SPEECH BY MR TAN KIAT HOW, SENIOR MINISTER OF STATE, MINISTRY OF DIGITAL DEVELOPMENT AND INFORMATION & MINISTRY OF HEALTH, AT THE GOOGLE FOR HEALTH SUMMIT, 29 OCTOBER 2026

Mr Yossi,

Dr Michael,

Ms Karen,

Dr Alan,

My colleagues from the Ministry of Health,

Mr Ben King, Country Managing Director of Google Singapore,

Distinguished guests, ladies and gentlemen, friends,

1.               A very good morning to everyone.

2.               It is my real pleasure to join you at the very first Google for Health Summit in Singapore. It’s not every day that I get to participate in a very exciting event that ticks both boxes: my Ministry of Digital Development and Information role, as well as my Ministry of Health role, so the intersection of both technology and health. So very happy to be here today, and to allow me to share some perspectives with all of you.

3.               We have heard several exciting and significant announcements from speakers earlier. Aa new Google Research Hub in Singapore; new work with National University Health System (NUHS) exploring how AI can support clinicians with Google DeepMind, and, over time, participate more actively in care workflows; and a collaboration between Google and NHG Health to explore how AI can identify health risks earlier and support residents before illness becomes serious.

4.               Taken individually, each effort is important. But together, they tell a bigger picture, a bigger story — from discovery, to clinical care, to population health. And they point to a broader question: as AI becomes dramatically more capable, how do we turn that capability into better outcomes for people, safely, responsibly, and at scale? The answer is not AI alone. It is what AI can do when it is connected to a capable health system.

From Possibility To Impact

5.               Google has shown before what frontier AI can make possible. AlphaFold transformed our ability to predict protein structures. But the lasting impact did not come from the model alone — it came because researchers could use those predictions, build on them, and translate them into further advances. So the lesson is simple: a breakthrough creates possibility. A system converts possibility into impact in the real world.

6.               And the same is true for healthcare AI. A powerful model does not automatically produce better health. It has to connect to trusted data, fit into real workflows, help professionals make better decisions, generate evidence and value, earn the trust of patients and users, and if it works, the system must be able to scale it.

7.               The early gains from AI are real — faster documentation, faster information retrieval, less administrative work. But these are not the main prize. The real prize is not automation. It is transformation. AI should help remove wasteful friction — repetition, duplication and delay — while preserving the human work that matters most: judgement, learning and connection.

8.               With that, we can focus on realising our vision of what better healthcare looks like. Can AI help move care beyond the hospital walls and into the community and into the home? Can it help us move from reacting to illness towards identifying risk earlier? Can it help us move from one-size-fits-all care towards more personalised support? The greatest value of AI may not come from doing today's healthcare faster. It comes from enabling forms of care that were previously difficult to deliver at scale.

9.               And that is why the population health collaboration between NHG Health and Google matters. It will explore whether predictive insights can identify health risks earlier, and whether an AI-enabled health coach can provide more personalised preventive guidance while directing clinically sensitive or urgent care towards appropriate care channels. Earlier insight. Earlier intervention. Support between healthcare visits. Potentially, a health system that can reach people before problems become crises. AI should not only help healthcare become faster. It should help healthcare become earlier.

From Answers To Actions

10.           AI is also beginning to play a more active role in care. The AI Co-Clinician initiative goes beyond decision support. It explores a future in which AI could eventually participate in patient-management workflows under physician oversight — beginning in simulation before any progression towards real-world use. That creates opportunity, but it also raises new questions about assurance.

11.           It is no longer enough to ask did the AI produced the right answer? We also have to ask: was the action appropriate? Was it within clearly defined boundaries? When should the system escalate? Which decisions should remain with a healthcare professional? How do we detect error, drift or unexpected behaviour? This is why rigorous evaluation before deployment matters.

12.           And we need trusted benchmarks that test not only whether AI performs well in general, but whether it is safe, robust and suitable for the specific workflow, population and context in which it will operate. So simulation first. Careful observation next. Real-world deployment only when the evidence supports it. And assurance cannot stop at deployment — once a system enters real care, we need to keep watching how it performs. Pre-deployment evaluation tells us whether a system is ready to enter care. Continuous assurance tells us whether it continues to deserve our trust once it is there.

13.           That is the standard we should hold ourselves to. Because in healthcare, we should not scale because AI is impressive — we should scale because the evidence compels us to. And as machine agency grows, human accountability must become clearer, not weaker.

Why Singapore?

14.           So this brings me to the next question. Why does this work here in Singapore? Not because Singapore has all the answers — we do not. We are not the world's largest healthcare market, and we do not have the largest research budget. But we have spent years building the foundations needed to turn technological possibility into real-world impact.

15.           First, our systems and data. We have connected health information, national digital infrastructure and common platforms that allows innovation to move beyond isolated pilots. We also have distinctive research assets, including multi-ethnic Asian genomic resources linked with longitudinal health information. Together with our health system, these strengthen our ability to study how technologies perform across different populations and in real-world care settings, not only in controlled settings. So that’s system and data.

16.           Second, our public healthcare institutions and professionals, many of whom are seated here today. Our healthcare clusters are responsible for defined populations and increasingly work across primary, acute, intermediate and community care. That matters because the best AI tools still need professionals with excellent judgement — people who know when to trust AI, when to question it, and when to override it.

17.           Third, our incentives. Our clusters increasingly have stronger incentives to keep residents healthy by investing upstream and working with community partners. Technology alone cannot move healthcare upstream. The incentives around healthcare have to move upstream too.

18.           And finally, trusted evaluation and continuous assurance. Singapore's healthcare AI governance increasingly follows AI across its lifecycle — from pre-deployment evaluation to real-world monitoring, incident reporting and learning. Taken together, all these factors create something distinctive. Singapore's advantage is not that we have all the pieces. It is that we can connect all the pieces — research to care, data to decisions, technology to workflows, innovation to outcomes, hospitals to community, pilots to scale. We have enough scale to test meaningfully, but are compact enough to learn as a system. That can be useful not only for Singapore, but for companies and health systems around the world.

Trust Must Be Earned

19.           But let me bring us to an important ingredient for all this to succeed. None of this works without trust. And trust cannot simply be declared. Trust is not what comes after innovation — trust is what allows innovation to scale. In healthcare, trust has to be earned through how AI is developed, evaluated, deployed and continuously assured.

20.           And the real world is different from the test environment. That is why we need to measure not only whether the technology works, but whether it improves care and outcomes — where does performance change in real-world use? Where do users over-rely on the system? What does incidents and near misses reveal that pre-deployment testing did not? Healthcare AI governance therefore cannot be a one-time approval. It has to become a continuous learning loop: evaluation, implementation, real-world evidence, assurance, and improvement.

21.           The measure of progress cannot simply be how many AI tools we deploy. The real measure is how quickly we learn what works, scale what works, and stop what does not. That is how trust is earned. And that is how we move from pilots to responsible scale.

Build Here. Learn Here. Prove Here.

22.           This is also the model of partnership we should pursue. The work with NHG Health and NUHS gives us a chance to go deep inside real health systems and understand what works. Build here. Learn here. Prove here. Then work out together what can travel.

23.           Our proposition to Google and the other frontier technology companies is not that Singapore already knows the answers — almost the opposite. Google brings frontier research, models and engineering capabilities, talents, and the best minds at global scale. Our health system brings clinical reality: professionals, researchers, population health teams, community partnerships, and responsibility for the people they serve. Google brings the frontier of what technology can do. Our health systems bring the frontier of what healthcare needs. And that is why this partnership matters.

24.           The problems should be defined together, the solutions developed together, the evidence generated together, and learning should continue together. Ultimately, Singaporeans should not experience the outcome as "more AI". They should experience better care — earlier intervention, more personalised support, care closer to home, smoother journeys, better tools for healthcare professionals. If patients, caregivers and healthcare professionals cannot see or feel the value of what we do, then I would say we have not succeeded.

A Glimpse Of What Could Go Right

25.           There is also a wider significance to this work that I would like to talk about. We meet at a time of understandable anxiety about AI across the world, and we see it in many different countries — what will more capable systems mean for jobs, for professions, for accountability, for safety, for trust? These are real questions, real anxieties. Healthcare is one of the hardest places in which to answer them. The stakes are high, the data is sensitive, workflows are complex, professionals remain accountable, and trust is indispensable. And that is exactly why healthcare matters.

26.           Healthcare can be a proving ground for how society learns to use increasingly capable AI. If we can show that AI creates real value here while making professionals more effective, preserving human judgement, earning trust, being rigorously evaluated before deployment and continuously assured in real-world use — then the lessons can travel. To finance, to education, to government, to professional services, to other sectors where judgement, accountability and trust matter. And they can travel across markets too, because what we are building is not only a set of healthcare solutions. We are also building ways of working with AI in high-trust environments.

27.           At a time when much of the global conversation is focused on what could go wrong, Singapore, working our partners, has an opportunity to offer a glimpse of what could go right. Not because we believe we are uniquely clever, not because we have all the answers, but because we have spent years building institutions that allow technology, people, incentives, governance and trust to work together.

Closing

28.           So let me end with a concluding thought. The world does not need another demonstration of how impressive AI can be — we already know that. What we need now are demonstrations of how AI can improve lives, earn trust, and strengthen human agency.

29.           Singapore's opportunity is to be a place where frontier technology meets real-world needs; where AI is rigorously evaluated, continuously assured and deployed in high-trust environments. If we can make AI work safely, responsibly, and at scale here, the lessons can travel far beyond healthcare — and far beyond Singapore. The announcements today, spanning discovery, clinical care and population health, give us an opportunity to do exactly that.

30.           The next frontier of healthcare AI is not frontier models alone. It is frontier collaboration between AI companies and health providers, and healthcare systems. We do not have all the answers, but we have built an environment where we can discover them together — through evidence, experience and partnership. The question is no longer simply what AI can do. It is what we choose to become capable because of it.

31.           That is the future I hope we can help build here together, in partnership with Google — and share with the world.

32.           Thank you.