CLA News / From Answer to Authority A Citation-First Discipline for AI-Assisted Legal Research By Shini Kumar & Frederick Antoine
AI can give a lawyer an answer in seconds, but it may not give that answer legal authority.
For a young lawyer, the distinction between an answer and a legal authority is not merely semantic; it is the difference between a useful lead and a proposition that can be advanced to a client, counsel or court. In a 2025 LexisNexis survey of 1,037 Australian legal practitioners, 45.6% identified legal research as the leading use case for generative AI, while 39.3% identified reliance on inaccurate or fabricated information (hallucinations) as their leading concern.[1] Although the data is jurisdiction-specific, it reveals a tension with wider Commonwealth relevance.
This risk is not confined to wholly invented facts or information. An AI response may cite a real authority but misconstrue the ratio, draw on case law that makes sense only in a different legal, factual, or procedural context. In any of these scenarios, there is an ever-present risk that the information generated is read as persuasive based on the familiar language used across common-law systems, accompanied by confident written style of the output.
The practical distinction between an AI-generated answer and a legal authority is therefore clear. An AI-generated response is a research lead. The citation provided in the response is a route to a source. Authority is obtained from a source that, after thorough validation, can support the proposition in the relevant forum, at the relevant time, on the relevant facts. Only when a source is read, and materially tested, does it become an authority. A lawyer needs to be able to open cited material and assess its treatment, rather than simply accept a confident output.[2]
While several countries across the Commonwealth offer guidance, this does not create a single, uniform rule for lawyers. It does, nevertheless, broadly point in a similar direction: generative AI may assist with legal work, but it does not displace or transfer responsibility for accuracy, relevance, currency or source verification.
Guidance from the Singapore Judiciary makes the point that lawyers remain professionally obligated to ensure that the materials they put before the Courts are independently verified, accurate, true and appropriate.[3] Ensuring the accuracy of court documents therefore includes verifying that any AI-generated references exist, that they support the legal propositions that are attributed to them, and that AI-generated content does not infringe on intellectual property rights. Where that AI-generated content includes extracts or quotations, lawyers must also confirm that these are accurate, faithfully reproduced, and attributed to the correct source.[4] The Guide additionally states that asking a generative AI tool for confirmation, or using one generative AI tool to confirm content generated from another, does not constitute sufficient verification.[5]
The Federal Court of Australia has been equally direct. In a practice note issued on 16 April 2026, the Court highlights that generative AI may produce fictitious authorities, misleading legal analysis, factual errors, and false confirmations of accuracy.[6] It emphasises that the presentation of false or inaccurate material to the Court is unacceptable, and inconsistent with the duties of parties not to mislead the Court or other parties.[7] In line with this, the Court clearly states that it expects responsible persons to ensure that legal authorities cited in submissions exist, and support the proposition stated.[8]
The Supreme Court of New South Wales has adopted a similar approach, stating that where generative AI has been used to prepare written submissions, summaries or skeleton arguments, the author must verify, in the body of these documents, that all citations, legal and academic authority, and case law and legislative references exist, are accurate, and are relevant to the proceedings.[9] The Court also requires similar verifications to be carried out for references to evidence in these documents, and states that any verifications must not be solely carried out by a generative AI tool or program.[10] The use of generative AI in preparing these documents does not qualify or relieve their authors of their professional and ethical obligations to the Court or to the administration of justice.[11]
The Malaysian Bar has cautioned that generative AI should not be used as a substitute for law books, legislation, case law, legal databases, secure work systems, or trained legal minds.[12] In a 2025 circular, the Bar warns against using such tools to verify the existence or accuracy of case law, legislation or legal rules, and advises that AI-generated legal advice or final legal documents be independently validated and subject to manual legal review.[13] It advises practitioners to treat AI output as unverified third-party commentary rather than as a primary legal source, and to verify it against official legal databases.[14]
Similar messages are carried in the guidance across other regions. The Courts and Tribunals Judiciary in England and Wales cautions judicial office holders that public AI chatbots do not provide responses from authoritative databases, making them a poor means of conducting research to find new information you cannot verify.[i] While the quality of response received may depend on how an individual engages with the relevant AI tool, even with the best prompts, information provided may be inaccurate, incomplete, misleading or biased. The New Zealand Law Society emphasises that all lawyers are ultimately responsible for the legal services they provide, and that a lawyer is not absolved of their responsibility for legal advice, or defects in an end-product, merely because it is derived from generative AI.[15] In light of this, careful human oversight of the use of generative AI is vital to ensure that it is used ethically and responsibly, including fact-checking outputs for accuracy and relevance, and ensuring that staff are using Gen AI only as authorised, and with proper disclosure where used.[ii] The Federal Court of Canada also highlights the importance of retaining a human in the loop to ensure accuracy and trustworthiness, noting that this manner of verification is aligned with the standards generally required in the legal profession.[16]
These are not merely warnings about ‘hallucinations’. They identify the discipline that should govern AI-assisted research: start from authority, not plausibility.
For young lawyers, the question is not whether AI should be used, but whether they can demonstrate the authority behind every material proposition it has helped them find. Generative AI can help frame issues, identify possible sources, and improve efficiency. It cannot decide whether an authority is current, applicable, properly understood or sufficient for the proposition advanced. That remains a matter of legal judgement.
A citation first discipline requires young lawyers to move away from an AI-generated answer to the underlying source, reading it in context, and explaining why it supports the advice, submission or draft. Lawyers should therefore be discerning, not only about the answers AI produces, but also the tools they choose to use. AI-assisted research is most useful where its outputs are grounded in trusted and authoritative content that enables lawyers to trace, test and defend every material proposition. Those features support, rather than replace, the lawyer’s independent duty to verify the accuracy, relevance, currency and application of each authority and source. Used in this way, AI can strengthen legal research, without weakening the verification, accountability, and professional care on which clients, courts and the rule of law depend.
Shini Kumar & Frederick Antoine
LexisNexis India & Southeast Asia
[1] LexisNexis, Legal AI and the Future of the Legal Profession: LexisNexis 2026 Australia AI Sentiment Survey Findings (2026) 8, 13 https://www.lexisnexis.com/assets/en-au/pdf/lexisnexis-2026-australia-ai-sentiment-survey-legal-ai-future-of-legal-profession.pdf accessed 1 September 2026.
[2] ibid paras 5(1)(b) and 5(3)(c).
[3] ibid para 5(3)(e).
[4] Federal Court of Australia, Use of Generative Artificial Intelligence Practice Note (GPN-AI) (16 April 2026) para 4.3 https://www.fedcourt.gov.au/law-and-practice/practice-documents/practice-notes/gpn-ai accessed 2 September 2026.
[5] ibid para 4.4.
[6] ibid para 4.5.
[7] Supreme Court of New South Wales, Supreme Court Practice Note SC Gen 23: Use of Generative Artificial Intelligence (Gen AI) (28 January 2025) para 16 https://supremecourt.nsw.gov.au/documents/Practice-and-Procedure/Practice-Notes/general/current/PN_SC_Gen_23.pdf accessed 2 September 2026.
[8] ibid paras 16–17.
[9] ibid para 18.
[10] Malaysian Bar, Circular No 242/2025: Updates on the Use of Generative Artificial Intelligence in Legal Practice (3 July 2025) para 2 https://www.malaysianbar.org.my/cms/upload_files/document/Circular%20No%20242-2025.pdf accessed 1 September 2026.
[11] ibid para 4.
[12] ibid paras 6–8.
[13] Judiciary of England and Wales, Artificial Intelligence (AI) Guidance for Judicial Office Holders (31 October 2025) s 3(I) https://www.judiciary.uk/wp-content/uploads/2025/10/Artificial-Intelligence-AI-Guidance-for-Judicial-Office-Holders-2.pdf accessed 1 September 2026.
[14] New Zealand Law Society Te Kāhui Ture o Aotearoa, Lawyers and Generative AI (March 2024) 3 https://www.lawsociety.org.nz/assets/Professional-practice-docs/Rules-and-Guidelines/Lawyers-and-AI-Guidance-Mar-2024.pdf accessed 2 September 2026.
[15] ibid 4.
[16] Federal Court of Canada, ‘Notice to the Parties and the Profession: The Use of Artificial Intelligence in Court Proceedings’ (7 May 2024) 2 (heading ‘Human in the loop’) https://www.fct-cf.ca/Content/assets/pdf/base/FC-Updated-AI-Notice-EN.pdf accessed 2 September 2026.
