Transcript of Keynote Address by Mr Murali Pillai SC, Senior Minister of State for Law and Transport, at the Pacific Rim Advisory Council 74th International Conference

26 October 2026 Posted in Speeches

“AI and its Impact on Business and Society” - Regulating Artificial Logics

Introduction

1.                  Distinguished delegates to the Pacific Rim Advisory Council Conference. Ladies and gentlemen. Good morning.

2.                  It is a pleasure to welcome the Pacific Rim Advisory Council back to Singapore.

3.                  Singapore’s connection with PRAC goes back a long way. PRAC held its very first international conference here in Singapore in 1987. Dentons Rodyk is a founding member firm of PRAC, and is, of course, a worthy host of the conference today. Thank you very much for today.

4.                  Today’s theme is how AI will impact business and society. Business and society — we see them as if they are a continuum and AI impacts them both the same, as butter spreads on bread.

5.                  For the benefit of our overseas guests, we have an equivalent of butter in Singapore. It is called kaya, and I am sure our local hosts will be able to introduce you to that — it is good, if not better.

6.                  But really the main point I want to make is that AI does not affect business and society the same. There’s always been a bifurcation of market and societal logic, and nowhere is this more clearly seen than in the impact of AI today.

7.                  Put starkly, much of AI — including the vast sums that are invested in it, or the enormous profits and vast wealth that flow to those who supply the computing power and other raw ingredients to the rise of AI — is directed by markets.

8.                  Here, the action is quick, decisive. Firms live or die by their bottom lines, and those who cannot survive under the new structural hegemons are dismissed as “uncompetitive.” Why should we seek to regulate private investing decisions, or to put brakes on entrepreneurial adventures? There are therefore few people asking “Is this right? Is this good? Should we pause?” To do so would risk one sounding naïve, or uninformed, or overly conservative.

9.                  But today, I would take this risk, and urge us all to ask: The way we see businesses now, especially large tech companies — lobbying governments, training their AI models to draft bills that are meant to regulate themselves, amassing profit and power. We, especially those of us in the business of law and regulating both markets and businesses, might do well to ask “Is this right? Is this good? Should we resist or must we acquiesce?”

10.             Businesses themselves have not stood by and allowed the unfettered progress of AI — they have blocked its use where it does not serve their business interests.

11.             At least two banks have cut AI access to bank accounts since AI agents can potentially shift capital out from accounts in search of better returns. On the face of it, this is due to security concerns, but on the other, idle capital — and in general, human inattention — does return profits to companies.

12.             A large online retailer too has blocked agents from its platform, as such agents bypass its ads and search across the web for better deals. Needless to say, consumers may gain from such use of AI.

13.             Compared to business, in societies, we are slower to place constraints on AI, partly because we are more complex than businesses and care for a whole line of public goods, not just private profits. The same complexity attends to how we define and think of the “harms” from AI.

14.             But we too must constrain AI where it does not serve our interests.

15.             A recent piece by mathematician Terence Tao in the New York Times is instructive. After the release of a series of mathematical proofs by AI agents, he is, he said, and I quote him, on a “war footing”.

16.             One does not know who the enemy is, but it can be natural to feel that humans are under siege. AI, in releasing these proofs that humans have not written, nor understood, has stolen something from humanity: the elegance of mathematical logic, the thrill of the chase, and the moment of epiphany.

17.             I am not sure how many in this audience know the Fermat’s Last Theorem. Lawyers are not known for their mathematical prowess — I have been practising for some time, so I think not much has changed. You may be more familiar with Pythagoras’ theorem: the right-angled triangle, a² + b² = c².

18.             Fermat’s Last Theorem, 300 plus years ago, basically said that if the power is more than two, you are not going to have that aⁿ + bⁿ = cⁿ. That was the theorem more than 300 years ago. It was a source of delight and mystery to mathematicians. The mystery was not whether it could be solved — Fermat himself said he solved it — but what exactly was the solution.

19.             This conjecture, scribbled into a margin of a book, stood unsolved for, as I said, 350 years, until a British mathematician, Andrew Wiles, in 1993 gave a lecture where he wrote out the proof, line by line, by hand, on a blackboard, in chalk.

20.             And when he had finished, he turned to the audience and humbly said “I think I will stop here.” He was greeted by a storm of thunderous, prolonged applause.

21.             The recent release of 722 proofs by OpenAI had no such human hand. AI has done so, at the instruction of private interests, for their benefit and for their particular gain.

22.             This is the conduct of business. And we may well say — well, they breached no law, we suffered no harm. Let society work itself out — and let the rules evolve as each jurisdiction sees fit.

23.             I agree that rules are anchored in specific contexts. But rather than an organic pace of evolution, there is an unassailable argument for speed and collective action. AI does not move at human scale, nor at human speed. It moves at an artificial speed — as fast and as long as businesses are willing to fund and empower it.

24.             The laws regulating such businesses and technologies must not lag. To do so would be to see money and power to trump rights and justice and the moral rules of society. Let me take these in turn.

A. AI adoption across different jurisdictions

25.             Let me speak across my two portfolios — law and transport. I begin with the rules of the road, and with six ports on the Pacific Rim: the US, China, Australia, South Korea, Japan and Singapore. Each offers a different perspective on AI adoption and governance.

United States

26.             In the US, the conversation has moved beyond experimentation to adoption. Clio’s 2025 Legal Trends Report found that 79% of surveyed US legal professionals used AI in their firms.

27.             But adoption does not necessarily mean institutional readiness. Nearly half reported using general-purpose AI tools, while more than half said their firm either had no AI policy, or they were unaware of one.

28.             This US experience illustrates how quickly AI adoption can proliferate, while firm-wide training and governance trail behind.

China

29.             China represents a different model of a large AI ecosystem, supported by domestic AI tools and open-source models.

30.             Shenzhen offers a concrete example. Its courts have widely deployed an AI-assisted adjudication system, supporting tasks from case filing and document review to hearings and drafting judgments. The system assists judges, who still remain responsible for the final decisions.

31.             China also regulates generative AI services and has introduced labelling requirements for AI-generated content.

32.             This reflects how legal-sector AI adoption is shaped by both domestic technological capabilities and the wider regulatory environment.

Australia

33.             Australia offers a particularly relevant comparison for Singapore as another common-law jurisdiction.

34.             In April this year, the Federal Court of Australia issued a Generative AI Practice Note, setting expectations around matters such as verification, confidentiality, and disclosure of AI use to the Court.

35.             Professional bodies have also developed practical guidance for responsible AI use.

36.             Australia is therefore an example of how courts and professional bodies can translate longstanding professional obligations into practical safeguards.

South Korea and Japan

37.             Other Pacific markets offer further contrasts. South Korea’s AI Basic Act includes transparency obligations for generative AI, while Japan’s AI Promotion Act takes a more promotion-oriented approach.

38.             These illustrate the range of approaches, from direct regulation to guidance and promotion.

Key takeaways

39.             The common lesson is that AI adoption and accountability must develop together.

40.             For firms operating across borders, this also means navigating the different legal and professional requirements that apply in each jurisdiction.

Singapore’s approach

41.             For Singapore, we are helping firms, particularly the smaller practices, adopt technology in a meaningful way, not simply to acquire tools. The Legal Innovation and Future-Readiness Transformation, or LIFT initiative is a good example.  Firms receive practical support to identify workflow needs, select suitable solutions, and implement them effectively. To date, 28 small and medium Singapore law firms have participated in the LIFT pilot.

42.             Second, we are promoting responsible adoption. In March this year, MinLaw published its Guide for Using Generative AI in the Legal Sector. The Guide provides practical, non-binding guidance anchored in three principles: professional ethics, confidentiality and transparency.

43.             Third, we are working with Infocomm Media Development Authority and the Singapore Academy of Law to build AI capabilities through AIxLegal. This is a capability building programme which provides practical AI training tailored to lawyers and in-house counsel. The aim is not to turn every lawyer into a technologist, but to help legal professionals use AI confidently and responsibly in their daily work.

B. Client expectations

44.             As AI reshapes legal work, it will also reshape the client–lawyer relationship. Clients want to know what the law means for their businesses: the risks, the available options, and the next steps.

45.             That puts a premium on lawyers who understand their clients’ business, industries and regulatory environments.

46.             Clients are also changing what they outsource.

47.             Across the Asia-Pacific, in-house teams are also using AI to handle more routine work themselves. One general counsel told us that the in-house legal department is now the law firm’s main competitor. Another said they would pay a premium only for work that could not be readily done by their own team, another provider, or a machine.

48.             While AI may change what clients expect and what lawyers do, it does not change who remains responsible for the advice.

C. Professional responsibility and trust

49.             Let me highlight three practical issues that the profession must address as AI becomes more deeply embedded in legal practice.

50.             First, professional ethics and training. We can certainly leverage AI to expand capacity and drive productivity. But embracing technology does not mean an uncritical surrender to it. We must use AI with our eyes open, because it can produce answers that sound impeccably convincing while being fundamentally wrong. This is particularly important for the next generation of lawyers.

51.             Sound legal judgment cannot be automated nor fast-tracked. Junior lawyers must still read cases, construct arguments and learn to think critically for themselves, even where AI may appear to perform these tasks faster. If developing legal acumen requires them to spend time doing what a machine can do in seconds, then that time is well spent.

52.             We should embrace technology, but we should also enable the use of technology to make sure that the lawyers, the young lawyers, continue to develop.

53.             Second, confidentiality. The duty to protect client information remains unchanged, but AI introduces new risks. Different tools have different arrangements for handling confidential information, including whether client data may be retained or used to train AI models.

54.             Firms must therefore understand how their tools handle client information and put appropriate safeguards in place, whether through anonymisation or secure, enterprise-grade solutions.

55.             For international firms, this is particularly important when working across jurisdictions with different data protection requirements as well as client expectations.

56.             Third, billing for AI-assisted work. AI also raises important questions about how lawyers should charge for work performed with AI assistance.

57.             Clients increasingly expect lawyers to use AI to reduce their hours spent working on the client’s matter, and for the lawyer’s bill to be reduced significantly – even disproportionately – as a result.

58.             Indeed, AI can reduce the time spent on certain tasks. However, determining reasonable fees is not simply a matter of counting hours. Firms must also account for the professional expertise, judgment and responsibility involved in delivering the work.

59.             The challenge is to ensure that law firms’ billing practices fairly reflect both AI-driven efficiencies and the value delivered to clients, balancing transparency in how fees are determined and managing clients’ expectations.

Closing

60.             Let me conclude. I have urged for greater attention to governance and regulation of AI. Our laws and regulatory approaches will continue to differ, as will the ways and speed of how we deploy technologies. The challenge is how we work across those differences.

61.             Our task, in the age of AI, comes down to three things: common principles that we can navigate together; knowledge we share across borders; and last, responsibility that we cannot delegate.

62.             My main point is that the impact of AI falls differently on business and society. Businesses obey the laws of the market, and their logic is economic – along this logic, they have enabled or impeded the use of AI.

63.             But beyond this, businesses can also be compelled by the laws of the jurisdictions in which they operate. These laws are made according to a different logic – that of the social and moral community to which they apply.

64.             This logic is complex and at times, imperfect and boundedly rational, but in its best version, strives to be good and just and fair. It is the logic of human beings, used to moving at human speed.

65.             My provocation is that this speed may not match the artificial logics we need to govern. We must therefore, as a community, be galvanised into action and a collective ambition not just to harness the power of artificial technology but to constrain it where and when we feel it must not go.

66.             I wish you a productive conference. Thank you.

Last updated on 26 October 2026