Remarks by Minister Josephine Teo for Digital Development and Information at Digital@UNGA

Distinguished colleagues and friends, good evening and thank you for inviting Singapore to share our views on digital inclusion, digital development and in particular AI.

There is tremendous excitement about what AI can do but in Singapore we try to focus on a slightly different question – Not simply: How capable is the technology? But: What difference does it make to people?

We have been fortunate – we are not starting from scratch.

Over many years, Singapore has invested in digital connectivity, digital identity, cloud infrastructure, data, and digital capabilities within government. For us, infrastructure is less of an immediate constraint. Our challenge is how to make good use of the foundations and strengthen them.

Our first challenge is therefore to deploy AI for impact, including within the government. We see two opportunities in the public service: helping our officers work better, and serving our citizens better.

Almost 70% of public officers use AI tools regularly. They were encouraged to, and they do so for tasks such as analysing information and preparing drafts. We are now exploring the use of AI agents that could help us redesign entire workflows while keeping officers accountable for important decisions.

AI is also helping us serve our citizens better. We use machine learning to route feedback so that a broken swing or faulty streetlamp gets to the right agency more quickly. At our borders, AI helps identify anomalies in millions of cargo shipments. To combat scams, AI helps us detect signs of phishing and impersonation.

These examples cut across multiple agencies and functions. But the principle is the same: the real measure of AI innovation is not how impressive the model is. It is whether it improves people's lives.

We are also very mindful that over time, people will only rely on AI if they trust it.

AI can be used for many types of harms. For example, scams can become more convincing and personalised; cyberattacks are easier to execute.

Building trust means managing these risks and putting the right safeguards in place, so that AI can be adopted with confidence. From that perspective, trust does not constrain innovation. In fact, trust allows innovation to scale.

There is a third dimension that matters, and that is: how do we broaden participation?

This is linked to the question of access. Southeast Asia has extraordinary linguistic and cultural diversity. Yet many AI models have weak understanding of our languages, accents, and cultural contexts.

So we built SEA-LION. It stands for Southeast Asian Languages in One Network, as well as another model that is multimodal in nature. These are open models designed around Southeast Asian languages and cultures, including the multilingual speech and code-switching that come very naturally to us.

AI Singapore is a research institute that is also working with partners like the Gates Foundation, Google, and the United Nations Development Programme (UNDP) on Project ATLAS, to build AI-ready data sets for underrepresented languages across the globe.

For us, this is how AI inclusion can be meaningfully expanded. Countries should not all have to build frontier models. Indeed, most of us will not have the capacity to do so. But we should be supported to adapt AI to our own languages, cultures, and needs. That's what we believe.

We also want to share what Singapore has learnt through the Digital Pillar of the Forum of Small States (FOSS), a community of 108 small countries, Singapore shares practical experiences, open-source digital tools and ways to build digital capabilities.

We have also created the Singapore Digital Gateway, where partners can access our digital toolkits and resources, and find capacity building opportunities through initiatives such as the Singapore Cooperation Programme. We also work with other partners such as the World Bank to deliver such opportunities.

To summarise, there are three key learnings from our AI journey thus far.

1. First, deploy for impact including within government.

2. Second, build trust for sustained use.

3. And third, broaden participation as much as we can to as many communities as we can reach.

 Ultimately, the success of AI should not be measured by how powerful our machines become. It should be measured by how many people are empowered by them.

On this note, I want to thank you once again for inviting us. Have a good discussion.