IAPQB

Courts write rules for their own AI use

How courts in England and Wales and in India limit AI use by judges and court staff

IAPQB Report No. 29 · Rule explainer · Published 17 November 2026
International AI Practice Qualification Board (IAPQB)

In one minute

  • Judges in England and Wales have had AI guidance since December 2023. They may not put non-public information into public chatbots, and only approved AI tools may be used for judicial work [1, 2].
  • In July 2025 the Kerala High Court became the first in India to adopt an AI policy for its district courts. It bars AI from findings, orders and judgments [3, 4, 5].
  • The Gujarat High Court followed in April 2026 with a detailed list of banned uses [5, 6, 7]. India's Supreme Court has called a decision based on fake AI-made judgments misconduct, and set aside a tribunal ruling built on invented precedents [8, 9].
  • The rules share four points: the judge decides, case work needs approved tools, case material stays out of public tools, and the person who signs answers for each citation. India's national rules were still a draft on 16 November 2026 [1, 2, 6, 10, 11, 12].

1. A familiar situation

A district judge has a long list of cases and a judgment due by evening. A free chatbot could summarise the witness statements and suggest three Supreme Court decisions. That would mean pasting names and private details into an outside service. The decisions would come with neat citations, and nothing on the screen would show whether they exist.

2. Why courts write their own rules

India's Supreme Court noted in July 2026 that AI's abilities, amid heavier workloads, are pushing professionals to use it [9, para 2]. Court rules address three concerns, all named in the guidance for judges in England and Wales. Anything typed into a public chatbot "should be seen as being published to all the world". AI may "make up fictitious cases, citations or quotes". Judges are also "personally responsible for material which is produced in their name" [1]. The rules differ on research. England and Wales does not recommend AI for legal research [1]. Gujarat allows it with checks against approved law journals [6, cl. 7].

IAPQB view. Each rule in this report makes the person who signs answer for AI output [1; 3; 6, cl. 11; 10, reg 8]. In the tribunal case (see 3.4), the bank said the invented precedents came from the tribunal's own research. They escaped the appeal tribunal's scrutiny and were pointed out by the losing side at the Supreme Court [9, paras 14, 16].

3. Four cases

3.1 England and Wales, 2023 to 2026: from guidance to approved tools

The judiciary first issued AI guidance in December 2023 and updated it in April and October 2025 [2]. The October 2025 version says: "Do not enter any information into a public AI chatbot that is not already in the public domain" [1]. A position statement of 9 October 2026 adds: "Only approved AI tools may be used for judicial purposes." The tools must keep data in the UK, have enterprise data protection and no web access, and they must not determine the outcome of a case. A steering group chaired by the Lead Judge for AI reviews new uses [2].

3.2 Kerala, 2025: no AI for findings or judgments

In July 2025 the Kerala High Court issued an AI policy for the judges and staff of the state's district courts, reported by three outlets as India's first [3, 4, 5]. AI tools may not be used to arrive at any findings, reliefs, order or judgment "under any circumstances", because responsibility lies fully with the judges [3, 4]. Cloud services such as ChatGPT and DeepSeek are to be avoided unless approved, to protect case facts, personal identifiers and privileged communications [3, 4, 5]. Judicial officers must check results from approved tools, including citations from case-law databases that use AI [3, 4].

3.3 Gujarat, 2026: a detailed list of banned uses

On 4 April 2026 the Gujarat High Court issued its own policy [5, 7]. It covers judicial officers, staff, legal assistants and interns, also on personal devices [6, cl. 5]. AI "shall never be employed" for decision-making, judicial reasoning, order drafting, judgment preparation or bail and sentencing considerations, and may not weigh evidence or summarise testimony [6, cl. 8]. It may help with legal research, the language of drafts, translation and scheduling, and a human officer must check each result [6, cl. 3, 7]. Public tools are limited to work that concerns no particular case. Wider use needs an approved enterprise tool whose contract bars training on inputs [6, cl. 9].

3.4 New Delhi, 2026: decisions built on invented precedents

On 27 February 2026 India's Supreme Court took note that an Andhra Pradesh trial court had relied on four judgments the High Court found to be AI-generated. It said a decision based on such judgments "would be a misconduct" [8].

On 2 July it set aside a ruling of the National Company Law Tribunal in Mumbai and the appeal ruling upholding it. Of six precedents the tribunal relied on, three did not exist, and three were quoted with paragraphs not found in the real reports [9, paras 15, 18]. The court called such a decision "no decision in the eyes of the law" and declared zero tolerance for the Bar and the Bench [9, para 7].

A month earlier, its AI Committee had published draft Regulations for Use of Artificial Intelligence in Courts, 2026. The draft bars judicial outcomes reached "solely on the basis of AI-generated information" [10, reg 20]. Comments were due by 15 July 2026 [11]. On 16 November 2026 the court's notices page listed no final version [12].

4. How practitioners handle it now

The judge decides, and the rules say so. The position statement says judicial decisions "will be taken by a judicial office holder" [2]. India's draft says AI must "function solely in an assistive capacity" [10, reg 4].

Tools are approved, with conditions. England and Wales requires approved tools that keep data in the UK and have no web access [2]. Gujarat allows wider use only for enterprise tools whose contracts bar training on inputs [6, cl. 9]. India's draft requires written approval, a register in each court and yearly audits [10, regs 19, 37, 38].

Case material stays out of public tools. England and Wales allows only public information into public chatbots [1]. Gujarat bars names of parties and witnesses, details of pending cases, privileged communications and evidence [6, cl. 8].

Citations are checked at the source. England and Wales requires AI output to be checked before it is relied on [1]. Gujarat requires a check against the full text of the judgment, in sources such as SCC Online [6, cl. 8, 10].

Staff are covered, and breaches have consequences. England and Wales covers clerks, judicial assistants, legal advisers and other support staff [1]. In Gujarat, assistants must tell the judge when they use AI, and any breach is misconduct [6, cl. 11, 12].

5. What to take from this

For judges, tribunal members and court staff. The person who signs answers for every citation, summary and translation that AI helped to produce [1; 6, cl. 11]. Gujarat adds that a correct format "shall not be treated as evidence" that a citation exists [6, cl. 10]. In the tribunal case, invented precedents passed two levels of decision [9, para 16].

For court leaders and administrators. England and Wales and Gujarat tie approved tools to written safeguards, and India's draft adds registers, audits and an incident database [2; 6, cl. 9; 10, regs 37 to 39]. In both Indian cases, the losing side raised the invented precedents after the decision [8; 9, para 14].

For judicial academies, law schools and training providers. The Judicial College in England and Wales trains judges on AI, and Gujarat's policy assigns AI training to its state judicial academy [2; 6, cl. 3]. Checking the full judgment would have shown the errors in both Indian cases [8; 9, para 15].

6. Limits of this report

The report covers only England and Wales and India. The Kerala policy could not be opened because the court's website did not respond. Only points found in at least two of three news reports are used [3, 4, 5]. England and Wales does not name its approved tools [2]. Even a final version of India's rules takes effect only when chief justices notify it [10, reg 1]. The draft lets an officer skip the check of AI output for reasons recorded in writing [10, reg 8], while the July judgment calls for zero tolerance of unchecked AI precedents [9, para 7]. A report on a third high court had no second source and is left out. No AI tool was tested.

References

[1] Courts and Tribunals Judiciary (31 October 2025). Artificial Intelligence (AI) Guidance for Judicial Office Holders. https://www.judiciary.uk/wp-content/uploads/2025/10/Artificial-Intelligence-AI-Guidance-for-Judicial-Office-Holders-2.pdf

[2] Courts and Tribunals Judiciary (9 October 2026). Position Statement on the Judicial use of Artificial Intelligence (AI). Signed by Sir Colin Birss, Lead Judge for Artificial Intelligence. https://www.judiciary.uk/guidance-and-resources/position-statement-on-the-judicial-use-of-artificial-intelligence-ai/

[3] Pratap, G. (19 July 2025). In a first, Kerala High Court issues policy for use of AI by judges; says no AI for judgments. Bar & Bench. https://www.barandbench.com/news/in-a-first-kerala-high-court-issues-policy-for-use-of-ai-by-judges-says-no-ai-for-judgments

[4] Ramdas, P. (20 July 2025). ‘No AI tools for judicial verdicts’: Kerala HC brings out landmark guidelines. The New Indian Express. https://www.newindianexpress.com/states/kerala/2025/Jul/20/no-ai-tools-for-judicial-verdicts-kerala-hc-brings-out-landmark-guidelines

[5] Bhalla, V. (5 April 2026). Gujarat High Court issues AI policy; bans use in decision-making, mandates human oversight. The Indian Express. https://indianexpress.com/article/legal-news/gujarat-high-court-ai-policy-judicial-decision-making-ban-human-oversight-10619719/

[6] High Court of Gujarat. Policy on the use of Artificial Intelligence in the Judicial and Court Administration. https://gujarathighcourt.nic.in/hccms/sites/default/files/miscnotifications/Policy%20on%20the%20use%20of%20Artificial%20Intelligence%20in%20the%20Judicial%20and%20Court%20Administration.pdf (accessed 16 November 2026; no date shown in the document)

[7] TNIE online desk, with inputs from PTI (4 April 2026). Gujarat HC announces new AI policy, prohibits its use in decision-making, judgment drafting. The New Indian Express. https://www.newindianexpress.com/india/2026/Apr/04/gujarat-hc-announces-new-ai-policy-prohibits-its-use-in-decision-making-judgement-drafting

[8] Gummadi Usha Rani & Anr. v. Sure Mallikarjuna Rao & Anr., Petition for Special Leave to Appeal (C) No. 7575/2026, Supreme Court of India, record of proceedings (order), 27 February 2026. Copy published by Bar & Bench. https://images.assettype.com/barandbench/2026-03-02/vyy7083e/GUMMADI_USHA_RANI___ANR_vs_SURE_MALLIKARJUNA_RAO___ANR_____.pdf

[9] Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr., Civil Appeal No. 11950 of 2025, 2026 INSC 668, Supreme Court of India, judgment, 2 July 2026. Copy published by Bar & Bench. https://images.assettype.com/barandbench/2026-07-02/npcfwf41/Pooja_Ramesh_Singh_vs__Jammu_and_Kashmir_Bank.pdf

[10] Supreme Court of India, Artificial Intelligence Committee (3 June 2026). Notice: Seeking views/suggestions of all stakeholders and the general public on draft ‘Regulations for Use of Artificial Intelligence (AI) in Courts, 2026’, with the draft Regulations for Use of Artificial Intelligence in Courts, 2026. https://cdnbbsr.s3waas.gov.in/s3ec0490f1f4972d133619a60c30f3559e/uploads/2026/06/2026060342.pdf

[11] Supreme Court of India, Artificial Intelligence Committee (16 June 2026). Notice: Extension of time for submission of views/comments by stakeholders and the general public on draft ‘Regulations for Use of Artificial Intelligence (AI) in Courts, 2026’. https://cdnbbsr.s3waas.gov.in/s3ec0490f1f4972d133619a60c30f3559e/uploads/2026/06/2026061691-1.pdf

[12] Supreme Court of India. Notices and Circulars [web page]. https://www.sci.gov.in/notices-and-circulars/ (accessed 16 November 2026)

How to cite

IAPQB (2026). Courts write rules for their own AI use: how courts in England and Wales and in India limit AI use by judges and court staff. IAPQB Report No. 29. International AI Practice Qualification Board. https://iapqb.org/research/029-courts-rules-for-ai.html

About this report

Prepared by IAPQB. Each fact comes from the sources listed, and each source was opened and checked on 16 November 2026; anything that could not be checked was left out. AI tools were used to search and to draft, and IAPQB checked every claim against its source and is responsible for the text. IAPQB runs an AI work qualification; this report does not promote it. Corrections: contact@iapqb.org or an issue at github.com/iapqb/iapqb.github.io. Licence: CC BY 4.0.

Archived with the series on Zenodo: doi.org/10.5281/zenodo.24248199 · Markdown sources: github.com/iapqb/research