidiary legislation giving effect to the EU AI Act's institutional requirements: the Malta Digital Innovation Authority is market-surveillance authority, notifying authority, single point of contact and operator of the national AI sandbox, with appeals to the Administrative Review Tribunal. A companion instrument allocates high-risk and prohibited-practice responsibilities to the Information and Data Protection Commissioner.
Regulation · Risk classification
2025-10-10In force
House of Representatives
A pending Philippine House bill, filed 2 October 2025, that would establish a Philippine Artificial Intelligence Commission as the country's national AI regulator. It had not been enacted as at 4 September 2026 and contains no provisions specific to courts, judges or the legal profession.
Statute · Risk classification
2025-10-02Proposed
STJ Rio NegroRio Negro
Rio Negro's judiciary adopted a generative-AI good-practice protocol in October 2024 covering validation of AI content, human control at every stage, bias evaluation and personal-data protection, and in October 2025 extended it to lawyers, experts and other external auxiliaries, authorising corrective sanctions and mandatory referral to bar ethics tribunals for fabricated citations.
Court rule · Verification duty · Supervision · Confidentiality
2025-10-01In force
Parliament (IT)
Italy's national AI framework law, in force 10 October 2025 and read in conformity with the EU AI Act. Article 13 confines AI in the intellectual professions to support tasks and requires clients to be told in clear language which AI systems a professional uses. Article 15 reserves every judicial decision to the magistrate.
Statute · Disclosure · Permissive · Prohibition · Supervision · Competence
2025-09-25In force
ÖRAK
The Austrian Bar's four-page guide of September 2025 on using AI in law firms, with an annexed checklist that an AI provider must sign before a firm may instruct it. It applies § 9(2) RAO confidentiality and the binding § 40(3) RL-BA service-provider requirements to AI suppliers and makes staff AI competence a precondition of use.
Guidance · Confidentiality · Verification duty · Competence · Supervision · Record-keeping
2025-09-23In force
Congreso
Latin America's first AI statute, published 5 July 2023: it promotes AI within Peru's digital transformation under principles of human dignity and rights, names the PCM Secretariat of Government and Digital Transformation as national authority, declares justice among priority sectors, and requires annual reports to Congress. The 2025 implementing decree introduces risk classification.
Statute · Risk classification · Permissive
2025-09-09In force
Advokatforeningen
Guidance produced by the Norwegian Bar Association's professional ethics committee and approved by its board on 5 September 2025, on using AI in legal practice. It is issued under the Association's ethics recommendations and holds that the lawyer always carries final responsibility even where AI is used as a tool.
Guidance · Verification duty · Confidentiality · Competence · Permissive
2025-09-05In force
JFBA
Members-only guidance issued in September 2025 by the JFBA's AI Strategy Working Group setting five points for appropriate use of generative AI in legal practice: keep confidential and client-identifiable information out of external AI services; treat outputs including cited cases as unverified until checked against primary sources; the attorney bears final responsibility.
Guidance · Confidentiality · Verification duty · Competence
2025-09-01In force
Mexican federal collegiate court
Criteria published in Mexico's Semanario Judicial de la Federacion on 22 August 2025 holding that AI may assist a judge only as an auxiliary tool, under an exercise of self-restraint and with respect for human rights, and that deliberation and the final decision remain strictly within the jurisdictional sphere.
Judgment (guidance) · Supervision · Verification duty · Prohibition
2025-08-22In force
Judiciary of Kenya
Kenya's Judiciary is developing an AI Adoption Policy Framework, announced officially in August 2025, and circulated a draft AI Policy with draft Practice Directions for public participation in May 2026. Reporting describes a three-tier risk model, a certificate of human verification for AI-assisted filings, and a bar on AI displacing judicial decision-making. Not in force.
Policy · Certification · Verification duty · Disclosure · Supervision · Risk classification
2025-08-11Consultation
Camara RosarioSanta Fe
Argentina's first appellate pronouncement on AI hallucinations: after the lawyer admitted the cited cases were invented by a chatbot, the court held that delegating research to AI and filing unverified citations is extremely risky and even reckless, declined a fine but referred the lawyer to the Rosario bar for disciplinary review and preventive measures.
Judgment (guidance) · Verification duty
2025-08-01In force
Ministère de la Justice
The French Ministry of Justice's working-group report, published August 2025, on deploying AI across the judicial order. It follows the February 2025 announcement of four priority use cases and sets out safeguards — a charter of AI use for staff, systematic verification of AI output, attention to bias, protection of confidential data and preference for nationally hosted solutions.
Report · Verification duty · Confidentiality · Competence
2025-08-01In force
CPACFCiudad Autonoma de Buenos Aires
Buenos Aires bar guide built on four pillars: responsible use minimising hallucinations and bias; permanent human control with the lawyer as final decision-maker; protection of personal data and professional secrecy, including client authorisation before sharing confidential information; and continuous training. It states that AI does not exempt the lawyer from responsibility for anything signed or filed.
Guidance · Verification duty · Confidentiality · Competence · Client consent
2025-07-11In force
Malaysian Bar
The Malaysian Bar's principal AI instrument, issued 3 July 2025 by the Bar Council Cyber and Privacy Laws Committee and the Legal Tech, AI and Sandbox Committee. Its annexed advisory sets out recommended and prohibited uses, a nine-risk mitigation table and a four-R framework for practitioners adopting generative AI.
Guidance · Verification duty · Confidentiality · Competence · Supervision · Record-keeping
2025-07-03In force
Senate of the Philippines
A pending Philippine Senate bill, filed 2 July 2025, proposing risk-based regulation of artificial-intelligence systems modelled on the European Union AI Act. It had not been enacted as at 4 September 2026 and does not separately classify judicial or legal-services uses of AI as high risk.
Statute · Risk classification
2025-07-02Proposed
Swedish Bar Association
Guidance dated Stockholm, 12 June 2025, from a Swedish Bar Association working group on using generative AI in legal practice. It ties GenAI use to the Code of Conduct (VRGA) across four areas: regulatory compliance, advisory liability, professional secrecy, and principal liability and office organisation.
Guidance · Verification duty · Confidentiality · Competence · Supervision · Client consent · Permissive
2025-06-12In force
Japan Diet
Japan's first framework AI statute, promulgated 4 June 2025 and fully in force from 1 September 2025. It establishes the AI Strategy Headquarters and a basic AI plan and sets duties of effort for the state, local governments, businesses and citizens. It contains no risk classification, no penalties, and no provisions specific to courts or legal services.
Statute · Permissive
2025-06-04In force
Norwegian Ministry of Digitalisation
A Norwegian Ministry of Digitalisation consultation on a draft Act on artificial intelligence that would incorporate Regulation (EU) 2024/1689 into Norwegian law through the EEA Agreement. It covers the division of responsibility between supervisory authorities, sanctions and geographic scope. The consultation closed on 30 September 2025.
Consultation · Risk classification
2025-06-01Consultation
Denmark
Denmark's national supplement to the EU AI Act, adopted 14 May 2025 and in force from 2 August 2025. It designates Digitaliseringsstyrelsen, Datatilsynet and Domstolsstyrelsen as national competent authorities, sets market surveillance powers, inspection rights and criminal penalties, and creates a bespoke split of supervision over courts' AI use.
Statute · Risk classification · Supervision · Prohibition
2025-05-14In force
STJ JujuyJujuy
Jujuy's Superior Court created a Judicial Innovation Laboratory and approved a Protocol for the Use of AI in the Judiciary as Annex I, with an authorisation form for AI use as Annex II, applying to all users within the provincial judiciary. The annex text was not accessible, so the substantive parameters remain unverified.
Court rule · Supervision · Record-keeping
2025-04-11In force
Supreme Court of Pakistan
Supreme Court of Pakistan judgment of April 2025 in a rent dispute in which Justice Syed Mansoor Ali Shah examined AI in the judiciary: AI can assist with research, drafting and case management but cannot replace judicial discretion, must respect the Article 10A fair-trial right, and hallucinated references require human verification.
Judgment (guidance) · Verification duty · Prohibition
2025-04-08In force
CSJ Santa FeSanta Fe
Minimum acceptable-use standards for judges, defenders and staff granted generative-AI tools by the Santa Fe Supreme Court: independent verification of all outputs, protection of sensitive case data, AI as a complementary assistant with decisions remaining exclusively the judge's, and transparency about AI-generated content, drawing on UNESCO and Council of Europe frameworks.
Guidance · Verification duty · Confidentiality · Supervision · Disclosure
2025-04-04In force
KIRP
A 22-recommendation handbook published by the Polish National Chamber of Legal Advisers in 2025, prepared with Microsoft and the Sołtysiński Kawecki & Szlęzak law firm, on adopting AI tools in legal practice. It runs from tool selection and contract review through professional secrecy, GDPR and client communication to training and incident management.
Guidance · Verification duty · Confidentiality · Competence · Supervision · Client consent · Disclosure · Record-keeping
2025-04-01In force
SAK
Binding Slovak Bar rules on AI in legal practice. Confidential information and personal data may enter AI tools only in one of three modes: a local on-premise tool, a GDPR Art. 28 processor with a no-training guarantee, or express informed client consent waiving confidentiality. Verification of output is mandatory; billing self-training time is barred.
Regulation · Confidentiality · Client consent · Verification duty · Competence · Supervision · Record-keeping · Prohibition
2025-03-31In force
MICDE
Kenya's non-binding national AI strategy, launched 27 March 2025. It sets three pillars — AI digital infrastructure, data ecosystems and governance, and AI research, innovation and commercialisation — with cross-cutting enablers on ethics, skills, finance and governance. It underpins the risk-based logic later carried into the Artificial Intelligence Bill 2026.
Policy · Permissive
2025-03-27In force
NDPC
The binding directive implementing Nigeria's Data Protection Act 2023. Article 43 imposes obligations on any controller or processor deploying artificial intelligence to process personal data — documented technical and organisational parameters, a DPIA filed with the Commission, anonymisation and low-risk testing, disparate-outcome review, and continuous monitoring after deployment.
Regulation · Risk classification · Record-keeping · Confidentiality · Client consent
2025-03-20In force
CNDP (Morocco)
Morocco's data protection authority's first public position on AI, issued in Rabat on 18 March 2025. It confirms that AI processing of personal data is governed by Law 09-08, flags automated decision-making as requiring particular safeguards, and announces that work on a dedicated deliberation on AI processing has begun.
Guidance · Confidentiality · Disclosure · Client consent
2025-03-18In force
Trib. FirenzeTuscany
First reported Italian hallucination case. In an IP interim appeal the respondent's brief cited Cassazione rulings that did not exist; counsel said a trainee had used ChatGPT without his knowledge. The court rejected an Article 96 c.p.c. reckless-litigation claim because the citations did not affect the defence and no bad faith was shown, but censured the failure to verify.
Judgment (guidance) · Verification duty · Supervision
2025-03-14In force
CAC
Regulation in force from 1 September 2025 requiring providers to attach explicit labels and implicit metadata labels identifying provider and generation details to AI-generated text, audio, images and video, and requiring distribution platforms to verify metadata and add notices. Relevant to authentication of AI-generated material in litigation, although not itself an evidence rule.
Regulation · Disclosure
2025-03-07In force
OVB / Avocats.be
Joint guidelines of Belgium's two bar federations, dated 1 February 2025, on lawyers' use of AI. They add no new deontological duties but apply existing ones: AI is neither forbidden nor mandatory, cited sources must be verified, professional secrecy must never enter a tool, and there is no duty to tell clients AI was used.
Guidance · Permissive · Verification duty · Confidentiality · Competence
2025-02-01In force
Korea Judicial AI Research Group
Non-binding guidelines finalised in February 2025 by the Judicial AI Research Group, a study society of judges founded in October 2023. They propose that courts may require parties to disclose AI use in submitted documents, including the tool, prompts and verification steps; warn about deepfake evidence; and state that existing procedural law governs admissibility.
Report · Disclosure · Verification duty
2025-02-01In force
Egypt NCAI
Egypt's second national AI strategy, launched by the President in January 2025 for 2025-2030. It identifies the judiciary as a target sector for the national Arabic large model and records that Egyptian-accent speech recognition has already been used to produce court minutes. It sets no rules for lawyers or judges.
Policy · Permissive
2025-01-16In force
Ministry of Justice (HR)
Implements the 2024 Courts Act duty to publish all court decisions. The ANON AI system performs automatic and semi-automatic anonymisation, metadata extraction and thematic classification; first-instance decisions are anonymised automatically, while higher-court and Supreme Court decisions are anonymised by authorised staff from ANON's proposals. The judge may order further anonymisation where persons could be identified.
Regulation · Supervision · Confidentiality
2025-01-01In force
NPC Philippines
The National Privacy Commission's advisory applying Republic Act 10173, its implementing rules and NPC issuances to AI systems that process personal data across development, training, testing and deployment. It is the Philippines' baseline data-protection instrument for legal-sector AI use by firms and public bodies alike.
Guidance · Client consent · Confidentiality · Record-keeping · Risk classification
2024-12-19In force
Danske Advokater
Two guides published on 17 December 2024 by Danske Advokater, the Danish legal-sector trade association. The first explains the EU AI Act's significance for law firms, including risk categories and provider versus deployer roles. The second is a practical checklist of considerations before buying and deploying AI in a firm.
Guidance · Confidentiality · Competence · Permissive
2024-12-17In force
CSJ
Binding guidelines for all Colombian judges and judicial employees. Article 8 bars free commercial chatbots, use of AI to assess evidence, weigh facts or decide, and entry of personal or confidential data into tools that retain inputs; article 9 requires verification of every output; article 10 requires express disclosure in the file of tool, model, date, purpose and prompts.
Regulation · Verification duty · Disclosure · Prohibition · Confidentiality · Competence · Supervision · Record-keeping
2024-12-16In force
Estonian Bar Association
The Estonian Bar Association's guide on artificial intelligence for advocates, published on 3 December 2024. It is addressed to advocates and law-office staff and sets out verification, competence and confidentiality expectations on the founding principle that the advocate, never the AI system, provides the legal service.
Guidance · Verification duty · Confidentiality · Competence
2024-12-03In force
BRAK
The German Federal Bar's orientation guide, dated December 2024 and announced on 9 January 2025, on using AI in a law firm consistently with professional law. Eight sections read §§ 43, 43a(2), 43e BRAO and § 5 BORA onto AI tools, set out the AI Act duties that fall on lawyers as deployers, and flag copyright, tax and professional-indemnity risks.
Guidance · Verification duty · Confidentiality · Competence · Supervision · Permissive
2024-12-01In force
AAIP
Argentina's data-protection authority's lifecycle guide for AI systems: privacy by design and impact assessment at design stage, bias and error testing at validation, documented privacy policies at implementation, and incident reporting, monitoring and complaint channels in operation. Not legal-sector specific but applies to firms and courts processing personal data with AI.
Guidance · Risk classification · Record-keeping
2024-09-01In force
CNB
A 42-page practical guide published by the CNB's AI working group in September 2024 to help French lawyers understand and adopt generative AI. It explains how generative AI is built, sets out use cases and risks, and gives guidance on formulating prompts, evaluating and improving generated output, and training. It is accompanied by a tool-selection grid.
Guidance · Permissive · Competence · Verification duty
2024-09-01In force
MOSTI
Malaysia's voluntary national AI governance and ethics guidelines, published by MOSTI in September 2024. They set cross-sector principles for AI developers, deployers and end users, sit alongside the National AI Action Plan, and contain no provisions specific to courts, judges or the legal profession.
Guidance · Permissive · Risk classification
2024-09-01In force
NBA-SLP
Nigeria's only profession-specific AI guidance. Issued by the NBA Section on Legal Practice on 28 August 2024, it treats AI as augmentative, insists on human oversight, and sets expectations on confidentiality, competence, critical review of outputs, client transparency and consent, supervision of staff, vendor selection and NDPA 2023 compliance. Non-binding.
Guidance · Verification duty · Confidentiality · Competence · Supervision · Client consent · Disclosure
2024-08-28In force
SIC
Binding instruction from Colombia's data-protection authority to all controllers it supervises: personal-data processing in AI systems must satisfy suitability, necessity, reasonableness and strict proportionality under Laws 1266/2008 and 1581/2012, with prior risk identification and mitigation, privacy by design and by default, staff training and continuous monitoring of security measures.
Regulation · Risk classification · Record-keeping
2024-08-21In force
Corte Constitucional
Reviewing a tutela in which a Cartagena circuit judge used ChatGPT to help reason a health-rights ruling for a child with autism, the Court found no breach of due process on the facts but set criteria for judicial AI use and ordered the Superior Council of the Judiciary to issue guidelines within four months.
Judgment (guidance) · Verification duty · Supervision · Disclosure
2024-08-02In force
FMCIDE / NITDA-NCAIR
Nigeria's national AI policy framework, published August 2024 and still marked draft. It sets principles of responsible conduct, inclusivity, innovation, sustainability, collaboration and data ethics across five pillars covering infrastructure, ecosystem, adoption, ethical development and governance. It does not mention the judiciary, the legal profession or the administration of justice.
Policy · Permissive
2024-08-01Proposed
TBMM
Türkiye's first general AI bill, submitted to the Grand National Assembly on 25 June 2024 with Ömer Faruk Gergerlioğlu as first signatory. It has not been reported out of committee. It contains no justice-sector or legal-practice risk classification, and no horizontal AI statute is in force in Türkiye as at 4 September 2026.
Statute · Risk classification
2024-06-25Proposed
SAV/FSA
The Swiss Bar Association's guideline on AI, approved 14 June 2024 and updated 16 February 2025. It focuses on generative AI, sets a three-route test for putting client information into a tool, requires independent verification of output, addresses liability, copyright and notification duties, and explains the Swiss, EU and Council of Europe regulatory position.
Guidance · Verification duty · Confidentiality · Client consent · Disclosure · Permissive
2024-06-14In force
DSK
Guidance of 6 May 2024 from the conference of Germany's federal and state data protection authorities on using AI applications, in particular large language models, in compliance with data protection law. It is addressed primarily to controllers deploying AI and is the reference text the BRAK directs law firms to.
Guidance · Confidentiality · Client consent · Record-keeping · Supervision
2024-05-06In force
Finnish Bar Association
The Finnish Bar's binding information-security guideline, adopted by its delegation and in force in its current form since 1 January 2024. Eighteen numbered duties on advocates covering competence and documented training, security policy, external audits, business-grade software, device and access management, encryption, backups, supplier contracts and secure disposal.
Regulation · Confidentiality · Competence · Client consent · Supervision · Record-keeping
2024-01-01In force
Gobierno de Espana
Digital-efficiency measures for the Spanish justice system. It distinguishes automated actions of simple processing, proactive actions such as alerts, and assisted actions in which systems including AI generate full or partial drafts to support judges, prosecutors and court lawyers; a draft cannot become a resolution without the operator's intervention and validation.
Statute · Supervision · Permissive
2023-12-20In force
Komdigi
Indonesia's only in-force AI instrument as at September 2026: a ministerial circular of 19 December 2023 setting nine ethical values for artificial intelligence, addressed to AI businesses and to public and private electronic system operators. It makes no mention of courts, judges or legal services.
Policy · Permissive · Risk classification · Confidentiality
2023-12-19In force
Bar Council Malaysia
The Bar Council's first AI circular, issued 24 November 2023 by its Technology, Cyber and Privacy Laws Committee, setting out the risks of ChatGPT in legal practice and the precautions practitioners should take. It remains in force alongside the 2025 circular that supplements it.
Guidance · Verification duty · Confidentiality · Competence
2023-11-24In force
ČAK
A six-point non-binding opinion of the Czech Bar Association. Lawyers may use any technology, including AI, provided they meet all statutory and professional duties. AI cannot itself provide legal services; entering client data into AI tools will normally breach confidentiality unless the client gives specific prior informed consent.
Ethics opinion · Confidentiality · Client consent · Verification duty · Prohibition · Permissive
2023-09-20In force
Judicial Yuan
Judicial Yuan press release of 27 August 2023 describing its generative-AI system that drafts judgment text for dangerous-driving and fraud-accessory cases, with drug offences piloted from December 2023. Fact-finding, application of law and sentencing remain entirely with the judge; the model is fine-tuned and hosted on-premises to protect privacy and official secrets.
Policy · Prohibition · Confidentiality
2023-08-27In force
Japan MoJ
Ministry of Justice guideline of August 2023 explaining when AI-assisted contract drafting, review and management services fall outside the Attorney Act Article 72 prohibition on unauthorised legal practice. A service breaches Article 72 only if provided for remuneration, concerning a legal case, and constituting legal affairs; generative-AI services are assessed on the same framework.
Guidance · Permissive
2023-08-01In force
CAC
Horizontal regulation on public-facing generative-AI services in force since 15 August 2023. Providers must use lawfully sourced training data, obtain consent for personal data, label generated content, undergo security assessment and algorithm filing where services have public-opinion attributes, protect user inputs, and handle access, correction and deletion requests. Not specific to legal services.
Regulation · Risk classification · Record-keeping · Client consent
2023-07-10In force
Japan PPC
Personal Information Protection Commission alert of 2 June 2023 cautioning businesses and public bodies on entering personal data into generative-AI prompts: use only within the stated purpose, and do not input data that may be used for model training without a legal basis. It also issued a notice to OpenAI on collection of sensitive data.
Guidance · Confidentiality
2023-06-02In force
Egypt NCAI / MCIT
Egypt's first responsible-AI instrument, adopted by the National Council for Artificial Intelligence in April 2023. It is voluntary and sets principles of human-centred design, transparency, justice, accountability and security for AI development and deployment. It contains no provision addressed to courts, judges, lawyers or legal services.
Guidance · Competence · Supervision · Permissive
2023-04-19In force
SPC
Supreme People's Court normative opinion of 8 December 2022 governing AI across Chinese courts. AI output may only be an auxiliary reference: it shall never substitute for the judge's decision in any case, users may decline AI assistance, and accountability remains with the decision-maker. Systems must be lawful, secure, fair and transparent.
Policy · Prohibition · Client consent · Risk classification
2022-12-08In force
Descriptive reference only: this page records what each rule requires. Missing a document? Tell us.
Afghanistan 
Albania 
Armenia 
Argentina 
Austria 
Azerbaijan 
Bosnia and Herzegovina 
Bangladesh 
Belgium 
Bulgaria 
Brunei 
Bolivia 
Botswana 
Belarus 
Switzerland 
Côte d'Ivoire 
Chile 
Cameroon 
China 
Colombia 
Costa Rica 
Cuba 
Cyprus 
Czechia 
Germany 
Denmark 
Dominican Republic 
Algeria 
Ecuador 
Estonia 
Egypt 
Spain 
Ethiopia 
Finland 
France 
Georgia 
Ghana 
Greece 
Guatemala 
Honduras 
Croatia 
Hungary 
Indonesia 
Iraq 
Iran 
Iceland 
Italy 
Jamaica 
Jordan 
Japan 
Kenya 
Kyrgyzstan 
Cambodia 
South Korea 
Kazakhstan 
Laos 
Lebanon 
Sri Lanka 
Lithuania 
Luxembourg 
Latvia 
Morocco 
Moldova 
Montenegro 
North Macedonia 
Myanmar 
Mongolia 
Macao 
Malta 
Mauritius 
Mexico 
Malaysia 
Mozambique 
Namibia 
Nigeria 
Nicaragua 
Netherlands 
Norway 
Nepal 
Oman 
Panama 
Peru 
Philippines 
Pakistan 
Poland 
Puerto Rico 
Portugal 
Paraguay 
Romania 
Serbia 
Russia 
Rwanda 
Sweden 
Slovenia 
Slovakia 
Senegal 
El Salvador 
Thailand 
Tunisia 
Türkiye 
Trinidad and Tobago 
Taiwan 
Tanzania 
Ukraine 
Uganda 
Uruguay 
Uzbekistan 
Venezuela 
Vietnam 
Zambia 
Zimbabwe