The Gambling Regulatory Authority of Singapore is implementing a program called Digitalization Master Plan 2.0.
“examine how the regulator is operationalizing its Digitalization Master Plan 2.0”

26 Jan 2026 · 20m · Kiong Kiong, Estelle Quag · 17 quoted facts
The Gambling Regulatory Authority of Singapore's CIO argues that AI should speed up and support regulatory work, but should not make high-stakes decisions. He says technology is useful where an error does limited harm, such as compiling reports. Where personal data, privacy or serious consequences are involved, humans must keep the decision. Trust in how the agency protects data is the core concern.
The central tension is between boldness and caution. He says technologists need 'the guts to try' and warns against over-caution, since humans also make errors, perhaps around 20%, against a possible 5-10% for technology. Yet he sets firm limits on where automation is acceptable. Shrinking manpower pushes the agency toward technology, while data leaks and AI-created security gaps keep it careful.
The CIO says regulators hold a lot of data from citizens and regulated entities. A hack or leak would break the trust of those who supplied it. He adds that AI adoption can open new loopholes and gaps in systems and processes.
He wants separate groups, such as auditors and government offices, to counter-check each other so the day-to-day team does not mark its own work. He also suggests using technology to check other technology implementations.
He compares a possible 5-10% technology error rate with roughly 20% for humans. He argues that rejecting AI for being imperfect, and going back to old methods, can be the worse option.
Low-harm tasks like compiling reports into indexes can be automated more freely. Where even small errors cause serious harm, he says technology should supplement decisions rather than take them over.
He says a shrinking population limits staffing, so the agency must use technology to make processes less people-dependent. The agency does not set numeric KPIs for cross-agency collaboration, but it keeps asking whether it invests enough in technology.
Only for lower-risk work. CIO Kiong Kiong says AI can supplement decisions and present data, but should not take over decisions where errors would cause serious harm.
He points out that humans also make errors. He compares a possible 5-10% technology error rate with about 20% for humans, and says too much caution would shrink the agency's appetite for risk unnecessarily.
He did not describe specific data or models. He said AI is used loosely where privacy and security are not affected, and that in sensitive areas humans make the decisions while technology is only an assistant.
He said it uses AI tools to monitor social media and public feedback on its policies. This shows whether its messages are being misrepresented or poorly understood, so it can refine how it communicates.
Yes. He said there is a data sharing platform between all government agencies, and that sharing is encouraged so agencies can reach common goals. He did not give specific metrics for it.
Stated by the speakers. Each quote matches the transcript, which shows what was said, not that it is true.
The Gambling Regulatory Authority of Singapore is implementing a program called Digitalization Master Plan 2.0.
“examine how the regulator is operationalizing its Digitalization Master Plan 2.0”
Kiong Kiong says the Singapore government invested heavily in computerization in the 1990s, when he graduated as an engineer.
“when I graduated back in the 1990s, that is the high time where the Singapore government had involved quite a bit of investment into computerization”
Kiong Kiong says the 1990s computerization drive aimed to turn manual Singapore government processes into automated ones.
“making Singapore manual processes into more automated leveraging technologies in government services.”
Kiong Kiong says the Gambling Regulatory Authority of Singapore holds a lot of data from citizens and regulated entities, so a hack or leak is a major concern.
“we have a lot of data from them. And now what happens if this data got hacked, got leaked?”
Kiong Kiong says the spread of AI technologies has added loopholes and gaps into systems and processes.
“it actually added more loopholes, gaps into systems, into processes.”
Kiong Kiong says AI governance at the Gambling Regulatory Authority of Singapore needs independence, with auditors and government offices counter-checking each other.
“We have to also ensure independence. How are the auditors or the group of government's offices counter-check on each other.”
Kiong Kiong proposes using technology to check other technology implementations as part of AI assurance.
“how can we also use technology to counter-check on technology implementations”
Kiong Kiong illustrates that technology might have a 5-10% error rate compared with about 20% for humans.
“maybe 5 % or 10 % error rate, as compared to humans, you know, past about 20%, right?”
Kiong Kiong says processes where even small errors are unacceptable should not use technology for decisions.
“for certain issues that are so crucial that we cannot accept even a small degree of errors, now those processes we should not even consider about using technologies.”
Kiong Kiong gives compiling lists of reports into indexes as an example of a lower-risk task suitable for automation.
“For example, like compiling a list of reports into indexes.”
Kiong Kiong says the extent of harm an error could cause determines where the agency applies automation.
“now the end results of the errors, how extensive the harm this error will create.”
Kiong Kiong says technology can be applied more freely in processes that do not affect personal privacy or security.
“technology will be more applicable into processes that come as a risk, doesn't impact personal privacies, doesn't impact personal security.”
Kiong Kiong says technology plays a supporting role rather than recommending actions in anti-money laundering related analysis of new game types.
“technology in this kind of area is usually playing a supporting role rather than recommending actions.”
Kiong Kiong says the agency uses AI tools to monitor social landscapes and public discussion of its policies.
“to use AI tools to monitor the social landscapes, understand what in general, what people are talking about”
Kiong Kiong says Singapore has a data sharing platform linking all government agencies.
“we have a data sharing platform between all the government agencies.”
Kiong Kiong says the agency does not set numeric KPI targets for its cross-agency technology collaboration.
“We don't really go into that portions like KPI measurements or, you know, what is the target in terms of numeric, you know, numbers to achieve.”
Kiong Kiong says manpower constraints from a shrinking population are a key driver for the agency's use of technology.
“Our constraint on man powers is always a key for, you know, population are shrinking.”
The guest explains how Singapore's 1990s computerization drive shaped his career. It sets up his view of technology as an enabler of better public services.
“I started as a young electrical engineer.”
He names data breaches and loss of trust as the main fear, made worse by AI-created gaps.
“there is always this possibility of having a risk factor.”
He lays out the governance foundation for AI: standards, separation of duties and technology checking technology.
“I think the most important part of it is to have a structured governance system.”
He argues that organizations should weigh AI errors against human ones rather than retreat from AI.
“technology also is not 100%. There is always a degree of errors that will happen.”
Automation is limited by how much harm an error could cause. High-stakes decisions stay with people.
“We can use technologies to supplement decision making but not take over decision making”
For personal data and at-risk individuals, technology is only an assistant and not the decision maker.
“technology can only be used as an assistant tool, not as a main decision maker”
The agency uses AI tools to track how its policies are discussed and understood, which helps it refine communications.
“use AI tools to monitor the social landscapes, understand what in general, what people are talking about”
He describes the government data sharing platform and why shrinking manpower drives technology investment.
“we have to be based on technologies to upscale ourselves, to make our processes less people-dependent.”
Welcome to CIO Leadership Live ASEAN. I'm Estelle Quag, Editorial Director for CIO ASEAN. Today, I have the great pleasure of speaking with Lee Kyeom Kyeom, the CIO of Gambling Regulatory Authority of Singapore. Together, we will examine how the regulator is operationalizing its Digitalization Master Plan 2.0, such as embedding generative and analytical AI into risk-based supervision, the governance, controls and metrics that make those systems auditable and effective. Our intent is to surface practical lessons for public sector CIOs, from predictive analytics and automation of lower risk approvals, to the use of AI and data models for harm minimization and the cyber posture against tech-enabled illegality and the data protection demands of PDPA and ASEAN frameworks.
We'll also cover workforce enablement, operator telemetry, interagency collaboration with a focus on measurable outcomes, transparency and resilience in a fast-evolving regulatory landscape. Welcome, Kiong Kiong. Great pleasure having you today.
Thank you. And happy to join the interview sessions.
Happy New Year. New Year to you, Kiong Kiong. Please share with our CIO audience your background and how did you rise to this leadership position at GRA?
I started as a young electrical engineer. I graduated my background as engineering graduates. So when I graduated back in the 1990s, that is the high time where the Singapore government had involved quite a bit of investment into computerization, making Singapore manual processes into more automated leveraging technologies in government services. So I started that journey and I get a hands-on into studying diplomat work in application systems, infrastructure implementations. And I think I'm lucky. My bosses give me the opportunities to design, to conceptualize, and I get the hang of it.
And I enjoy doing the implementation and testing and adopting technologies to help our services improve, help our citizen lives improve into a more efficient and effective way of doing things. So given this opportunity, I guess I am doing well. So my bosses give me these opportunities and then that's how I rose into the position of doing organizations in applying technologies and making technologies as a quick enablement in terms of business, our operations itself, getting more efficient and make our citizen life more easier in terms of doing government services.
Well, you're so talented, Kiong Kiong. Thanks for sharing with us your stellar career journey. Now, what keeps you awake at night? Could you share with us the top challenges you face when driving digital innovation and transformation in your role?
When we talk about technologies, Apart from the advantage in terms of benefits that help us get our life better, getting our operation more efficient, there is always this possibility of having a risk factor. How does the risk affect us? And that's always in the height of our mind. When we deal with citizens or in deal with our regulated entities, we have a lot of data from them. And now what happens if this data got hacked, got leaked?
There will be a big issue for us. the trust factor, the people who bring data to us, we have not been doing a good job if we didn't protect it well. So here's something that, especially in the eye of AI world, where AI technologies have been greatly been adopted and encouraged to be adopted, changed our lives, changed our way of works, it actually added more loopholes, gaps into systems, into processes. If we don't pay attention, if we are not way ahead of what is evolving, don't understand the architecture, the depth of how the technology has come about, we will be in a very terrible position, not able to protect data, not able to protect the insights of processes, then we
would seem to be not doing our job well and cross the trust factor to the people who trust us in delivering services to them. And that's something that I believe all CIOs will make them stay away at night.
Agree with you, Kiam Kiong. What governance, assurance and auditability controls are you instituting as GRA integrates generative and analytical AI across all facets of your operations under the Digitalization Master Plan 2.0?
I think the most important part of it is to have a structured governance system. Understanding what is our cloud rule, what is the values, what is the asset that we have in place. then how are this technology helping us to make our life easier, to make our process more efficient? Then, what will be this particular technology opening up possible backdoors, possible vulnerabilities that we have a whole set of standards, processes, framework and how to protect them.
Having us to do the job is one group of expertise executing the day-to-day tasks. We have to also ensure independence. How are the auditors or the group of government's offices counter-check on each other. As we said just now, technology has evolved and AI is getting smarter every day. They are challenging even outperforming human beings in certain areas. And how can we also use technology to counter-check on technology implementations to strike fences? I think that's very important.
So can you share with us any leading indicators that shows you that application of AI have strengthened your risk-based decision rather than adding a lot of noise?
I think that is a very important and crucial matters that any of the CIO is always thinking. Now, the thing is that I always believe technology aids us in terms of processes, in terms of decision-making, but we need to ensure there is correct operation boundaries as we just spoke about how to ensure fairness, non-binary, ethical behaviors. Now, these are easily said and done. What is the balancing point? What is so-called something that is acceptable?
what is the balance point that's right between the pros and cons that would that would not cause us to withdraw unnecessary because it is important as a technologist we want to apply we must have the guts to try but yet what are the descendants parameters or armoured to be around to save us against all the dangers. So we need to make sure that there are certain expectations in terms of profits and loss. For example, we always said that in the past, when we put in human beings as a check and balance controls, and we say that check and balance control always is the eyesight to say that human beings are not 100%.
We will have a certain degree of human errors, right? That's true. And today we talk about technologies. and we acquire technology, we ask ourselves, technology also is not 100%. There is always a degree of errors that will happen. And therefore, we say that we don't trust the technologies. We want to go back to the old way of doing things. But we have to ask ourselves, if we say technology will have a negative, or it's not 100%, maybe 5 % or 10 % error rate, as compared to humans, you know, past about 20%, right?
So you're saying which one is better off. So we should not be so cautious and that's all our risk appetite will be reduced, right? We have to have the same considered way of thinking and make sure we are taking a balance so that we can accept certain degree of risk, right? So we will be retracted unnecessarily.
Yeah, I agree with you. Don't put in the analysis paralysis and then you never stop an emerging tech that's like AI, right? Precisely. Thank you so much for that. Now, GRA has substantially automated lower risk approvals. Where do you draw the line between human judgment and automation? And are there any post-implementation controls that help you guard against, like what you mentioned just now, complacency with automation?
I think as what I said just now, that it's very important for us to think through the matters very seriously. Now, for certain issues that are so crucial that we cannot accept even a small degree of errors, now those processes we should not even consider about using technologies. We can use technologies to supplement decision making but not take over decision making In order to provide data, in order to present data in a more readable format, it gets upgraded.
But for certain processes, we think that the risk factors is not that high. For example, like compiling a list of reports into indexes. Well, if something goes wrong with that, it's not going to be the justice. So again, the key point here is that now the end results of the errors, how extensive the harm this error will create. That will help us in decision in which area we can have more usage and we can then be more open to apply them.
But in areas that the risk factors is very high, even though small errors we make, then we will have to be very cautious.
I agree with you. Thank you so much. Now I know GRA is already using AI and data models to identify potential at-risk individuals early. What are the features and datasets that drive those signals and how do you preserve due process and privacy?
Again here is that, as I mentioned just now, the using of personal data is almost important to the government. We do not will use technology alone to look at data sets. I would say that technology will be more applicable into processes that come as a risk, doesn't impact personal privacies, doesn't impact personal security. Those areas, I think technology can be applied more loosely. But in areas of high sensitivity and high data security perspective, I thought that a certain degree of human decision making is very important.
And technology can only be used as an assistant tool, not as a main decision maker on true drive decision making.
Thank you. Now, as you collect and analyse operators' data on new BAT types, how are you aligning technical telemetry with anti-money laundering and counter-finance of terrorism risk models?
I would say that in this area, it's a very complex thing. thinking about the type of bad games that has come about. It may not be that straightforward as a decision-making as technology can apply because there are human factors considerations. There are also regulatory considerations. So I would say technology in this kind of area is usually playing a supporting role rather than recommending actions. I think it's too risky. So I would say the traditional way of data analytics, data visualizations, helping us to contextualize information so that it can ease our offices from the traditional information compilations, it starts in that way.
Not into the portions of making automatic decisions or even structural steps in deriving decisions.
Sure. So let's change gears a little bit. Now, Nation's Office of Drug and Crime notes that organized crimes use of generative AI, blockchain and data mining in illegal online gambling. What is GRA's cyber posture for detecting AI-fabricated content, URL churn and crypto laundered flows?
I think from our perspective here is that we are also like other organizations. we keep watch whether information around us has been mistreated or misrepresented. So I think generally, Dave's AI tools is a good tool for internet information monitoring and tracking of what our policies, our visions, or even what we intend to do that we miscoded or misrepresented. So it's important, that's why you say, to use AI tools to monitor the social landscapes, understand what in general, what people are talking about, especially, you know, if a new piece of news is being broadcast or telecast.
Because there are always services that use AI to help us to check and monitor what's the general feedback, the general feelings that the general public have on these policies, and even help us in terms of whether we have been explaining well or certain representations of information have not been presented to structurally to get the understanding, it can help us in terms of refining messages, refining our portion of how this information can be communicated to the public.
Thank you, Kem Kyo. Is GRA aligning your data, typologies and tooling with Asia-Pacific Group on Money Laundering Expectations and also the home team's agencies to share intelligence at machine speed?
I would say that information sharing under the Singapore government, we have data sharing in what infrastructure already developed in Ipromet. Sharing information within the government agencies is always encouraged and we have a data sharing platform between all the government agencies. So today, I would say the structure there do support our sharing of information and it's always encouraged to share among government agencies information so that we are trying to make it easy for government agencies to get access to data to achieve our common objective.
Do you guys have any joint matrix together that shows you tech collaboration among yourselves reducing regulatory blind spots?
I would say that all picks itself. We don't really go into that portions like KPI measurements or, you know, what is the target in terms of numeric, you know, numbers to achieve. But we always believe that we have to advance with technologies. We have to change our processes according to what technology advances, right? And then, of course, taking into consideration our budgets, right? Our constraint on man powers is always a key for, you know, population are shrinking.
we have to be based on technologies to upscale ourselves, to make our processes less people-dependent. So it's always on our watch to say that have we been investing enough in terms of technologies to make our job more efficient and more productive?
Kim Kiong, thank you so much for your clear and candid insights into how Gambling Regulatory Authority of Singapore is operationalising Digitalisation Master Plan 2.0 by bringing analytics, automation and AI into risk-based supervision while keeping governance and public trust at the core.
Thank you.
We've covered vast ground on predictive modelling, lower risk approval automation, AI-enabled harm minimisation, cyber posture against tech-enabled illegality and the data protection demands of PDPA and ASEAN frameworks, along with the collaboration with the home team ecosystem. For our CIO audience, this has been Estelle Quack, Editorial Director for CIO ASEAN, together with Kim Keong, the CIO of Gambling Regulatory Authority of Singapore. Thank you so much for joining us and see you again in our next episode.