Remarks by Minister Josephine Teo at Foreign Policy’s Tech & AI Forum @ UNGA81

Moderator Rishi Iyengar, Technology Reporter, Foreign Policy: There's a lot of conversation around AI right now. Every country is taking their own approach. There are some universal truths, but each country is approaching it its own way. I'm wondering if you can lay out Singapore's approach to AI, how Singapore thinks about the technology as a whole and how Singapore uses the technology.

Minister Josephine Teo: That's a great question to start. It's fair to say that AI is not entirely new. Our own experimentation with AI probably took on a different pace around 2018 – this was before COVID. At the time, it was still quite modest. We identified some sectors where AI projects should be pursued as a way of experimenting with how we use the technology. But most importantly, even before we came up with a National AI Strategy, we came up with an AI governance framework. This went on for a while, and then COVID hit. Coming out of COVID, it also occurred to us that the technology was maturing, particularly around late 2022 and early 2023. The big pivot towards generative AI created a great sense of urgency and momentum, and also, a great sense of opportunity. So we decided to refresh our National AI Strategy.

The question was, in this new effort, what should be our North Star? What should we be setting our sights on? As we understood the excitement around AI, it became very clear that commercial applications and industry use were very likely to take off because the incentives were so strong. We decided that it was important to be very clear on the purpose of this technology – it is ultimately to improve people's lives and serve citizens well-being. So, in a slightly unusual way, when we put out our vision statement, we said: “AI for the public good, for Singapore and the world.” It was a reminder to ourselves not to disregard how profound the impact of AI would be on people – sometimes in a very good way, sometimes not so good. How do we think about the “public good” element in all of our endeavours? We said “for Singapore and the World” because we recognise that solutions tried out in Singapore could potentially be scaled elsewhere, and this made for a more compelling purpose for rallying the community to work on things that truly matter.

That was how we thought about AI. I sometimes say that the answer to “how” is “why”. Before we can identify the strategies, we have to figure out our purpose. Setting that as our North Star, and being focused on the public good, meant that not only were we interested in AI adoption and its transformative effects in the private sector and industry, but we were also interested in its use   in the delivery of public services within government and in achieving worthy social purposes. That was how we embarked on this second phase of our AI journey.

Moderator: As you just mentioned, “AI for the public good”, and the technology brings with it a lot of opportunities. This week and in the past few weeks, it seems like the conversation has centred a lot around the risks that the technology poses, both from an immediate perspective when you're talking about public good economic impacts, job losses and increasingly centred on kind of the more alarming kind of existential risk in the debate around that. So how does Singapore think about balancing with that public good as your North Star balancing the opportunities with the risks? Because with AI in particular, it seems like that's quite a fine line.

Minister: We have consistently thought of the need for technology to be trusted before it can be regularly relied upon. Take aviation, for example: you and I would not step onto a plane unless we trusted it to land safely and get us home. The same applies to many other technologies – we will use them regularly only if we trust that they will not harm us. But there is a range in terms of the harm that technology can cause us. While it is not irrelevant for us to think about extreme risks, if we set aside the speculation as to how extreme these risks are going to be and how soon they will be upon us, there are many other types of risks that we have to manage.

I think it is important to acknowledge that whilst we keep an eye on extreme and catastrophic risks, we must not disregard the importance of working on other types of risks, such as AI systems producing discriminatory effects, and being victims of or falling victim to cyberattacks. These are also risks that we have to manage.

The difficulty is that today, in aviation, we generally feel safe. However, we haven't arrived at the same point with AI. But if we ask ourselves, how do we get from people not being willing to step on a plane to people being generally unafraid to step on a plane, there is a whole bunch of things that we need to do. It's a very multi-layered approach to protection. First, you have the aircraft manufacturers. You could think of them as the equivalent of the model makers. But it's not just them. It's also air traffic control and air traffic management. It's also how pilots and crew are trained. It's how aircraft are maintained and the standards to which they must be maintained, as well as the degree of training that the individuals who maintain them must undergo.

It is an entirely holistic approach that provides useful ways for us to think about the question of AI safety. For Singapore, the choice we made was that, in the absence of a ready playbook on AI safety, our objective must be to learn as much as we can, as fast as we can. That means rolling up our sleeves and getting to it.  In 2018, we had a governance framework. Fast forward to 2023, 2024, as we pushed out our updated AI strategy, we said that our governance framework also needed to be updated. Moreover, it was important for us to ask: how can we test for safety? We therefore decided to create a testing framework and a software toolkit to test certain aspects of AI systems. We knew that these tests were not going to be perfect, but you have to learn. We continue to be on this journey of learning.

No country can hope to tackle the questions of AI safety on its own. You really need collaborators. This means working with like-minded partners that have invested in setting up AI safety institutes, finding meaningful projects on which to collaborate, and being very watchful of use cases where risks are imminent. Those are the things that we concern ourselves with.

We've not taken our foot off the pedal. In fact, we think that this is an area that deserves a greater amount of research resources. We have funding for R&D set aside for AI safety and responsible AI research. At the same time, we want to bring together experts. For two years in a row, Singapore has convened the International Scientific Exchange on global AI safety research priorities. It resulted in something called the Singapore Consensus. I hope that we can continuethis work.

Moderator: Regarding a couple of things you said on the airplane analogy, the one difference I see is that, one, AI is moving very fast, and two, there is a real sort of race and strategic competition around this technology. We have President Trump and President Xi meeting – the two preeminent powers. Singapore is one of many middle powers, but a very advanced open economy. How do you navigate that strategic competition when you are looking at sometimes asked to choose between the US tech stack and the Chinese tech stack? How does Singapore navigate that kind of landscape?

Minister: In the first place, we cannot simply expect to make progress by accessing any single platform or ecosystem. This approach is not sustainable and does not build resilience. Instead, we try to stay connected with diverse sources of technology and innovation. We have longstanding cooperation with the US in many areas. Similarly, our cooperation with China is long and deep. We also cooperate with our partners in Europe and other parts of the world.

Singapore is a very open place. What we also try to be is a trusted and credible partner to those we work with. This means that if we give a commitment, we will honour it, and we will be candid about what we require our partners to provide within this relationship.

Importantly, being a trusted and credible partner does not relieve us of the responsibility to build our own   capabilities where we can. If you want to be able to plug into different tech ecosystems, you must have the capabilities in your workforce. You must have the capabilities in your enterprises. The government itself must understand how these different technologies are designed and implemented, and how the risks they carry are different from one another.  The government must also have the ability to govern them.

Another layer of engagement that will remain important is finding ways to support international arrangements that strengthen interoperability and connectivity. So there is no single silver bullet to deal with the challenge of navigating a very complex, contested environment. It requires us to be focused and thoughtful in the way we engage.

Moderator: On the subject of multilateral institutions, Singapore is about to take over the chairmanship of ASEAN. What role do you see for multilateral institutions like ASEAN in the regional context, like the UN in the global context, in regulating this technology?

Minister: They remain critical to countries like Singapore, which hope very much that the world can remain as open and connected as it can be. Fortunately for us in ASEAN, we are able to work on positive agenda items such as a Digital Economy Framework Agreement. Beyond that, we work on practical issues. For example, data is critical to anyone who wants to see AI developed in their respective countries. But each one of us will have limited datasets. So how do we promote trusted cross-border data flows, taking into consideration legitimate privacy concerns and national security concerns when it comes to certain types of data? Those are the things that very often don't make the headlines, but they are the things that really matter to businesses. We are also working together to promote AI adoption, particularly among small and medium enterprises.

These are some of the areas that Singapore will advance during our ASEAN chairmanship, building on the good work of the current ASEAN Chair, the Philippines. I'm also very happy to connect with our colleagues in Thailand because they will take on the chairmanship after Singapore in 2027. Although the chairmanship rotates, the priorities need not change with each term. There can be sustained interest as well as commitment. That's what we work towards.