Judicial AI Use Tracker (How are Judges using AI?)
Below is the Judicial AI Use Tracker, which monitors how judges are using AI and records any official guidance, commentary, or policy developments on judicial use.
This tracker captures both the benefits and challenges of AI in judicial contexts ranging from tools that enhance efficiency, consistency, and access to justice, to cases and discussions highlighting risks. It aims to provide a balanced picture of how AI is shaping judicial reasoning, case management, and evidence assessment.
Tracker Status: Active/Monitoring
Publication Date: 21 May 2025
Last Verified: 5 August 2026
Latest Case Chronologically: Campos v Munoz (US)
Latest Legal Article: 1,000+ Suspected or Confirmed AI Hallucination Cases: UK Reaches 54, Concerns of Judgments containing Hallucinations Mount in India and Possible Citation Errors in Canada
Author and Contact: Matthew Lee (Barrister) click here for details.
Ad/Marketing Communication
UK‑based legal commentary and comparative analysis of international case law on AI related legal issues. This Judicial AI Use Tracker (how are judges using AI) forms part of lecturing/teaching law and writing/editing law articles/reports and is communicated solely in connection with promoting or advertising Matthew Lee’s practice. Not legal advice. Not Direct/Public Access. All instructions via clerks at Doughty Street Chambers.
What the Tracker and Charts Will Show (when research is complete)
The public Judicial AI Use Tracker below lists Dates, Case Name/Guidance, Country, Area of Law, Actor, AI Tool, a Brief Description of AI Use and the year. There are hyperlinks on the Case Name/Guidance column to show the source where available.
Key Stats Line and Charts
Judicial AI Use Tracker
| Date | Case Name/Guidance | Country | Area of Law | Actor | AI Tool | Notes |
|---|---|---|---|---|---|---|
| 31-Jul-2026 | F.J.L. v T.G.O (2025/220239) [2026] ZAGPJHC 875 | South Africa | Family | Judge | AI suspected not confirmed | A Full Court identified serious citation and quotation errors in an Acting Judge’s earlier reasons, including a fictitious authority. AI hallucination was considered a plausible explanation, but the Court expressly made no finding that AI had been used. |
| 29-Jul-2026 | Practice Direction HC 142: Responsible Use of Generative Artificial Intelligence in Court Documents | Ireland | Civil | Irish High Court | All | The High Court permitted responsible GenAI use in civil proceedings subject to independent human verification and control, with specific requirements concerning witness statements, expert reports and disclosure. Non-compliance may attract procedural or costs consequences. |
| 8-Jul-2026 | Expediente Administrativo n.º 0000376-49.2026.2.00.0826 (TJSP) | Brazil | | | | Awaiting Credible Translation |
| 2-Jul-2026 | Singh v Jammu and Kashmir Bank Ltd and another, Civil Appeal No. 11950 of 2025, 2026 INSC 668 | India | Company | Judge | Unclear | The NCLT relied on six fake or inaccurate authorities apparently obtained through its own research, which were then repeated by the NCLAT. The Supreme Court set both decisions aside and strongly condemned judicial reliance on unverified AI-generated authorities. |
| 10-Jun-2026 | Henry County Schools v Grant; Henry County Schools v Douglas; Henry County Schools v Forbes, A26A0698–A26A0700 | US | Employment | Judge | AI suspected not confirmed | The trial court adopted an order submitted by the plaintiffs' lawyer which ignored binding law, misquoted cases and relied on at least one hallucinated case. The Georgia Court of Appeals reversed the state-law rulings, vacated the Title VII analysis and remanded for reconsideration using genuine authorities. |
| 9-Jun-2026 | Crown Court AI Pilots: AI Tech Ambition to Deliver Smarter Justice for Victims | UK | Crime | Judge | Unclear | The Ministry of Justice announced pilots and planned AI tools to assist judges with identifying trial-ready cases, grouping hearings, legal research and case analysis. The announcement concerned experimental or planned use rather than general deployment. |
| 3-Jun-2026 | Draft Regulations for Use of Artificial Intelligence (AI) in Courts, 2026 | India | All | Judicial Administration | All | Draft regulations proposed allowing AI for research, case management, transcription, translation and similar support while prohibiting AI-only adjudication, credibility assessment and certain predictive uses. Human judicial responsibility would remain paramount. |
| 28-May-2026 | H. C., a Minor, etc v CONTRERAS | US | Family | Judge | AI suspected not confirmed | Counsel relied on a fictitious authority and materially misstated the law. Although those errors were identified before judgment, the trial court’s ruling repeated them; there was no finding that the judge personally used AI. |
| 14-May-2026 | Mother v Father, B [2026] EWFC 107 (B) | UK | Family | Judge | Co-Pilot | With the advocates’ agreement, the judge used secure Judicial Copilot to generate simplified versions of the judgment for parents with learning and cognitive difficulties. The advocates considered the AI-generated summaries particularly useful. |
| 5-May-2026 | Hannah Renea Payne v The State, No. S26A0459, 324 Ga. 305 (2026) | US | Crime | Judge | Unclear | A prosecutor used AI to draft submissions and a proposed order without verifying the generated authorities. The trial judge substantially adopted the order, including fictitious or misattributed cases; the Supreme Court vacated it and warned courts to scrutinise AI-assisted proposed orders. |
| 30-Apr-2026 | SNV Aviation Pvt Ltd v ABS Tour and Travels, RFA(COMM) 284/2026 | India | Civil | Judge | AI suspected not confirmed | The Delhi High Court found that the district judgment attributed non-existent propositions to real Supreme Court decisions and said its drafting prima facie suggested AI use. It stayed the ₹1.088 crore decree subject to a ₹20 lakh deposit. AI use was not conclusively established. |
| 22-Apr-2026 | Speech by the Chancellor of the High Court: Legal professional privilege in the Age of AI | UK | All | Judge | All | "thanks to important work done internally involving collaboration between the judiciary, the Judicial Office, His Majesty’s Courts and Tribunals Service (HMCTS) and the Ministry of Justice’s AI unit, we now have two AI systems which we are satisfied are secure. One is a secure form of the Microsoft Copilot system and that is available to all judges in England and Wales. The other is a secure system based on in-house development with by HMCTS and MOJ." |
| 7-Apr-2026 | B. S. M. s/ determinación de la capacidad jurídica, Expte. n.º 1-74985-2025 | Argentina | | | | Awaiting Credible Translation |
| 26-Mar-2026 | Comisión Nacional de Disciplina Judicial, Radicación n.º 150011102000202000358 02 | Colombia | | | | Awaiting Credible Translation |
| 5-Mar-2026 | Campos v Munoz | US | Family | Family Commissioner | AI suspected not confirmed | Counsel supplied fictitious authorities which were incorporated into a proposed order and signed by the family commissioner. The appellate court criticised the failures to verify the authorities and sanctioned counsel, emphasising the responsibility of courts and lawyers to guard against AI-generated fabricated law. |
| 3-Mar-2026 | High Court of Australia Generative AI Experimental Testing | Australia | All | Judicial Administration | All | The High Court adopted internal guidelines permitting Justices and staff experimentally to use an approved GenAI tool for supportive tasks including summarising lengthy documents and preliminary research. AI is prohibited from drafting substantive judicial reasoning or substituting for judicial discretion or independent analysis |
| 27-Feb-2026 | Gummadi Usha Rani & Anr v Sure Mallikarjuna Rao & Anr, SLP(C) No. 7575/202 | India | Civil | Judge | Unclear | A trial court relied on four non-existent authorities, subsequently identified as AI-generated. The Supreme Court treated judicial reliance on fake AI material as a serious institutional concern potentially affecting the integrity of adjudication. |
| 18-Feb-2026 | Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary | Philippines | All | Judge | All | The Supreme Court adopted a judiciary-wide framework permitting AI for specified support functions while prohibiting it from becoming the sole, primary or determinative basis for adjudication. Human oversight, disclosure, traceability and accountability remain required. |
| 1-Dec-2025 | Fiscalía General de la Nación v Luis Alberto Contreras Fandiño (Tribunal Superior de Bogotá, Sala Penal) | Colombia | | | | Awaiting Credible Translation |
| 18-Nov-2025 | Kim v Insurance Australia Ltd t/as NRMA (Motor Accident Injuries) [2025] ACAT 87 | Australia | Civil | Tribunal Judge | Bard | The ACAT reasons cited two non-existent authorities. The decision was removed from the tribunal website, and the registrar later confirmed misuse of generative AI. The underlying insurance review had been dismissed. |
| 31-Oct-2025 | Updated Artificial Intelligence (AI) Guidance for Judicial Office Holders | UK | All | Courts and Tribunals Judiciary | All | This refreshed guidance has been developed to assist judicial office holders in relation tothe use of Artificial Intelligence (AI). It updates and replaces the guidance documentissued in April 2025 |
| 22-Aug-2025 | Evans v HMRC | UK | Tax | Judge | Co-Pilot | Judge expressly stated how and why AI was used: “I have used AI in the production of this decision This application is well-suited to this approach. It is a discrete case-management matter, dealt with on the papers, and without a hearing. The parties’ respective positions on the issue which I must decide are contained entirely in their written submissions and the other materials placed before me. I have not heard any evidence; nor am I called upon to make any decision as to the honesty or credibility of any party.” |
| 23-Jul-2025 | re CorMedix Inc Securities Litigation, No. 2:21-cv-14020 | US | TBC | Law Student or Intern | ChatGPT | Judge Neals later confirmed that a temporary law-school intern used ChatGPT for legal research when preparing a draft opinion. The draft was docketed before adequate review and contained serious citation and quotation errors before being withdrawn. |
| 20-Jul-2025 | Jackson Federation of Teachers et al v Lynn Fitch et al | US | Education, Civil Rights, Constitutional | Law Clerk | Perplexity | A law clerk used Perplexity as a drafting assistant when preparing an order. An early draft containing factual and legal errors was accidentally docketed before normal chambers review, leading the judge to introduce additional verification safeguards. |
| 30-Jun-2025 | Shahid v Esaam, A25A0196 | US | Family | Judge | AI suspected not confirmed | A trial order apparently prepared by the husband's lawyer relied on two fictitious cases. The Georgia Court of Appeals said the wider filing pattern suggested generative AI, vacated the order, required a new hearing and imposed a $2,500 penalty on counsel. |
| 14-Apr-2025 | Judicial Guidance on AI: England & Wales | UK | All | Courts and Tribunals Judiciary | All | Updated guidance addressed responsible judicial use of AI and the secure Microsoft Copilot Chat available to judicial office holders. It was subsequently replaced by the October 2025 guidance. |
| 26-Mar-2025 | Rotterdam District Court AI Sentencing-Reasons Pilot | Netherlands | Crime | Judge | Unclear | Judges and a court lawyer used AI as a writing aid when drafting sentencing reasons in two criminal judgments. The court stressed that AI did not make or weigh the judicial decisions and that all suggested wording remained subject to human review. |
| 20-Feb-2025 | Ross v. United States (No. 23-CM-1067) | US | Crime | Judge | ChatGPT | Judges used ChatGPT to test an asserted matter of “common knowledge”, with different prompts producing materially different answers. The opinions used the exercise principally to illustrate the limitations of relying on LLM outputs as evidence. |
| 30-Dec-2024 | Buckeye Trust v Principal Commissioner of Income Tax, ITA n.º 1051/Bang/2024 | India | | | | Awaiting Credible Translation |
| 23-Oct-2024 | Tribunal da Relação de Lisboa judgment concerning Helena Lopes da Costa and others, 23 October 2024 | Portugal | | | | Awaiting Credible Translation |
| 5-Sep-2024 | United States v Deleon | US | Crime | Judge | ChatGPT, Claude, Gemini | In a concurrence, Judge Newsom queried several leading LLMs about the ordinary meaning of “physically restrained” as part of his continuing consideration of whether AI might assist judicial interpretation. |
| 7-Jun-2024 | Rechtbank Gelderland, ECLI:NL:RBGEL:2024:3636 | Netherlands | Civil | Judge | ChatGPT | The judge disclosed using ChatGPT when estimating the lifespan of solar panels and other factual matters relevant to calculating damages. The judgment did not identify the prompts, underlying sources or verification process. |
| 23-May-2024 | Md Zakir Hussain v State of Manipur & Ors, WP(C) No. 70 of 2023 | India | Public Law, Employment | Judge | ChatGPT | After the State failed to explain the governing procedures, the judge used Google and ChatGPT 3.5 for additional research. The Court ultimately applied conventional legal principles to find that the petitioner’s removal breached natural justice. |
| 1-Mar-2024 | Re: Bail Order | India | Crime | Judge | Unclear | Used AI tool to get a broader picture of bail jurisprudence when an assault is one that involves cruelty. |
| 28-May-2023 | Snell v United Specialty | US | Civil | Judge | Bard, ChatGPT | In a concurrence, Judge Newsom used ChatGPT and Google Bard to explore the ordinary meaning of “landscaping” and whether installing an in-ground trampoline fell within that meaning, as part of a wider discussion of LLMs as interpretative tools. |
| 15-Sep-2023 | Article in Guardian | UK | Unclear | Judge | ChatGPT | Lord Justice Birss disclosed that he had used ChatGPT to produce a paragraph summarising an area of law he already understood and then checked the result. The judgment in which the paragraph appeared was not identified. |
| 23-Mar-2023 | AM (a juvenile) v State | Pakistan | Crime | Judge | ChatGPT | After determining the bail application on conventional grounds, the judge conducted an experimental exchange with ChatGPT-4 and reproduced it in the order. He expressly stated that the decision to grant bail was not based on the AI responses. |
| 30-Jan-2023 | Padilla autism‑therapy ruling | Colombia | TBC | Judge | ChatGPT | The judge posed four questions to ChatGPT and reproduced the answers in a constitutional health judgment. He stated that AI did not replace the judicial decision and that the information used had been checked. |
| 2-Jun-2022 | China | TBC | TBC | TBC | Zhejiang courts deployed the “Xiaozhi” AI assistant to analyse case materials, identify issues in dispute and assist in producing judgment documents in financial-loan cases. The system supports rather than replaces human judges. | |
| 30-Aug-2018 | Project VICTOR, Brazil | Brazil | TBC | TBC | Victor | Brazil’s Supreme Federal Court developed Project VICTOR to use machine learning to classify extraordinary appeals and identify issues concerning “general repercussion”. It assists judicial triage rather than determining the merits of cases. |
| 13-Jul-2016 | State v Loomis | US | Crime | TBC | TBC | The sentencing court considered a COMPAS risk assessment alongside other sentencing factors. The Wisconsin Supreme Court permitted such consideration subject to strict limits, including that COMPAS must not determine incarceration or sentence severity. |
Judicial AI Use Tracker FAQ
What does this data represent?
Each row represents an instance where AI has been referenced, used, or discussed within a judicial or official context. This includes both formal judicial reasoning and informal judicial commentary, such as speeches or ethical guidance.
What is the purpose of the Judicial AI Use Tracker?
The tracker provides an overview of how AI is entering judicial practice capturing both positive uses (e.g., enhanced efficiency, improved access to justice, analytical support) and potential risks (e.g., bias, overreliance, opacity). Its purpose is to understand emerging patterns in judicial engagement with AI.
What kind of AI tools are included?
The tracker includes a range of tools such as legal research assistants, case prediction systems, document analysis tools, transcription aids, and language models used for drafting or reasoning support.
How often is the Judicial AI Use Tracker updated?
The tracker is updated periodically as new cases, judicial comments, or official guidance emerge. Each update refines the dataset and contributes to a clearer picture of the evolving relationship between AI and judicial practice.
