Mass Call Address 2023
21 August 2023
"Answering the Call in the Age of Artificial Intelligence"
The Honourable the Chief Justice Sundaresh Menon
Supreme Court of Singapore
1. Good morning. Let me begin by extending my heartiest congratulations to all of you on your call to the Bar. This is the first time that we are holding the Mass Call in a fully physical format since 2019, and I am delighted to be able to welcome each of you and your guests, in person, today.
2. Barely nine months ago, on 30 November 2022, a little-known startup launched what it called a “research release” of a new digital product.(1) Over the next few days, that product took the world by storm, and by January, just two months later, it had around 100 million users.(2) I am referring, of course, to ChatGPT. The rise of this tool and other tools employing generative artificial intelligence (or “AI”) has profound implications for our societies. The eminent American statesman, Henry Kissinger, and Eric Schmidt, the former CEO of Google, contend that generative AI has ushered in a new epoch in human history that they call the Age of Artificial Intelligence. They suggest that generative AI will have far-reaching philosophical and practical consequences, because it will challenge our very concepts of thought and knowledge, and will also transform many aspects of our world including education and international relations.(3)
3. So, what does this all mean for the legal profession? And how should we respond to the risks and opportunities of generative AI? As we gather to welcome you to the profession at a very exciting time, I propose to outline today some thoughts on how generative AI has begun and will continue to transform legal systems, and then to touch on how we should start thinking about lawyering in the Age of Artificial Intelligence. As I will explain, amidst a rapidly evolving legal landscape, we must hold fast to our enduring values as a learned and honourable profession and recommit ourselves to the continuous pursuit of these fundamental ideals.
II. Generative AI and its impact on legal systems
4. Let me begin with an overview of the impact of generative AI on legal systems. For today, I will focus on two main aspects: namely, the new legal tools and new legal issues that are arising from generative AI.
5. First, generative AI will engender a range of new legal tools. In a recent survey of lawyers conducted by the Thomson Reuters Institute, 82% of respondents agreed that generative AI can be used in legal work.(4) Indeed, generative AI tools have already been developed for the legal sector. A leading example is Harvey, which can answer legal queries, summarise information and prepare drafts of documents. Harvey has been adopted by Allen & Overy, a leading global law firm,(5) and by PwC, a major alternative legal service provider that is rolling it out to its 4,000 legal professionals.(6)
6. I believe that such products will transform not only the legal services sector, but also our justice system. Taking these in order, generative AI is likely to affect the methods of lawyering in at least two main ways.(7)
(a) First, generative AI tools will greatly enhance existing uses of AI in the law such as in legal research and the creation of legal documents. The current crop of tools are already highly adept at interpreting user queries and synthesising complex information. One illustration of this is the remarkable performance of GPT-4, the current version of the AI model underlying ChatGPT, in the United States Uniform Bar Examination (or “UBE”). GPT-4 was able not only to outperform the median human test-taker on the multiple choice section of the UBE, it evidently could pass the essay and problem solving components.(8) This suggests it has the potential to assist lawyers with research and related legal tasks.
(b) But beyond this, generative AI will likely bring new ways of using AI to the legal industry. Whereas it used to be thought that legal work requiring creativity or emotional intelligence lay beyond the scope of AI, at least for the foreseeable future, it has more recently been suggested that existing generative AI products can already be used for creative purposes, such as brainstorming how to draft bespoke contracts.(9) Further, some generative AI tools can even mimic empathy and emotions, raising the prospect that such tools might eventually take on some client-facing roles.
7. Apart from the day-to-day work of lawyering, generative AI will also likely affect the structures and composition of legal service providers, especially law firms.(10) Modern law firms tend to be organised in a pyramid model, with a broad base of associates supporting a small group of partners. But as generative AI takes over basic legal tasks, many entry-level roles are likely to fall away. Instead, we might expect growing demand for legal technologists – by which I mean legally trained persons with expertise in technology – as well as allied legal professionals (or “ALPs”) with skills in data analysis, design thinking and related areas. On this view, law firms of the future will likely comprise multidisciplinary teams of lawyers, legal technologists, and ALPs, working together and developing and using digital legal products.(11)
8. Beyond legal practice, generative AI tools will also have a major impact on our justice system. In particular, they will likely play a key part in the efforts of our courts to enhance access to justice. This is a central priority for our Judiciary, especially given the steady rise in the number of self-represented persons who navigate our justice system each year. We have sought to harness technology to enhance access to justice in two main ways.(12)
(a) First, we have used digital tools to enhance the delivery of legal information. For example, our courts collaborated with the Singapore Academy of Law to develop a Motor Accident Claims Online Simulator (or “MACO”) that was launched in October 2020. This is a free online tool that enables parties involved in a road accident to get a sense of their liabilities, and the likely amount of compensation that would be ordered for personal injuries. In its first one and a half years, MACO recorded over 13,000 simulations.(13)
(b) Second, we have used technology to provide legal assistance to court users, including by creating tools that help laypersons to generate legal documents. In November 2021, we launched the Divorce eService, an electronic platform that allows laypersons to generate and file the documents needed for divorce cases.(14) And in April this year, we rolled out the Probate eService which enables an executor named in a will to apply for and obtain the grant of probate within a matter of days.(15)
9. Our Judiciary will continue to explore the use of generative AI to further advance these two strategies of promoting access to justice. Apart from transforming the delivery of justice, these developments will likely have implications for legal practice. Work involving personal injury, divorce and probate were, not long ago, the exclusive preserve of lawyers, but new digital tools have emerged and these will affect the demand for the more routine aspects of legal services in these and other areas. Ultimately, the value proposition of lawyers will, to an increasing extent, lie in the sort of complex legal work that will be much more resistant to being displaced by technology.(16)
10. The second main way in which generative AI will affect our legal systems is by raising new legal issues. Let me illustrate this point with two emerging issues in intellectual property (or “IP”) law related to generative AI.
(a) The first is whether a generative AI tool may be the named inventor of a patent. This issue arose in applications filed by Dr Stephen Thaler, a computer scientist, to register patents in the name of an AI system called DABUS. The applications have been rejected in several common law jurisdictions, on the basis that only a human may be named as the inventor under existing law.(17) But the deeper issue remains as to whether the law should allow AI models to be recognised as inventors. This raises a number of knotty questions, including how to determine whether an innovation is an AI invention, and the proper legal approach we should take to inventions that were not conceived by any human.
(b) The second issue relates to whether using data to train generative AI models may infringe on IP rights. For instance, Getty Images, the well-known stock-image licensing service, has sued the company behind Stable Diffusion, a text-to-image tool, alleging infringement of its copyright and trademark rights in its database of more than 12 million photographs.(18) The case raises several complex issues, including whether and to what extent AI-generated output amounts to a “derivative work” under IP law.
III. Lawyering in the age of AI
11. How then should lawyers respond to these significant shifts in the legal landscape that are arising from developments in AI? I suggest that the answer lies in returning to our core values as a learned and honourable profession, and devoting ourselves to upholding those cherished ideals. This calls for two things. First, at an individual level, we who are engaged in the legal profession need to commit to a lifelong journey of learning so that we can contribute meaningfully to the discussion of and around these issues. And second and more specifically, we need to think deeply about the ethical and responsible use of AI tools. Let me expand on both these two points.
A. A learned profession: committing to lifelong learning
12. Specialised knowledge and expertise lie at the heart of every profession.(19) But this is perhaps especially true of the legal profession, because our status as a learned profession is central to our identity. Yet, as generative AI transforms our operating environment, much of what we know about the law, legal practice, our justice system, and our societies more broadly, will become obsolete or irrelevant. This reflects a broader trend known as the decreasing “half-life of knowledge,(20) which refers to the ever-diminishing amount of time that it takes for half of the knowledge in a particular field to be superseded or to become obsolete. In recent decades, advances in science and technology have driven a rapid fall in the half-life of knowledge. Generative AI is the most recent and perhaps most powerful catalyst of this phenomenon.
13. The decreasing half-life of knowledge presents a serious challenge to the continuing relevance and value of our profession. To combat it, we must commit to a path of lifelong learning. As I put it on another occasion, “we can only remain a ‘learned profession’ by becoming a profession of learners”.(21) Hence, admission to the Bar cannot be understood as the terminus of a legal education, but rather, as an early marker on a ceaseless journey to learn, unlearn, and relearn, so that we may earn and preserve the title of a learned profession.(22)
14. What then should you strive to learn in relation to generative AI? I suggest that there are at least three areas of critical importance.
15. The first is core legal knowledge. You must apprise yourselves of the new and emerging legal issues arising from generative AI, including those that I have outlined earlier. This will be a major undertaking, because these issues are novel, and it is unlikely that they can all be resolved properly by applying existing legal principles and rules. Rather, basic legal concepts and doctrines will have to be developed or extended, and this might transform whole areas of the law. Consider the ongoing debate over whether cryptocurrency and other digital assets are a species of property. This discussion has led some commentators, including the UK Law Commission, to propose that the law should recognise a third category of personal property, besides things in possession and things in action, comprising cryptocurrency and other digital assets.(23) Similarly, the rise of generative AI will likely prompt reconsideration of important concepts in IP and other areas. It will not be possible to be effective legal practitioners without keeping abreast of the latest developments in these areas.
16. Second, you will need to acquaint yourselves with the emerging digital tools that will become available for legal practice. This will include learning about how to use such tools, such as how to craft effective prompts. But you will also need to be aware of the risks and limitations of generative AI products. These include the risk of what has become known as “hallucinations” as well as concerns regarding client confidentiality, and I will come back to these issues shortly.
17. Third, it is imperative that you broaden your knowledge beyond the legal domain. In particular, you should expose yourselves to some of the following areas:
(a) The first is data science and design thinking.(24) You should strive to acquire at least a basic understanding of these areas because, as I noted earlier, law firms will likely come to comprise multidisciplinary teams of lawyers, legal technologists and ALPs, and such knowledge will be crucial for you to function effectively in such teams. Hence, it is worth exploring courses such as the Singapore Management University’s Graduate Certificate in Law and Technology, which includes modules on data, design thinking, and other topics related to law and technology.
(b) Another area is digital ethics and regulation. Professor Michael Sandel, the renowned American philosopher, has identified three key areas of concern relating to the use of AI. These are privacy, bias, and what he describes as the “hardest question”: namely, whether human judgment has an indispensable place in certain important decisions.(25) It is critical that you familiarise yourselves with these and other normative issues arising from the use of AI and other digital tools, so that you can contribute to the discussions on how we should resolve them and also be alive to how we might avoid the pitfalls of such technologies.
B. An honourable profession: committing to the ethical use of AI
18. That brings me to the second pillar that should underpin our response to these developments. In my Mass Call address last year, I emphasised that honourability is foundational to our profession. As lawyers, we are accorded certain privileges and these come with weighty responsibilities, and it is crucial to the administration of justice and, indeed, the legitimacy of our legal system, that we discharge these duties honourably.(26) This takes on renewed significance in the context of generative AI, because such tools raise several ethical risks and concerns.
19. First, a significant limitation of many generative AI tools is the risk of “hallucinations” that I have already mentioned. Such tools are obviously not bound by values like honesty and integrity, and can therefore provide wholly incorrect answers.(27) Earlier this year, it was widely reported that two New York lawyers had filed a brief referring to several purported authorities that were entirely invented by ChatGPT. The lawyers failed to verify the fictitious cases, and were ultimately fined and sanctioned.(28) Indeed, the same problem has already reared its head in Singapore. There has been at least one case in which a self-represented person used ChatGPT to create submissions that included fabricated case law. This was only uncovered after counsel for the opposing party embarked on an ultimately fruitless search, and pointed out to the court that there were no such cases.
20. The risk of “hallucinations” is closely related to a second problem, which has been called “automation bias”. This is the propensity to treat algorithmic outputs as authoritative and reliable, because they appear to be the product of an objective or scientific process, especially when we do not understand them.(29) Indeed, it is evident that the two New York lawyers succumbed to this bias – their defence was that they did not imagine that “a piece of technology could be making up cases out of whole cloth”.(30)
21. Lawyers owe their clients a duty of reasonable competence and diligence, and have an obligation to the court to refrain from advancing inaccurate factual or legal propositions.(31) Hence, lawyers who use generative AI tools must verify the accuracy and currency of all AI-generated output.(32) Such tools can serve at best as a supplement, and not as a substitute, for the independent judgment and work of lawyers. It has also been suggested elsewhere that lawyers should keep records of the prompts that they use and the output generated by AI tools, so as to be able to address any concerns or queries that might subsequently arise in connection with their use of generative AI.(33)
22. As I noted earlier, generative AI products also raise confidentiality concerns. In particular, lawyers who include client information in the prompts that are fed into generative AI products may infringe duties of client confidentiality. This can have serious consequences, especially if the tools are publicly accessible systems such as Bing or ChatGPT. This risk has led companies like Samsung, JP Morgan and KPMG to ban or restrict the use of such generative AI tools in the workplace.(34) Lawyers must therefore exercise care when using generative AI products, to ensure that they do not breach their confidentiality obligations to their clients.(35)
23. The point is to recognise that our growing use of generative AI tools will come with various ethical risks, and resolutely commit to their appropriate and ethical use, so as to preserve our status as an honourable profession. The central importance of honourability to our vocation is reflected in the motto of the Singapore Academy of Law (or “SAL”): Honor est in Honorante. This reminds us that honour lies in honouring: one becomes honourable only through the honour one bestows upon fellow members of the profession, the court and the public.(36) To reinforce this message, the SAL will start a new tradition this year of presenting each newly called member of the Bar with a customised lapel pin, which bears the SAL’s motto and each member’s unique AAS number. I hope this will serve as a reminder to you of the integral commitment to honour that is foundational to our profession.
24. You are entering our profession at a time of disruption and transformation, precipitated by the advances in technology. Generative AI and other emerging digital tools will present significant challenges for our profession, but they will also offer many opportunities. My message today is that in navigating the evolving landscape, you should take our core values as a learned and honourable profession as your constant lodestars. These tenets should guide you to pursue lifelong learning and lead you to use the tools that will emerge in the Age of Artificial Intelligence responsibly.
25. Let me close by thanking the families and, in particular, the parents of the newly called members of the Bar. Your unwavering support for these young men and women has brought them to where they are today, and I salute you for the countless contributions and sacrifices that you have made. Their milestone achievement today is as much your accomplishment and your moment for celebration as it is theirs.
26. Finally, on behalf of the Judiciary and the profession, let me extend my heartiest congratulations once again to all of you on your call to the Bar. I warmly welcome you to our profession.
27. The court is now adjourned.
(1) OpenAI, “Introducing ChatGPT”: https://openai.com/blog/chatgpt.
(2) Krystal Hu, “ChatGPT sets record for fastest-growing user base – analyst note”, Reuters (2 February 2023): https://reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/.
(3) Henry Kissinger, Eric Schmidt and Daniel Huttenlocher, “ChatGPT Heralds an Intellectual Revolution”, The Wall Street Journal (24 February 2023): https://www.wsj.com/articles/chatgpt-heralds-an-intellectual-revolution-enlightenment-artificial-intelligence-homo-technicus-technology-cognition-morality-philosophy-774331c6.
(4) Thomson Reuters Institute, ChatGPT and Generative AI within Law Firms (17 April 2023), p 7 (albeit a smaller majority of 51% of respondents felt that generative AI should be applied to legal work, with some raising concerns about confidentiality and other risks involved in the use of such tools).
(5) It was reported that up to a quarter of Allen & Overy’s lawyers use Harvey every day: see Chris Stokel-Walker, “Generative AI is Coming for the Lawyers”, Wired (21 February 2023) (“Stokel-Walker”): https://wired.co.uk/article/generative-ai-is-coming-for-the-lawyers.
(6) PwC, “PwC announces strategic alliance with Harvey, positioning PwC’s Legal Business Solutions at the forefront of legal generative AI” (15 March 2023): https://www.pwc.com/gx/en/news-room/press-releases/2023/pwc-announces-strategic-alliance-with-harvey-positioning-pwcs-legal-business-solutions-at-the-forefront-of-legal-generative-ai.html.
(7) Sundaresh Menon CJ, “Legal Systems in a Digital Age: Pursuing the Next Frontier”, Opening Address at the 3rd Annual France-Singapore Symposium on Law and Business (11 May 2023) (“Legal Systems in a Digital Age”), paras 17–19.
(8) Daniel Martin Katz, Michael James Bommarito, Shang Gao and Pablo David Arredondo, “GPT-4 Passes the Bar Exam”: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4389233.
(9) Esther Ajao, “How ChatGPT can advance AI in the law industry”, TechTarget (24 February 2023): https://www.techtarget.com/searchenterpriseai/news/365531776/How-ChatGPT-can-advance-AI-in-the-law-industry.
(10) Legal Systems in a Digital Age, para 31.
(11) Armour, Richard Parnham and Mari Sako, “Augmented Lawyering” (2022) University of Illinois Law Review 71 at 95.
(12) Legal Systems in a Digital Age, para 43.
(13) SG Courts, Annual Report 2021, p 28: https://judiciary.gov.sg/docs/default-source/publication-docs/sg_courts_annual_report_2021.pdf.
(14) SG Courts, Divorce eService: https://www.judiciary.gov.sg/services/e-platforms/divorce-eservice.
(15) SG Courts, Probate eService: https://www.judiciary.gov.sg/services/e-platforms/probate-eservice.
(16) Legal Systems in a Digital Age, para 29(b).
(17) The applications have been rejected in Australia, Europe, the United Kingdom, and the United States. On the other hand, the application was allowed in South Africa which does not have a substantive patent examination regime under which the merits of a patent application are evaluated: see John Villasenor, “Patents and AI Inventions; Recent court rulings and broader policy questions”, Brookings Techtank (25 August 2022): https://www.brookings.edu/blog/techtank/2022/08/25/patents-and-ai-inventions-recent-court-rulings-and-broader-policy-questions/.
(18) Blake Brittain, “Getty Images lawsuit says Stability AI misused photos to train AI”, Reuters (7 February 2023): https://www.reuters.com/legal/getty-images-lawsuit-says-stability-ai-misused-photos-train-ai-2023-02-06/.
(19) Richard Susskind and Daniel Susskind, The Future of the Professions: How Technology Will Transform the Work of Human Experts (Oxford University Press, 2022 Edition), p 50
(20) Sundaresh Menon CJ, “A Profession of Learners”, Mass Call Address 2019 (27 August 2019) (“A Profession of Learners”), paras 9–10; Fritz Machlup, The Production and Distribution of Knowledge in the United States (Princeton University Press, 1962); Samuel Arbesman, The Half-Life of Facts: Why Everything We Know Has an Expiration Date (Penguin, 2013).
(21) A Profession of Learners, para 41.
(22) A Profession of Learners, para 2.
(23) UK Law Commission, Digital Assets: Final Report (27 June 2023), ch 3 and 4: https://www.lawcom.gov.uk/project/digital-assets.
(24) Václav Janeček, Rebecca Williams and Ewart Keep, “Education for the provision of technologically enhanced legal services” (2021) 40 Computer Law & Security Review 1 at 5–8.
(25) Christina Pazzanese, “Ethical concerns mount as AI takes bigger decision-making role in more industries”, The Harvard Gazette (26 October 2020): https://news.harvard.edu/gazette/story/2020/10/ethical-concerns-mount-as-ai-takes-bigger-decision-making-role/.
(26) Sundaresh Menon CJ, “The Legal Profession as an Honourable Profession”, Mass Call Address 2022 (23 August 2022), paras 6–7
(27) Legal Systems in a Digital Age, para 20(a).
(28) “New York lawyers sanctioned for using fake ChatGPT cases in legal brief”, The Straits Times (23 June 2023) (“New York lawyers sanctioned"): https://www.straitstimes.com/world/new-york-lawyers-sanctioned-for-using-fake-chatgpt-cases-in-legal-brief.
(29) The Australasian Institute of Judicial Administration, “AI Decision-Making and the Courts: A guide for Judges, Tribunal Members and Courts Administrators”, at pp 34–35.
(30) New York lawyers sanctioned.
(31) See rr 5(2)(c) and r 9(2) of the Legal Profession (Professional Conduct) Rules 2015 (Cap 161, S 706/2015) (“Professional Conduct Rules”).
(32) Allen & Overy lawyers who use Harvey are shown a list of rules upon logging into the platform, including a requirement that they “validate everything coming out of the system: see Stokel-Walker. Judges in Illinois and Texas have also made standing orders requiring lawyers to certify either that their court filings were not created using generative AI, or that a human had verified the accuracy of any AI-generated content: Devin Bates, “Taming the AI Beast: Judges Set Rules to Control Use of Generative AI in Their Courts”, JD Supra (8 June 2023): https://www.jdsupra.com/legalnews/taming-the-ai-beast-judges-set-rules-to-2716202/.
(33) New South Wales Bar Association, Issues Arising from the Use of AI Language Models (including ChatGPT) in Legal Practice (12 July 2023), p 2.
(34) Mark Gurman, “Samsung Bans Staff’s AI Use After Spotting ChatGPT Data Leak”, Bloomberg (1 May 2023): https://www.bloomberg.com/news/articles/2023-05-02/samsung-bans-chatgpt-and-other-generative-ai-use-by-staff-after-leak; Katie Prescott, “Companies Block ChatGPT Amidst Privacy Fears”, The Times (23 February 2023): https://www.thetimes.co.uk/article/companies-block-chatgpt-amid-privacy-fears-5bgb5dlv9.
(35) See r 6 of the Professional Conduct Rules.
(36) Sundaresh Menon CJ, “The Singapore Academy of Law: An Essential Dedication to Honour and Service”, Singapore Academy of Law Annual Lecture 2018 (11 October 2018), paras 36–41.