![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
The surge in the utilization of generative artificial intelligence within the legal profession, particularly in preparing legal documents, has prompted a flurry of responses from the judiciary, including practice directives and standing orders. These varied responses have either directly or implicitly referred to litigants using “hallucinated” cases generated by large language models (LLMs) as a major concern. This article briefly evaluates these responses, making conclusions on their propriety and relevance.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 May be used interchangeably with generative AI in this article.
2 “What are Large Language Models?”, IBM, https://www.ibm.com/topics/large-language-models.
3 Ibid.
4 Ibid.
5 See “Generative AI & the Legal Profession,” April 2023, LexisNexis, https://www.lexisnexis.com/pdf/ln_generative_ai_report.pdf. See also IIona Logvinova, “Its not automagical: Adopting generative AI and legal tech takes strategy, intuition and systems thinking,” August 30, 2023, McKinsey, https://www.mckinsey.com/featured-insights/in-the-balance/breaking-barriers-unleashing-the-power-of-collaboration-in-a-global-legal-department .and. See International Legal Generative AI report: Detailed survey findings, August 22, 2023, Lexis Nexis, https://www.lexisnexis.com/pdf/lexisplus/international-legal-generative-ai-report.pdf
6 See “Generative AI & the Legal Profession,” April 2023, LexisNexis, https://www.lexisnexis.com/pdf/ln_generative_ai_report.pdf.
7 The term “hallucination” may have different meanings depending on the context in which it is used (See Negar Maleki, Balaji Padmanabhan and Kaushik Dutta, “Ai Hallucinations: A Misnomer Worth Clarifying,” January 9, 2024, online, https://arxiv.org/pdf/2401.06796.pdf.
9 Ibid.
10 Ibid. For an expert summary of this problem in the context of the legal profession, See Maura R. Grossman, Paul W. Grimm and Daniel G. Brown, “Is disclosure and certification of the use of generative AI really necessary,” online, https://scholarship.law.duke.edu/cgi/viewcontent.cgi?article=6969&context=faculty_scholarship, 72–74.
11 See Matthew Dahl, Varun Magesh, Mirac Suzgun and Daniel Ho, “Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models,” online, https://arxiv.org/pdf/2401.01301.pdf. For more general surveys and findings on LLM hallucinations, see Ji et al, “Survey of Hallucination in Natural Language Generation,” online, https://arxiv.org/pdf/2202.03629.pdf.; Shawn Curran, Oliver Bethell and Sam Lansley, “Hallucination is the last thing you need”, online,. https://arxiv.org/ftp/arxiv/papers/2306/2306.11520.pdf; Denis Peskoff and Brandon M. Stewart, “Credible without Credit: Domain Experts Assess Generative Language Models”, https://aclanthology.org/2023.acl-short.37.pdf; Joy Buchanan and Olga Shapoval, GPT-3.5 Hallucinates Non-Existent Citations: Evidence from Economics, June 2023, C:\Users\fogunde\Downloads\SSRN-id4467968.pdf;
12 See for example “Terms of use”, November 14, 2023, OpenAI, https://openai.com/policies/terms-of-use (OpenAI).
13 (2023) 22-CV-1461(PKC).
14 Ibid.
15 Ibid.
16 Ibid.
17 Ibid.
18 Ibid.
19 See Clara Geoghegan, Colorado Lawyer Cited Fake Cases in Motion Written with ChatGPT, June 21, 2023, LawweekColorado, https://www.lawweekcolorado.com/article/colorado-lawyer-cited-fake-cases-in-motion-written-with-chatgpt/.
20 The lawyer was subsequently suspended for ninety days (See People v Zachariah C. Crabill 23PDJ067, November 22, 2023, https://www.documentcloud.org/documents/24036102-michelnewtrialmotion101623.
21 .United States of America v Prakazrel Michel (2023) Case No. 19-148-1,. https://www.documentcloud.org/documents/24036102-michelnewtrialmotion101623.
22 United States District Court 18-CR-602, UNITED STATES DISTRICT COURT (documentcloud.org).
23 See Morgan v Community Against Violence (2023) 23-CV-353-WPJ/JMR (D.N.M October 23, 2023). The Court in this case did not state whether the fake citations were AI-generated or not.
24 (2023) United States Court of Appeals for the Second Circuit 22-2057, available at https://ww3.ca2.uscourts.gov/decisions/isysquery/134fd2c7-d7fe-4e70-9be1-9fef8c687742/1/doc/22-2057_opn.pdf#xml=https://ww3.ca2.uscourts.gov/decisions/isysquery/134fd2c7-d7fe-4e70-9be1-9fef8c687742/1/hilite/.
25 Ibid.
26 EDI 11172 (Mo.Ct. App. Feb 13, 2024).
27 Chris Dolmetsch, “Lawyers use ChatGPT to add up fees, Judge faults their math,” February 22, 2024, Bloomberg News, https://news.bloomberglaw.com/business-and-practice/lawyers-use-chatgpt-to-add-up-fees-judge-faults-their-math
28 Ibid.
29 (2024) BCSC 285.
30 Ibid.
31 Ibid, p.45.
32 See Chloe Xiang, “Judge Bans AI-Generated Filings in Court because it just makes stuff up”, June 1, 2023, VICE, https://www.vice.com/en/article/qjvk7w/judge-bans-ai-generated-filings-in-court-because-it-just-makes-stuff-up.
33 See “Disclosure and Certification Requirements- Generative Artificial Intelligence- Chambers of United States District Judge Leslie E. Kobayashi”, online, https://www.hid.uscourts.gov/cms/assets/95f11dcf-7411-42d2-9ac2-92b2424519f6/AIGuidelinesLEK.pdf.
34 “Disclosure and Certification Requirements- Generative Artificial Intelligence- Chambers of United States District Judge Scott L. Palk”, online https://www.okwd.uscourts.gov/wp-content/uploads/AI_Guidelines_JudgePalk.pdf.
35 See “Standing Order for Civil Cases before Magistrate Judge Fuentes,” online, https://www.ilnd.uscourts.gov/_assets/_documents/_forms/_judges/Fuentes/Standing%20Order%20For%20Civil%20Cases%20Before%20Judge%20Fuentes%20rev%27d%205-31-23%20(002).pdf.
36 See “Order on Artificial Intelligence, United States Court of International Trade,” online, https://www.cit.uscourts.gov/sites/cit/files/Order%20on%20Artificial%20Intelligence.pdf.
37 Ibid.
38 Sara Merken, More judges, lawyers confront pitfalls of artificial intelligence, June 16, 2023, Reuters, https://www.reuters.com/legal/transactional/more-judges-lawyers-confront-pitfalls-artificial-intelligence-2023-06-16/.
39 See United States Bankruptcy Court, Northern District of Texas, July 21, 2023, online, https://www.txnb.uscourts.gov/sites/txnb/files/news/General%20Order%202023-03%20Pleadings%20Using%20Generative%20Artificial%20Intelligence-signed.pdf. Similar order was issued on June 9, 2023 by another Judicial District Court Judge in Texas. See “Standing Order Regarding use of Artificial Intelligence, 394th Judicial District Court,” June 9, 2023, https://img1.wsimg.com/blobby/go/2f8cb9d7-adb6-4232-a36b-27b72fdfcd38/downloads/Standing%20order%20Regarding%20Use%20of%20Artificial%20Int.pdf?ver=1690737636792.
40 See “Individual Practices in Civil Cases,” Southern District of New York, https://www.nysd.uscourts.gov/sites/default/files/practice_documents/AS%20Subramanian%20Civil%20Individual%20Practices.pdf, p.7.
41 Ibid
42 CV 23-69-M-DWM (D. Mont. Jun 22, 2023).
43 See United States District Court: Eastern District of Missouri, https://www.moed.uscourts.gov/self-represented-litigants-srl.
45 See United States Court of Appeals for the fifth Circuit, Notice of Proposed Amendment to 5th CIR. R.32.3, https://www.ca5.uscourts.gov/docs/default-source/default-document-library/public-comment-local-rule-32-3-and-form-6.
47 See Judge Iain D. Johnston, United States District Court, Northern District of Illinois, https://www.ilnd.uscourts.gov/judge-info.aspx?IuUaWzNcEoPWNpdOx+5lSeRQvpEAF5l/.
48 See “Practice Direction, Re: Use of Artificial Intelligence in Court Submissions,” June 23, 2023, https://www.manitobacourts.mb.ca/site/assets/files/2045/practice_direction_-_use_of_artificial_intelligence_in_court_submissions.pdf. Manitoba’s Chief Justice cited as reasons for his directive the “legitimate concerns” about the reliability and accuracy of the information generated by AI programs and the need for ongoing discussions about its responsible use in case. (See Zena Olijynk, “Canadian courts turning an eye to how artificial intelligence is used in the legal system”, July 11, 2023, Canadian Lawyer, https://www.canadianlawyermag.com/resources/professional-regulation/canadian-courts-turning-an-eye-to-how-artificial-intelligence-is-used-in-the-legal-system/377757.
49 Ibid.
50 See Practice Direction, Use of Artificial Intelligence Tools, June 26, 2023, https://www.yukoncourts.ca/sites/default/files/2023-06/GENERAL-29%20Use%20of%20AI.pdf.
51 See “Notice to the Profession & Public- Ensuring the Integrity of court submissions when using large language models,” October 6, 2023, online, https://albertacourts.ca/kb/resources/announcements/notice-to-the-profession-public---use-of-ai-in-citations-submissions.
52 See “Ensuring the Integrity of Court Submissions when using generative Artificial Intelligence, Supreme Court of Nova Scotia,” October 18, 2023, https://courts.ns.ca/sites/default/files/notices/Oct%202023/NSSC_Court_Submissions_AI_Oct_18_2023.pdf.
53 See “Notice to Profession and Public: Integrity of Court Submissions When Using Large Language Models, Superior Court of Quebec,” October 24, 2023, https://coursuperieureduquebec.ca/fileadmin/cour-superieure/Communiques_and_Directives/Montreal/Avis_a_la_Communite_juridique-Utilisation_intelligence_artificielle_EN.pdf?trk=public_post_comment-text.
54 See “Use of Artificial Intelligence (AI) and Protecting the Integrity of Court Submissions in Provincial Court”, October 27, 2023, Nova Scotia Provincial Court, https://www.courts.ns.ca/sites/default/files/notices/Oct2023/NSPC_Artificial_Intelligence_Oct_27_2023.pdf.
55 See “Notice to the Parties and the Profession: The Use of Artificial Intelligence in Court Proceedings, Federal Court,” December 20, 2023, 2023-12-20-notice-use-of-ai-in-court-proceedings.pdf (fct-cf.gc.ca). The Federal Court on the same date also issued interim Principles and Guidelines on its use of Artificial Intelligence. See “Interim Principles and Guidelines on the Court’s Use of Artificial Intelligence,” December 20, 2023, https://www.fct-cf.gc.ca/en/pages/law-and-practice/artificial-intelligence.
56 See Cristin Schmitz, SCC considers possible direction on use of AI in top court as more trial courts weigh in, July 7, 2023, Law 360, https://www.law360.ca/articles/48377/scc-considers-possible-practice-direction-on-use-of-ai-in-top-court-as-more-trial-courts-weigh-in?article_related_content=1. A similar incident occurred.
57 This is the description expressed by Grossman et al in Maura R. Grossman, Paul W. Grimm and Daniel G. Brown, “Is disclosure and certification of the use of generative AI really necessary, online, https://scholarship.law.duke.edu/cgi/viewcontent.cgi?article=6969&context=faculty_scholarship.
58 See Adam Allen Bent, “Large Language Models: AI’s Legal Revolution” (2023) 44:1 Pace Law Review 91, 132.
59 Some authors have created a “hallucination vulnerability index”, a comparative evaluation of LLMs based on their vulnerability to producing hallucinations (See https://arxiv.org/pdf/2310.04988.pdf?trk=public_post_comment-text).
60 For a general survey of legal LLMs, see Rawte et al, “The troubling Emergence of Hallucination in Large Language Models- An Extensive Definition, Quantification and Prescriptive Remediations”, online, https://arxiv.org/pdf/2312.03718.pdf.
61 See Adam Allen Bent, “Large Language Models: AI’s Legal Revolution” (2023) 44:1 Pace Law Review 91, 132 Large Language Models: AI's Legal Revolution (pace.edu).
62 Andrew Perlman, “The Legal Ethics of Generative AI”, online, SSRN-id4735389.pdf, p.11.
63 See Serena Wellen, “How Lexis+AI delivers Hallucination-Free Linked Legal Citations”, April 30, 2024, LexisNexis, https://www.lexisnexis.com/community/insights/legal/b/product-features/posts/how-lexis-ai-delivers-hallucination-free-linked-legal-citations.
64 Ibid.
65 Ibid. Users are nevertheless warned to review a citation for accuracy where the citation provided by Lexis+AI does not include a link. Some experts have also warned that RAG, while undeniably useful, cannot stop LLMs from hallucinating (See Kyle Wiggers, “Why RAG won’t solve generative AI’s hallucination problem”, TechCruch, May 7, 2024, https://techcrunch.com/2024/05/04/why-rag-wont-solve-generative-ais-hallucination-problem/.
66 See “LexisNexis Launches Second-Generation Legal AI Assistant on Lexi+AI”, April 23, 2024, LexisNexis, https://www.lexisnexis.com/community/pressroom/b/news/posts/lexisnexis-launches-second-generation-legal-ai-assistant-on-lexis-ai.
67 See Bob Ambrogi, “True to Its Name, LexisNexis Unveils New AI Capabilities for Both its Lexis Legal Service and Its Nexis News Service”, LawSites, April 24, 2024, https://www.lawnext.com/2024/04/true-to-its-name-lexisnexis-unveils-new-ai-capabilities-for-both-its-lexis-legal-service-and-its-nexis-news-service.html.
68 See “Thomson Reuters Launches Generative AI-Powered Solutions to Transform How Legal Professionals Work”, November 15, 2023, Thomson Reuters, https://www.thomsonreuters.com/en/press-releases/2023/november/thomson-reuters-launches-generative-ai-powered-solutions-to-transform-how-legal-professionals-work.html
69 See “GPT-4 alone is not a reliable legal solution-but it does enable one”, May 5, 2023, Casetext, https://casetext.com/blog/cocounsel-harnesses-gpt-4s-power-to-deliver-results-that-legal-professionals-can-rely-on/.
70 Ibid.
71 A system where human evaluators rank factual correctness of LLM generated responses.
72 See Tom Martin, “Hallucinations: What are They, Why Do They Happen, How to Fix Them?”, LawDroid Manifesto, May 8, 2024, https://bootstrap.rbi.skyhigh.cloud/clientless/#url=https://www.lawdroidmanifesto.com/p/hallucinations-what-are-they-why. See also Ziwei Xu, Sanjay Jain and Mohan Kankanhalli, “Hallucination is Inevitable” An Innate Limitation of Large Language Models”, January 22, 2024, https://arxiv.org/pdf/2401.11817, 12-13; Megan Ma, Aparna Sinha, Ankit Tandon and Jennifer Richards, “Generative AI Legal Landscape 2024”, March 2024, Stanford Law School, https://law.stanford.edu/publications/generative-ai-legal-landscape-2024-2/.
73 See Harry Surden, “ChatGPT, AI Large Language Models, and Law”, (2024) 92 Fordham Law Review 1942, 1969–1972.
74 See for example American Bar Association Model Rules of Professional Conduct, Rule 1.1 (Comment 8), Rule 3.3 and Rule 1.6, Federation of Law Societies of Canada Model Code of Professional Conduct, https://flsc-s3-storage-pub.s3.ca-central-1.amazonaws.com/Model Code Oct 2022.pdf, Rule 3.1-2 (Comment 4A and 4B).
75 Ibid.
76 Maura R. Grossman, Paul W. Grimm and Daniel G. Brown, “Is disclosure and certification of the use of generative AI really necessary”, online, https://flsc-s3-storage-pub.s3.ca-central-1.amazonaws.com/Model Code Oct 2022.pdf.
77 David Marshall alludes to this point as well in observing that courts may themselves not be in the best position to understand and assess the risks and uses of Generative AI by the legal profession (See David Marshall, Court Directives on Generative AI: Recent Attempts at Balancing Concerns with Ethical Duties and Effective Representation, July 18, 2023, https://go-gale-com.legislative.idm.oclc.org/ps/retrieve.do?tabID=T004&resultListType=RESULT_LIST&searchResultsType=SingleTab&retrievalId=139cb60b-8e02-4620-9b24-a5559073bae8&hitCount=967&searchType=BasicSearchForm¤tPosition=331&docId=GALE%7CA757603031&docType=Article&sort=Relevance&contentSegment=ZGPN&prodId=ITOF&pageNum=17&contentSet=GALE%7CA757603031&searchId=R8&userGroupName=saskatch_legis&inPS=true.
78 See “Practice Direction, Re: Use of Artificial Intelligence in Court Submissions,” June 23, 2023, https://www.manitobacourts.mb.ca/site/assets/files/2045/practice_direction_-_use_of_artificial_intelligence_in_court_submissions.pdf .(Manitoba),. See “Use of Artificial Intelligence (AI) and Protecting the Integrity of Court Submissions in Provincial Court”, October 27, 2023, Nova Scotia Provincial Court, https://www.courts.ns.ca/sites/default/files/notices/Oct 2023/NSPC_Artificial_Intelligence_Oct_27_2023.pdf. (Nova Scoitia). These directives were clearly intended to respond to LLMs based on their timing and the fact that AI in general has been a major part of law practice for well over a decade.
79 See for example American Bar Association Model Rules of Professional Conduct, Rule 1.1 (Comment 8), Rule 3.3 and Rule 1.6, Federation of Law Societies of Canada Model Code of Professional Conduct, https://flsc-s3-storage-pub.s3.ca-central-1.amazonaws.com/Model Code Oct 2022.pdf, Rule 3.1-2 (Comment 4A and 4B). See also Federal Rules of Civil Procedure, Rule 11. https://www.law.cornell.edu/rules/frcp/rule_11, Rule 11, Rule 26(g).
80 (2023) United States Court of Appeals for the Second Circuit 22-2057, available at https://ww3.ca2.uscourts.gov/decisions/isysquery/134fd2c7-d7fe-4e70-9be1-9fef8c687742/1/doc/22-2057_opn.pdf#xml=https://ww3.ca2.uscourts.gov/decisions/isysquery/134fd2c7-d7fe-4e70-9be1-9fef8c687742/1/hilite/.
81 Ibid. Although beyond the scope of this paper, some scholars have argued that existing Rules of Court are no match for the complexities created by LLM usage, they still consider the judicial response to have negative effect (See See Jessica Gunder, “Rule 11 is no match for generative AI” (2024) 27 Stanford Technology Law Review, C:\Users\fogunde\Downloads\SSRN-id4769448.pdf.
82 Andrew Perlman, “The Legal Ethics of Generative AI”, online, C:\Users\fogunde\Downloads\SSRN-id4735389.pdf.
83 In Thomson Reuters 2024 State of the Courts Report, only 15% of respondents believed generative AI could be used in court settings and 9% believed it should be used in a court setting. Nearly a third believed it should not be used in a court setting. (See “State of the Courts Report 2024,” Thomson Reuters, https://legal.thomsonreuters.com/content/dam/ewp-m/documents/legal/en/pdf/reports/state-of-the-courts-report-2024.pdf?form=thankyou&gatedContent=%252Fcontent%252Fewp-marketing-websites%252Flegal%252Fgl%252Fen%252Finsights%252Freports%252F2024-state-of-the-courts-report.
84 Inaccuracy of information was cited as the greatest concern for majority of respondents when asked about the use of generative AI in court settings (See State of the Courts Report 2024, Thomson Reuters, https://legal.thomsonreuters.com/content/dam/ewp-m/documents/legal/en/pdf/reports/state-of-the-courts-report-2024.pdf?form=thankyou&gatedContent=%252Fcontent%252Fewp-marketing-websites%252Flegal%252Fgl%252Fen%252Finsights%252Freports%252F2024-state-of-the-courts-report.
85 See Joe Patrice, “73 Percent of Lawyers Expect to be Using Generative AI Next Year … But to do what exactly”, November 8, 2023, Above the Law, https://abovethelaw.com/2023/11/73-percent-of-lawyers-expect-to-be-using-generative-ai-next-year-but-to-do-what-exactly/.
86 See for example Chief Justice Robert’s comments relating to generative AI in 2023 Year-End Report on the Federal Judiciary, December 31, 2023, online, https://www.supremecourt.gov/publicinfo/year-end/2023year-endreport.pdf and Shweta Watwe, “Judges Reflect on GenAI Use One Year after ChatGPT’s Debut,” November 28, 2023, Bloomberg Law, https://news.bloomberglaw.com/litigation/judges-reflect-on-genai-use-one-year-after-chatgpts-debut. In the Canadian context, the Chief Justice of the Yukon Supreme Court has affirmed the relevance of AI in increasing time and cost efficiency and stating that the Yukon Directive was not intended to discourage use of AI in legitimate ways (See Zena Olijynk, “Canadian courts turning an eye to how artificial intelligence is used in the legal system”, July 11, 2023, Canadian Lawyer, https://www.canadianlawyermag.com/resources/professional-regulation/canadian-courts-turning-an-eye-to-how-artificial-intelligence-is-used-in-the-legal-system/377757 .
87 For a more general discussion on the relationship between risk misperceptions and emerging technology, see Steve Calandrillo & Nolan Kobuke Anderson, “Terrified by Technology: How Systemic Bias Distorts U.S. Legal and Regulatory Responses to Emerging Technology” [2022] 2022:2 U Ill L Rev 597.
88 See Jessica Gunder, “Rule 11 is no match for generative AI” (2024) 27 Stanford Technology Law Review, C:\Users\fogunde\Downloads\SSRN-id4769448.pdf
89 OpenAI’s AI classifier, launched in January 2023, was discontinued in July 2023 due to low accuracy. See more generally; Ruixiang Tnag, Yu-Neng Chuang and Xia Hu, “The Science of Detecting LLM-Generated Texts”, online, https://arxiv.org/pdf/2303.07205.pdf; Chua Chin Hon; “AI vs AI”: Detecting LLM-Generated Text”, June 21, 2023, Medium, https://chuachinhon.medium.com/ai-vs-ai-detecting-llm-generated-text-eec92ea63e42; Wang et al, “LLM-Detector: Improving AI-Generated Chinese Text Detection with Open-Source LLM Instruction Tuning”, February 2, 2024, online, LLM-Detector: Improving AI-Generated Chinese Text Detection with Open-Source LLM Instruction Tuning (arxiv.org); 2307.07411.pdf; 2301.11305v1.pdf (arxiv.org), https://arxiv.org/html/2402.01158v1
90 “Judges Reflect on GenAI Use One Year after ChatGPT’s Debut,” November 28, 2023, Bloomberg Law, https://news.bloomberglaw.com/litigation/judges-reflect-on-genai-use-one-year-after-chatgpts-debut.
91 While the ethical or legal responsibility for verifying the accuracy or reliability of information rests on the party seeking to rely on such information, lawyers have an ethical duty to provide thorough service to their clients (See Model Code of Professional Conduct, Federation of Law Societies of Canada, https://flsc.ca/wp-content/uploads/2024/05/2024-Model-Code-of-Professional-Conduct.pdf, Rule 3.2-1, American Bar Association Model Rules of Judicial Conduct, American Bar Association, https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/model_rules_of_professional_conduct_table_of_contents/ (Rule 1.3). Judges are expected to perform judicial duties in a skillful, careful, attentive and timely way (See Ethical Principles for Judges, Canadian Judicial Council, https://cjc-ccm.ca/sites/default/files/documents/2021/CJC_20-301_Ethical-Principles_Bilingual_Final.pdf, Chapter 3, American Bar Association Code of Judicial Conduct, American Bar Association, https://www.americanbar.org/groups/professional_responsibility/publications/model_code_of_judicial_conduct/, Canon 2). It is far-reaching to conclude that this imposes a duty on judges or opposing counsel to verify information filed by another party. It does however suggest that it is prudent to review such documents thoroughly in determining the next course of action or in the case of judges, issuing a judgment or ruling.
92 See Stephanie Wilkins, “ChatGPT faces a timeout as legal turns its attention to more mature, legal-specific AI tools”, May 25, 2023, Law.com, https://www.law.com/2023/05/25/chatgpt-faces-a-timeout-as-legal-turns-its-attention-to-more-mature-legal-specific-ai-tools/?slreturn=20231001155143; Cristina Criddle, “Law firms embrace the efficiencies of artificial intelligence”, May 4, 2023, https://www.ft.com/content/9b1b1c5d-f382-484f-961a-b45ae0526675.
93 See Augunstein et al, “Factuality Challenges in the Era of Large Language Models”, October 10, 2023, online, https://arxiv.org/pdf/2310.05189.pdf#page=11&zoom=100,75,685; Tonmoy et al, “A comprehensive survey of Hallucination Mitigation Techniques in Large Language Models”, January 8, 2024, online, https://arxiv.org/pdf/2401.01313.pdf.
94 See Kim Martineau, “What is Retrieval-Augmented Generation”, August 22, 2023, IBM Research, https://research.ibm.com/blog/retrieval-augmented-generation-RAG.
95 LexisNexis launches Lexis+AI, a generative AI solution with linked Hallucination-free legal citations, October 25, 2023, https://www.lexisnexis.com/community/pressroom/b/news/posts/lexisnexis-launches-lexis-ai-a-generative-ai-solution-with-hallucination-free-linked-legal-citations.;. “Rhys Dipshan, LexisNexis Announces Generative AI platform Lexis+ AI, automating search, drafting and summary Tasks,” May 4, 2023, Law.com, https://www.law.com/legaltechnews/2023/05/04/lexisnexis-announces-generative-ai-platform-lexis-ai-automating-search-drafting-and-summary-tasks/?slreturn=20240002112313.
96 For an overviews on the evolution of generative AI between November 2022-date, see “A year in review: How AI transformed the legal profession in 2023,” January 8, 2024, Thomson Reuters, https://legal.thomsonreuters.com/blog/how-ai-transformed-the-legal-profession-in-2023/;
97 For studies on LLM hallucination mitigation methods, see Ji et al, “Towards mitigating Hallucination in Large Language Models via Self-Reflection,” online, https://openreview.net/pdf?id=up8EYzyrKV; Wei et al, “Measuring and Reducing LLM Hallucination without Gold-Standard Answers via Expertise-Weighing,” February 16, 2024, online, https://arxiv.org/html/2402.10412v1; Tonmoy et al, “A comprehensive survey of Hallucination Mitigation Techniques in Large Language Models”, January 8, 2024, online, https://arxiv.org/pdf/2401.01313.pdf; Dhuliawala et al, Chain of Verification reduces Hallucination in Large Language Models, September 25, 2023, online, https://arxiv.org/pdf/2309.11495.pdf?trk=public_post_comment-text.
98 Some express skepticism in this respect, arguing that hallucination will not decrease with better versions of LLMs because new versions will manifest “irresistible and enhanced language” (See Paul D. Callister, “Generative AI and Finding the Law”, (2023), online, https://irlaw.umkc.edu/cgi/viewcontent.cgi?article=1902&context=faculty_works, p.48. Others consider this as a concern mostly in more generic LLMs like ChatGPT (See .Shawn Curran, Oliver Bethell and Sam Lansley, “Hallucination is the last thing you need”, online,. https://arxiv.org/ftp/arxiv/papers/2306/2306.11520.pdf; Samuel Dahan, Rohan Bhambhoria, David Liang and Xiaodan Zhu, “Lawyers should not trust AI: A call for an Open-Source Legal Language Model,” October 27, 2023, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4587092.
99 (2024) File No. 2016-2501/A&B, available at https://caselaw.findlaw.com/court/sur-s-crt-new-yor-kin-cou/115735333.html.
100 Experts in this area adopt a similar approach. See generally; Michael Murray, “Artificial Intelligence and the Practice of Law Part 2: Working with your new AI Staff Attorney,” January 2023, https://www.researchgate.net/profile/Michael-Murray-8/publication/371842830_Artificial_Intelligence_and_the_Practice_of_Law_Part_2_Working_With_Your_New_AI_Staff_Attorney/links/64d14b3fd394182ab3b224e3/Artificial-Intelligence-and-the-Practice-of-Law-Part-2-Working-With-Your-New-AI-Staff-Attorney.pdf; Michael Murray, Artificial Intelligence and the Practice of Law Part 1: Lawyers must be Professional and Responsible Supervisors of AI, July 19, 2023, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4478588...
101 See Harry Surden, “ChatGPT, AI Large Language Models, and Law”, (2024) 92 Fordham Law Review 1942, 1969. See also “Techniques for writing effective legal AI prompts”, January 26, 2024, CaseText, https://casetext.com/blog/writing-effective-legal-ai-prompts/.
102 See Florida Bar Ethics Opinion 24-1, January 19, 2024, The Florida Bar, https://www.floridabar.org/etopinions/opinion-24-1/.
103 See “Practice Resource: Guidance on Professional Responsibility and Generative AI, Law Society of British Columbia,” online, https://www.lawsociety.bc.ca/Website/media/Shared/docs/practice/resources/Professional-responsibility-and-AI.pdf (British Columbia); Guidelines for the Use of Generative Artificial Intelligence in the Practice of Law, February 2024, Law Society of Saskatchewan, https://www.lawsociety.sk.ca/wp-content/uploads/Law-Society-of-Saskatchewan-Generative-Artificial-Intelligence-Guidelines.pdf (Saskatchewan); “The Generative AI Playbook: How Lawyers can Safely Take Advantage of the Opportunities Offered by Generative AI, Law Society of Alberta,” online, https://www.lawsociety.ab.ca/resource-centre/key-resources/professional-conduct/the-generative-ai-playbook/ (Alberta); “Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law”, State Bar of California, https://www.calbar.ca.gov/Portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf (California).
104 See “Artificial Intelligence: Guidance for Judicial Office Holders, Courts and Tribunals Judiciary,” December 12, 2023, https://www.judiciary.uk/wp-content/uploads/2023/12/AI-Judicial-Guidance.pdf.
105 Bob Ambrogi, a legal technology commentator makes a similar point in suggesting that any rulemaking surrounding generative AI use in law practice should be done in a “uniform and deliberative manner”. (See David Lat, “AI use in Law Practice Needs Common Sense, Not more Court Rules”, February 28, 204, Bloomberg Law, https://news.bloomberglaw.com/us-law-week/ai-use-in-law-practice-needs-common-sense-not-more-court-rules Grossman et al allude to this point in arguing against the necessity of individualized judicial standing orders. In their view, guidance and education being provided by bar associations and law societies in LLM usage takes away the burden from judges (See Maura R. Grossman, Paul W. Grimm and Daniel G. Brown, “Is disclosure and certification of the use of generative AI really necessary,” online, https://scholarship.law.duke.edu/cgi/viewcontent.cgi?article=6969&context=faculty_scholarship.
106 Maura R. Grossman, Paul W. Grimm and Daniel G. Brown, “Is disclosure and certification of the use of generative AI really necessary,” online, https://scholarship.law.duke.edu/cgi/viewcontent.cgi?article=6969&context=faculty_scholarship.
107 Andrew Perlman, “The Legal Ethics of Generative AI”, online, C:\Users\fogunde\Downloads\SSRN-id4735389.pdf
108 This is also gradually being acknowledged by courts. For example in the Matter of Samuel, the Surrogate Court of the King’s County in New York observed that while the court may be dubious about using AI to prepare legal documents, the attorney’s failure to review AI-generated cases without proper examination was the cause of concern rather than the use of AI (See In the Matter of Samuel (2024) WL 238160.
109 See David Lat, “AI use in Law Practice Needs Common Sense, Not more Court Rules”, February 28, 204, Bloomberg Law, https://news.bloomberglaw.com/us-law-week/ai-use-in-law-practice-needs-common-sense-not-more-court-rules.; Stephanie Wilkins, “The Problem with the ‘Bogus’ ChatGPT Legal Brief? It’s not the Tech,.” .online,. https://businessday.ng/news/legal-business/article/the-problem-with-the-bogus-chatgpt-legal-brief-its-not-the-tech/.