Opening Address by Minister Josephine Teo at SID Digital & Tech Forum

1.      Good afternoon, everyone. I am very happy to join you today at this year’s Digital & Tech Forum.

2.      I wanted to take this opportunity to register three points. These are the top-of-mind things that I hope you focus on.

3.      First, where Singapore stands today with regard to AI. There are many rankings and there is a certain consistency in how Singapore is being regarded. It is not a bad thing that we are generally thought of as progressive, ranking third, fourth, or fifth.

4.      The important thing is not to be too comforted by the fact that we are number three, four, or five, because the distance between us and the one before is vast.

5.      At the same time, the distance between us and whoever comes slightly behind is actually very small. In other words, it's very competitive, and there are many countries seeking to beat Singapore.

6.      We should take that as a compliment, challenge, and encouragement. We need to continuously look at ways to better ourselves, and there are many good reasons as well as opportunities for us to do so. Our people are already using AI in a very significant way.

7.      I see Andy (Brown) here from OpenAI. I think the intensity of use across all the markets in which OpenAI operates speaks for itself. Some of this data is available, and if you track on a per capita basis, how much the top AI websites and their mobile versions are being accessed, Singapore does quite well.

8.      Having said that, if you take a look at the enterprise level, there is a gap. Whether it is SMEs or non-SMEs, this is one technology where the experience at the individual level and the enterprise level has reversed. We usually experience most of the technologies because our employers are using them and we have to keep up. With AI, this is quite different.

9.      The individual experience of using AI is often far more intense. It is also far more sophisticated than what many enterprises have adopted. You have a completely different challenge. You may be attempting to implement AI in a bigger way. Some of your employees are already using AI at an individual level. So, the question is how do you harness their energy, knowledge, and their abilities to create an environment where they can bring their individual-level productivity improvements into the workplace?

10. That is something company executives and organisational leaders have to think about. Many of your employees are ahead of you. Some of them have used vibe coding to build apps to keep track of their children’s weekend roster. They use tools to build their own news agent to collect and scan the internet for interesting developments. There is an opportunity to harness that kind of ability already present in your workforce.

11. The challenge is that what individuals bring to the workplace must come within institutional guardrails. For example, if individuals can vibe code tools for themselves, they can also do so at the workplace.

12. The problem is: what happens when that individual leaves the organisation? Where is the documentation? How was this tool deployed across customer and supplier interactions, if nobody else knows how to use it? Therefore, the question of documentation and what you, as organisational leaders, should require in terms of governance, becomes very important.

13. In addition to that, you have to consider the touchpoints with the use of AI, and whether safety requirements have been met. The Government's approach is not to stop meaningful experimentation by introducing excessive legislative or regulatory requirements. At the same time, this domain cannot be left unaddressed. Our preference is to develop frameworks and articulate clear principles to guide the responsible use of AI.

14. We will also provide guidelines and introduce tools. However, we will stop short of making it mandatory until we are very certain.

15. We will build on existing frameworks so that you have a better sense of what you need to do.  You need to be able to ask your teams the right questions. These frameworks and guidelines are intended to do just that. With the help of AI, you can ask it to interpret it for your own specific circumstance. I think they are now much more ready to be used than they were before.

16. The second thing is that cyberattacks are going to be part and parcel of everyday life. As we like to say, when it comes to law enforcement in Singapore, low crime does not mean no crime. Just because Singapore is a trusted business environment doesn't mean these threats do not exist.

17. It’s hard to talk about cybersecurity because it is a very technical field. But in today's context, no board can afford to be underinvested in this domain. It would be beneficial to build up your depth of understanding, and I'm very glad that SID has taken it upon itself to do this.

18. On my way to the ballroom, I passed by quite a number of the companies showcasing their wares. What they bring to the table ought to gain your attention, and I encourage you to engage more deeply with the issues that will confront your organisation.

19. As with AI adoption, cybersecurity is also K-shaped. If you don't have the cybersecurity foundations that keep pace with leading practices, you will find yourself falling further behind, and it will become harder and harder to catch up.

20. Some of the cyber risks posed by AI-enabled threats are quite well discussed and I won’t go into detail. I would only suggest one thing: as much as AI can intensify the threats, AI itself is very often the target. Models can be hijacked and jailbroken. You need to encourage people to use AI as a tool to fight against these threats. It is not a threat that you can fight with human effort alone. The more you are able to understand how AI can help you in the defence of your systems, the more likely you are able to succeed in it.

21. The third point I want to make is about how much we allow AI to act on its own. The power of AI agents is obviously of great interest to Singapore. We are often short of people and the right skills, so agents could potentially help us overcome this significant gap. However, it is also well documented that agents can go rogue. It is no longer a question of whether they can cause harm; they have already demonstrated that they can.

22. So, the question is what kinds of measures you introduce to minimise the risks. One of the sponsoring organisations is in the business of helping you to protect your systems against intrusion by an agent. Just as you authenticate the person that you interact with, you also need to identify and authenticate an agent. Solutions are emerging to help you do that, and you will have to judge how effective they are. But suffice to say that these capabilities are scaling up.

23. There is only one parting thought. When we think about safety, we should recognise that, today, almost all the risks we know how to mitigate arise primarily from human-managed systems. Agent-managed systems will create risks that we are not yet able to fully identify. As a result, our risk mitigation and safety measures are also not yet fully designed for agent-enabled systems.

24. The big question then is: to what extent can we push ahead with agent-dependent systems when the risk management measures and the safety guardrails haven't been designed? In that sense, there is some reassurance: for a long time, we are still going to have to depend on human-enabled systems.

25. Having said that, for your organisation to remain progressive and competitive, you cannot afford not to think about the usefulness of agents. My encouragement to you is that proofs of concept and small-scale projects are a good way to start, but you do need to move beyond them.

26. My only request is that you do so while empowering people, rather than replacing them. There is still so much we can do to unleash human potential and creativity, and it would be a loss to all of us if we believed that AI could simply take over everything humans do.

27. On that note, I hope this has been a useful presentation, and I wish you fruitful discussions for the rest of the day.

28.  Thank you.