部長何晶在2026年國際AI安全科學交流主會議上的開幕演講

Newsroom Opening Speech by Minister Josephine Teo at the International Scientific Exchange for AI Safety 2026 Main Conference Speeches Opening Speech by Minister Josephine Teo at the International Scientific Exchange for AI Safety 2026 Main Conference 18 May 2026

Good morning, everyone.

I am truly delighted to welcome you and want to thank you for being part of the International Scientific Exchange 2026. We deeply appreciate your commitment to AI Safety and meaningful dialogue. The fact that Singapore can be a host for such dialogue is an honour for us, and we do not take this privilege lightly.

In the year past, AI developments have continued to accelerate. For many more organisations and individuals, it is moving from occasional use to a shaper of everyday life. We can see and feel its influence on how we learn, how businesses operate, how governments deliver services, and even how citizens form opinions and make decisions.

There is enormous opportunity before us. Countries like Singapore that are resource-constrained can least afford to disregard AI. We need it for productivity and economic competitiveness, as much as to provide better healthcare and education. It can also amplify our ability to contribute to scientific progress and strengthen national resilience.

It is no surprise therefore that Singapore has outlined a clear ambition to become a leading AI hub — a place where top researchers, engineers, companies and investors come together to build useful and impactful AI systems.

We want AI to help solve real problems, such as improving healthcare, strengthening transport and logistics, boosting manufacturing capabilities, enhancing education and supporting areas such as climate resilience. Our National AI Strategy and national AI Missions – they reflect these aims.

Apart from seeking to become an AI hub, we also ask another very important question: what kind of AI hub do we aim to be? Because there are many ways to compete in AI. Some countries compete on market size. Others compete on compute power or access to capital.

Singapore will not, and cannot, try to outspend the world’s biggest spenders on AI. We do not have such deep pockets.

Therefore, we will focus on creating an environment of trust – where technology is deployed responsibly, risks are well understood, and effective protections thoughtfully implemented. That is how we are thinking about it.

This is where AI safety is essential. Too often, safety is framed as being in tension with innovation, as though more safety must equate to less innovation. This is not always a helpful way to think about the challenge.

For Singapore, safety is not viewed as a brake on innovation, but a vital aspect of our value proposition. We want businesses, public sector agencies, and citizens to integrate AI deeply into important systems.

But to do so, they must have the confidence that these systems are reliable, secure and properly governed. We believe countries and companies that can provide such confidence will have a better chance of sustaining AI innovation.

This matters especially because AI is moving beyond chatbots and search tools. Increasingly, even in Singapore, we see AI systems being used in areas that affect people’s lives directly: healthcare, finance, transport, public services, education, cybersecurity, and critical infrastructure.

In such areas, trust matters enormously.

A company deciding where to build an AI research centre should consider not only access to talent and infrastructure, but also whether the surrounding AI ecosystem is stable, trusted and well-governed.

Researchers deciding where to work should ask whether they are anchored in environments where bold innovations coexist with serious thinking about long-term societal impact.

Governments deciding who to partner with will increasingly look for those that combine capability with responsibility.

In other words, we believe that safety and innovation are becoming mutually reinforcing. A trusted AI ecosystem may ultimately become more attractive than a purely fast-moving one.

At the same time, we should also be clear-eyed about the risks. For an open and globally connected city like Singapore, the first major concern is trust. AI now makes it very easy to create convincing text, audio, images and videos at very low cost. This creates obvious risks around scams and fraud. We have already seen how online scams can become more sophisticated and more personalised. By now, this is so commonplace, that they are not even novel to talk about.

But the issue goes beyond scams. Over time, people may begin to question whether what they see online is real. Was that video genuine? Was that message written by a person? Was that online conversation authentic?

Once people start to lose confidence in what they see and hear, trust in institutions and in one another can easily weaken. This matters a great deal for a society like ours that depends heavily on trust: trust in transactions, trust in communications, trust in public institutions, and trust between communities.

The second concern we have is cybersecurity. AI is making cyberattacks easier to carry out and harder to detect. Criminals can now produce more convincing phishing emails, impersonate voices, and automate attacks at a much larger scale. Mythos is just one way in which we see this being presented so prominently.

At the same time, countries are becoming more digital and more connected. Inevitably, policymakers must worry about the resilience of essential systems: banks, telecommunications, transport, healthcare, and public services.

In fact, two weeks ago in Parliament, we answered the question on the impact of Mythos and what we are doing better to secure our critical infrastructure. These questions are only the beginning. As Parliamentarians acquire the language to ask deeper questions, I expect that the sophistication and the angles from which they will pursue them will only sharpen further.

As AI capabilities increase, cyberattacks may just be the tip of the iceberg. Experts are surfacing other security concerns like loss of control. There are already studies that suggest adversarial behaviour by AI Agents that seek to proliferate their existence at the expense of everything else that we care about, such as safety.

The third concern is the impact on jobs and skills. AI will create many opportunities, but it will also change work very quickly. Some tasks may disappear – we know that’s a fact. Others may be heavily automated. This creates understandable anxiety, especially among professionals and mid-career workers. People are asking: will my skills still be relevant? How do I adapt? Will opportunities remain open to younger workers entering the workforce? These are not just economic questions. They are social ones as well.

The fourth concern is over-dependence. Today, much of the world’s advanced AI capability is concentrated in a small number of companies and countries. This creates a new kind of dependence. Countries may become reliant on systems that they do not control and cannot fully inspect. When something goes wrong — whether a system failure, harmful content, or a major security incident — policymakers may have limited visibility of how those systems actually work, or how access and usage rights were even decided.

So, the question is: what can countries like Singapore do?

The first, we believe, is to build resilience, not fear. Singapore has generally taken a practical approach to technology. We do not reject new technologies simply because they carry risks. But neither do we assume that markets alone can be counted on to solve every problem.

We adopt the same balanced approach to AI. For example, Singapore has already introduced initiatives such as AI Verify to encourage responsible AI testing and transparency.

We are launching a new programme to accredit third party AI testers, to give deployers added assurance. We want deployers to go for proper testing, but they naturally ask the question: who can I trust as a tester?

And so, the scheme to accredit third-party AI testers, who are not the government, is intended to set standards and to uphold a certain level of confidence in the testers themselves.

We have also introduced guidance around Generative AI governance and broader efforts around online safety, scams, and digital trust.

Rather than being anti-technology, these efforts require a joint focus on innovation and accountability. Our aim is to ultimately support adoption by increasing confidence. This distinction is important.

Here, I want to especially call out the effort by Concordia, our partner in organising ISE, for doing a great job of describing the state of AI Safety in Singapore. Thank you, Brian, for letting me have a copy of the report – I’m sure he is making available more copies of this.

I’ve had the opportunity, in the last few hours since you gave it to me, to scan it. I would have to say a thank you for calling out the areas where we have made progress, but also highlighting the significant gaps, particularly around security by design. This is why we are very interested in continuing this conversation on AI Safety research priorities.

The second thing that we believe countries like Singapore can do, is to continue to invest in people. We are committed to ensuring that AI does not result in jobless growth. We will help our people stay relevant in the changing workforce and equip them to use AI safely. We do not think AI safety is the sole concern of engineers and computer scientists alone. We believe that it is also the concern of teachers, psychologists, public servants, cybersecurity experts, business leaders and even parents.

We believe a society that understands AI risks broadly and deeply will cope much better than one that does not. Digital literacy therefore becomes increasingly critical: how to verify information, how to recognise scams, how to question suspicious content, and how to use AI responsibly rather than blindly relying on it.

Third, governments need to continue building internal expertise. Public institutions need enough technical understanding to ask the right questions. If we do not, we are sunk – because we won’t even be able to assess for ourselves whether the risk levels that we are taking are acceptable. We need to evaluate risks and make sensible decisions. Governments may otherwise regulate too slowly or react in ways that are not effective.

Fourth, we will seek out like-minded countries and work with them on common challenges. Many AI risks will not respect national boundaries. Smaller countries may not individually shape global AI development, but together they can influence international norms, standards and expectations.

Now let me turn to what we ask of scientists, researchers and technology companies.

First, we ask for responsibility. The companies building the most powerful AI systems should recognise that they are no longer creating ordinary consumer products. These technologies may affect entire societies. Safety therefore cannot be treated as an afterthought.

Second, we ask for transparency. Governments and citizens need clear understanding of: what AI systems can do, what their limitations are, and what safeguards are in place. Trust becomes difficult when systems are opaque and unexplainable.

Third, we ask researchers to focus not only on making AI more powerful and capable, but also on making it more reliable and controllable. We need AI systems that are easier to test, easier to monitor, and less likely to produce harmful or misleading outcomes. Today’s discussions will be important in identifying critical areas for more attention and investment. We are encouraged by offers of funding to support the global research projects prioritised through the Consensus.

Earlier this morning, I met with Max Tegmark, and we were discussing how unexpectedly the work that begun 12 months ago has proven very timely. As the momentum has begun to shift, and interest has also begun to pivot, we are well placed to ride this change and hopefully to steer AI in a more safe direction.

We hope that these global research projects that are being prioritised and carried forward into action, will be a tangible way to move us from dialogue to action.

Fourth, we ask the global AI community to broaden participation. The countries shaping frontier AI today may be few, but the societies affected by AI will be many. Smaller countries should not merely be passive consumers of technologies developed elsewhere. Our experiences, constraints and perspectives matter too. I am glad that the UNDP is here, working with the Singapore AI Safety Hub and IMDA to connect today’s discussions with the UN Global Dialogue on AI Governance.

And finally, we ask for partnership. Countries like Singapore may not control the frontiers of AI development – we may never get that. But we can contribute meaningfully to areas such as trusted deployment, governance, cybersecurity, public-sector implementation, and real-world testing in complex urban environments. These are valuable capabilities in the AI age.

Singapore’s ambition is therefore not simply to become an AI hub, but one that is trusted to use it with care.

We aim to be home to talent who want to innovate quickly and responsibly, where impactful AI systems are tested and deployed to improve people’s lives and uphold public confidence.

We hope that you will continue to work with us to grow a hub where AI serves the public good, for Singapore and the world.

Thank you once again for being here.