Pablo Hernández de Cos
0:51 ↗
Good afternoon. Good afternoon, everyone. On behalf of the Group of Thirty, I'm very pleased to welcome all of you to our 41st annual International Banking Seminar. I thank Governor Vitai Ratna Korn and the Bank of Thailand for making this event possible, and also the IMF Managing Director Cristina Giorgva and her wonderful staff, who are always cooperative in helping us do these events.
As you may know, the Group of Thirty was established in 1978. It's an independent global body consisting of people from both the private sector as well as the public sector, as well as people from academia. What we aim to do is expand understanding of global economic and financial issues and to explore the international repercussions of decisions that are taken in both the public and private spheres. I believe our role is extremely important in these very difficult times, when we can help foster mutual understanding and dialogue between countries.
Now, we were last here in Bangkok 35 years ago, and since that time, Thailand has transformed its economy and positioned itself as a leading voice in the region. Of course, the global economy has also transformed itself over this period, though not always in desirable ways. Even as emerging markets have learned the virtues of openness and macro stability, developed countries seem to be forgetting them. Political and macroeconomic dysfunctionality is no longer just a preserve of emerging markets and developing countries; it spread elsewhere also.
Nevertheless, the global economy has been surprisingly resilient, and growth is strong — well, strong enough given the shocks the global economy has faced — but as always, it is unevenly distributed. We cannot, as you've seen in these meetings, afford to be complacent. The protests of youth across the world suggest growing worries about the future, even as cracks build below the surface.
Now we have a lot of policymakers in this room, and you know where the cracks lie. They include two wars, rising energy prices and a likely shortage of refined products even as winter looms in the northern hemisphere; an unprecedented El Niño which is affecting agriculture everywhere; a great dependency on a new technology both directly and indirectly through credit and equity prices; and cyber risks, which our host Governor Ratana of the Bank of Thailand will touch upon. And of course, behind all this, an opaque non-bank financial system facing a sharp rise in interest rates, which our panel on balancing innovation and stability in modern digital finance will no doubt address.
Unfortunately, with very high sovereign debt burdens everywhere, we have little room for mistakes. We will need all the mutual international support we can summon to see us through the next few years. Of course, current geopolitical differences will make any cooperation hard. We know that to obtain the limitless possibilities suggested by AI, we first require cooperation to mitigate its existential risks. So we have a full agenda today. We'll talk about all this.
Our opening talk is by our host, Governor Ratna Korn, followed by the panel discussion, which will be moderated by Pablo Hernández de Cos, who assumed the role of General Manager of the BIS on 1st July 2025. He is, I should note, also a member of the Group of Thirty. Please note that today's seminar is an invitation-only event, but it is on the record. Members of the press are present and will be watching online. With that, Governor Ratnakorn, the floor is yours.
Chair Raghuram Rajan, Managing Director Christina, Excellencies, distinguished guests, good afternoon. It is an honor to open the 44th annual Group of Thirty International Banking Seminar. I would like to thank the Group for convening this discussion here in Bangkok.
The G30 background notes highlight increasingly interconnected risks from high public debt, geopolitical fragmentation, non-bank finance, and technology dependencies. This provides an important backdrop for today's discussions: how AI and digital finance are changing financial markets, and what the change means for central banks and financial stability, and where stronger international cooperation is needed.
Balancing financial innovation and stability goes to the heart of modern central banking across economies. Digital finance is expanding access and changing how people use financial services. Today, nearly 8 in 10 adults worldwide have a financial account, up from around half in 2011, with mobile access driving much of this progress.
The next wave of innovations, particularly AI, offers opportunities that go well beyond making existing financial services faster or cheaper. It can raise productivity by automating complex tasks, improving decision-making, and enabling new products and business models. In finance, it can improve credit assessment, including for firms and individuals with limited credit history, helping channel capacity towards productive sectors.
Combined with tokenization, AI could also automate financial processes and make cross-border payments and securities settlement more efficient. Thailand's digital payment infrastructure can enable such innovations. Ten years ago, we introduced PromptPay, our instant payment system. Today, it has more than 83 million registered users.
In 2025, transfers through the system reached 54 trillion baht, more than twice the size of Thailand's GDP. That is for free transfers up to two million baht. PromptPay has transformed retail payment, extending digital financial services to people in even the remotest areas and to the smallest merchants.
However, technology brings new challenges. Greater speed, scale, and connectivity enable fraud to spread faster, while cyber and operational disruptions can affect multiple institutions and jurisdictions. We have seen a surge in both fraud and cyber events in the financial sector over the last decade.
AI and tokenization also pose risks — some old, some new — to the soundness of money and finance. Our responsibility as regulators is to enable innovation while putting in place safeguards that protect consumers, preserve inclusion, and maintain financial stability and public trust.
At the Bank of Thailand, we aim to achieve this through our responsible innovation framework, building on three pillars: smart technology, sound governance, and sound economics. The pillars are mutually reinforcing. Sound governance ensures that good technology is not abused and provides the clarity needed for businesses to invest.
Common standards prevent fragmentation, leading to stronger network effects and commercial opportunities, and profitable enterprises support investment in new technology that engenders better service. The upshot of responsible innovation is the delivery of financial services that are safe, inclusive, and resilient.
In this respect, an acute problem that we are currently facing is digital fraud. It is spreading rapidly, it is generating huge losses, and it is global. As digital fraud operates across borders, it requires global solutions; otherwise, activity will simply migrate to the weakest part of the system.
That's why the G30, together with the Bank of Thailand, launched the Bangkok Blueprint for Resilient Financial Services just a few days ago. It was an important paper for us, and I would like to thank IMF Managing Director Kristalina again for her strong support for this initiative, and colleagues from BIS institutions for turning our shared ambition into a practical framework.
The blueprint sets out 12 key considerations across five pillars, covering prevention, detection, response, information sharing, and accountability. It provides a reference that countries can adapt to their own circumstances, while promoting greater consistency and cooperation across borders.
Clear responsibility, effective information sharing, and appropriate incentives can support coordinated implementation at scale. Putting this into practice is not straightforward, especially as technologies, business models, and risks continue to evolve. I look forward to hearing the panel's perspective on these challenges, and more broadly on the implications of AI and financial innovation for central banks and financial stability. With that, let me turn to the panel. Thank you.
Well, good afternoon, everybody. Let me start by thanking the Thai authorities for the excellent organization of this meeting and the very warm hospitality. Thank you very much, Governor.
One of the main topics that we've been discussing through the whole week is related to innovation, technology, and in particular AI. We've been emphasizing how important AI has been for the resilience of the economy. The global economy has been subject to many different shocks, negative shocks, and AI has been one of the positive shocks that have been able to compensate for some of those negative impacts coming, in particular, from negative supply shocks.
We've been discussing as well the implications of AI on inflation, both in the short run, with in particular the effects that it is having on demand, but of course also on the potential consequences in the long run for some key variables for monetary policy, in particular for R-star, and also by the way on G-star.
And for sure, we've been discussing this week the potential consequences of AI for financial stability. And here, for example, our meetings of the FSB, including the one that we had this week, also put a lot of focus on this topic. So we wanted to wrap up, to use this panel as a kind of wrap-up discussion of the whole week.
And for this, we have the best panelists possible. They don't need much of an introduction, but still, let me say that we have the pleasure to have Governor Bailey of the Bank of England, Governor Cardoso of the Central Bank of Nigeria, Dr. Georgieva, Managing Director of the IMF, and Huh Shin, Governor of the Central Bank of Korea. I'm still tempted to say BIS, but no, it's the Central Bank of Korea.
So we've organized the discussion, starting first with the implications for monetary policy. Then we will move to financial stability, and then finally we will also focus on what is the role of international cooperation to face the challenges that surely they are going to emphasize in the first two parts of the discussion. So let's move directly, Andrew, to the implications of technology more in general on monetary policy.
Well, thanks, Pablo, and can I add my great thanks to the team from Thailand, Governor Ratan, and all his colleagues. We've really had excellent meetings and an excellent time here. It's been a real pleasure. So yes, AI and technology and monetary policy. I want to start the sort of inner economic historian in me. It doesn't take long for it to come out.
You know, I do think that the things over history that have actually raised, you know, potential growth rates over a sustained period are, you know, big changes, particularly in general purpose technology and the effects that that can have in terms of spreading across the economy. And I think it is possible that one of the causes of us having lower potential growth rates in many of our economies for the last 15, 20 years is actually, you know, that we may have been between a period of general purpose technology innovation, with the previous one, sort of ICT and the internet, having sort of, you know, had its effects on growth rates, and then, you know, it's leveled off.
And so you ask the question, what's the next one? Well, I think it's most likely some combination of AI and robotics. And if you then, as you said, in the context of monetary policy, then if you think, well, how do we go about thinking about this? I mean, if it's going to raise productivity levels and raise potential growth rates, I think the first thing to say is we're going to get demand effects and supply effects. And the question is, we do have to disentangle those two. And we have to disentangle them both in terms of their scale, but also their sequencing as well.
And there's an excellent paper by Sana and some colleagues at the Bank of England that came out a month ago, which I think does just that, and I strongly recommend it as a sort of framework for thinking about these things. I think at the moment, and of course it does differ from economy to economy, because clearly, you know, the US is in a different position in that respect, but I think we are seeing, you know, demand effects probably coming through more quickly, particularly via asset markets and asset valuations.
And so I think that's probably more so in the US than the rest of us, but I think that is the mechanism that's coming through probably sooner. I think that, and obviously we're seeing quite strong asset valuations in that particular sector, and that leads me to the question, well, you know, what could go wrong, and what could change those effects and then feed through into those demand effects? And this relates to the supply effects as well.
And I think four things that I have in my mind. So the first one is that obviously the demand effects, insofar as they come via asset prices, rest on markets forming a view on what the eventual productivity effects are going to be and what the effects on potential growth are going to be. Clearly, markets could suffer a correction if they change their view on what those are. Obviously, it could go either way, by the way.
The second one comes back to my point about timing. I think again, if you look back at history, it's typically taken longer for these general purpose technology effects to come through than people assume they will. You know, history doesn't always repeat itself, but there's every reason to think that, you know, in the age we live in now, these timing effects will be faster.
But you go back, I mean, obviously you go back to Bob Solow's famous comment about you can see computers everywhere except in the productivity statistics. You go back to that, it was about, I don't know, 40 years from Edison wiring a light bulb up to it appearing in productivity numbers. So I think timing effects are important, and that's because a lot of the general purpose technology effects to come through require what I might call secondary innovation, if you like, to actually put it to work.
Third thing is, you know, you look at the markets, everybody is priced to be a winner at the moment. And of course, again, typically when you look back, everybody isn't a winner. You know, I always quote this point, remember Netscape? And, you know, ask where Netscape is today, it isn't. So, you know, everybody doesn't win, and of course there are companies we've not even heard of who will eventually come through. So, you know, you will get sort of what I call churn effects.
And then the fourth one is the one that for me is sort of newest in my thinking anyway, but I think is actually quite critical in the world we're in now. These asset valuations are now increasing, and obviously there's a lot of borrowing going on behind this to fund the capital investment going on. They are increasingly dependent upon public trust in the technology, and particularly public trust in AI. And let's face it, there are questions around that public trust.
I think, you know, if you — this is a bit, get a bit philosophical for a second — if you try to sort of strip that down, I actually think at root it's about, you know, how much human control do we have over this thing, and how much do we have relative to what we want to have. And again, I think if that public trust was undermined, and let's face it, you know, Chris and I were on a panel earlier this week, you were quoting the French refrigeration situation, there was a headline I was quoting that morning from Anthropic saying that, you know, they can see places where AI agents could worm their way into much greater parts of our lives, as it were.
You know, that public trust clearly is to a degree fragile, and that could affect, again, you know, the asset valuation demand effect channel. So those are things that I look at. I think, you know, demand effects, you know, not surprised, we're probably seeing them faster than supply effects. Say that's backed up by history, and there are some fragilities.
Thank you very much, Andrew. So, Governor, you want to continue?
Thank you very much, and again to add my voice to say a big thank you for the hospitality we've been given while here. It's been really outstanding, and also to thank Cristiana for all the good work. Thank you so much. Well, let me start off by saying that from my perspective, the impact in the emerging markets would be different from, you know, the more advanced economies, naturally.
And, you know, I was quite struck by the last session, the plenary, when we were hearing that really this could be an opportunity for either convergence or divergence. And this is exactly what you say, and I fully agree with that. This is a great opportunity for either convergence or divergence.
And I do think that it is vitally important that emerging economies ensure that they have an AI strategy. And obviously the importance of that would vary from country to country. But I do think it is vitally important that an AI strategy is put in place so that the opportunities that arise as time goes on do not elude us. So that's number one.
Now, in terms of impact generally and on monetary policy, well, one area that I would say concerns me is with respect to the trade-off between productivity and job creation, to the extent that we have a situation where roughly 60% of the population in Africa is said to be under 25.
And by 2035, it's estimated that there'll be more people going ready for the job market in Africa than the whole of the world put together. That obviously has its own implications in a situation where the deployment of AI is able to bring about productivity without creating jobs. So I think that is certainly one area that needs to be properly interrogated.
Of course, the models that will be adapted need to be models that suit the peculiar environment, because then central banks will increasingly be judged not just by money supply but also the credibility of data. The credibility of data. So it's very important that the data and the models that are being used and adopted are models that are good for the environments in which they are.
There could also be, in my view, in a number of the emerging economies, there could also be a break with respect to transmission, quite frankly, would be much more effective. Transmission will be much more effective because you find that in many situations the transmission mechanism takes a lot longer through the banking system than when you adopt AI, and that's something that would be positive in many respects.
Then particularly in a country like mine, Nigeria, where we have a very strong fintech culture and where the fintechs have largely done very well. They've created unicorns not just in Nigeria but across the African continent. There's a possibility of looking at, at some point in time, the CPI basket, because then digital services and the impact of AI can affect it in a positive way, as the case may be. So obviously the impact of the disinflationary process from using AI is something that would need to be looked at very carefully. Thank you very much.
Well, I would take it from where we just got. We need to differentiate between the short-term impact of AI on inflation. We take a position that our main concern about monetary policy is price stability, the contribution to price stability. Then our number one question would be, is AI inflationary or not? In the short run, there is no question it is inflationary. It is a positive demand shock.
We have seen that trade in goods now is more than 10% AI-related. The valuations everywhere, almost not only in the big AI-developing countries but in other countries as well, you see valuations going up. What happens as a result? People feel rich. When they feel rich, what do they do? They go out and shop. And when you have other factors impacting inflation, in this case energy price, food price, defense spending in some countries is also inflationary.
You have all this, and then on top of it you have a positive demand shock from AI that goes through two channels. One is the direct impact of higher demand for AI goods, and the second one is valuations, in the sense that we are wealthier. So let's go out and spend. So there is no question about this. Then the question is, where is this short-term inflationary impact concentrated? Mostly in countries that are integrated in the AI value chain. It's not everywhere.
So then the question, the second question is, what next? And there you have two schools of thought. One says it will continue to be inflationary. Why? Because it would have this wealth effect that would continue, or because we would see a very rapidly changing cost structure of goods, or because there would be so much demand for energy because of AI that it would push prices up.
So there are those that are in the inflationary school, they have their multiple explanations. And then there is a more popular, to be fair, deflationary school of thought that says, oh no, no, no, this is going to cut costs quite dramatically. We would have abundance of goods and services, and they all would be cheaper, thus deflationary.
At the Fund, we don't have a position now because we want to work harder on understanding all these dynamics. For example, the energy pressure from AI, how big it is going to be, what that would mean for energy prices, who is going to be affected. But what I can tell you is that the most significant deflationary impact may very well come from what we would be talking next about, which is a massive shock, negative shock, because of either financial stability or other forms of hit, and then we are all in a panic.
Nobody buys anything. We actually have no financial means to buy a thing because the whole financial system was wiped out by rogue agents. So there is that. But what I want to say here is that there is one aspect of this conversation that I want to put in front of us, even if we are not going to advance too much, and it is monetary-fiscal policy coordination. And I think we have not talked much about it, but it is going to be quite pressing to understand how this monetary-fiscal policy coordination would go.
In the meetings we just completed, one of the questions we were asking is what would happen to tax policy? Who do you tax? Who or what do you tax? What happens to public revenues? What are the consequences for public spending, and from there, what is the consequence for monetary policy and central banks? So to draw my line here, we are in a place where we have many more questions than we have answers. And my alert to everybody is, do not fall for the guys that pretend they have the answers. They are tricking you.
Thank you, Pablo. It's great to be here, and thank you, Governor Ratan, for the hospitality. It's my quiet period before our monetary policy meeting next week. So what I'd like to do actually is to be the bridge between the monetary policy discussion and the next round, which is on financial stability. So let me maybe kick that off.
Now, some of you were at Jackson Hole, and there was a very interesting discussion at Jackson Hole. Markus Brunnermeier gave a very interesting paper, and Raghuram Rajan was discussant. It was about AI and financial stability, what we can really expect from that, and it went over many of the pressing issues that are in front of us. Now, one of the characteristics of AI, of course, is that it's highly non-transparent. It's a black box, basically, but it's a very, very clever black box.
And how do we shape the financial system to meet the challenges that come from that? Now, Raghuram made the point that actually it may be not as bad as we may think. So we're not actually up against AI, you know, it's not humans versus AI, it's humans plus AI versus other humans plus AI. And that actually changes the game somewhat.
And the lack of transparency is not unique to AI. Lack of transparency is something that we're all familiar with from the series of financial crises that we've had. And as Raghuram points out, we have built up a whole set of tools to deal with that, and they are the classical ones: capital, restrictions on leverage, collateral, and very, very simple rules that are robust.
Now, but Raghuram did say something that I didn't quite agree with, so maybe we can discuss this afterwards. Raghuram said, 'Look, AI models themselves are very diverse, and that diversity gives us some protection.' I think that's true as a general statement, but for the financial market, that may not be true, and that's because financial market players are acting against various constraints, and the constraints may actually induce correlated behavior in a way that will amplify shocks.
Now, what I mean by this is prices themselves can become coordination mechanisms. So if, for example, we're all leveraged up, we're all levered players, then when the price falls, everyone — well, many become forced sellers. And when that's the case, prices fall further, and then people sell more, and suddenly, even though we all use different AI models, we're all suddenly coordinated.
And some of you may know the example of the Millennium Bridge. Millennium Bridge is a bridge in London. It's very close to where Andrew works. And it's a pedestrian bridge which opened, as the name suggests, during the Millennium. And at the opening, it was swaying from side to side with lots of people on the bridge. It was closed on the same day.
And what actually turned out was that the bridge was the coordination mechanism. If the bridge sways, everyone adjusts their stance, and that adjusting of the stance is done in a coordinated way in reaction to the bridge itself. So the bridge is the price mechanism, and the pedestrians are the financial institutions who are levered. And I think that problem, I think, is endemic. It's not something which is specific to AI. It is something that is specific to the financial market itself.
Now, how do you deal with that? Well, I think we have to go to where Raghuram was pointing, which is how do we ensure diversity in a world where inherently there are forces for conformity? And I think the answer are exactly these robust rules. So if these were indeed diverse players, then they wouldn't be forced sellers when the price falls.
So lower leverage, higher capital, greater safeguards would be conducive to greater diversity. And I think that's the other reason why I think these classical tools, I think, will be more important than ever. So I think, you know, it's the great unknown, but I think some of the basic rules will still be relevant. So why don't I hand over back to you, Pablo.
Very good. Thank you very much, Huh. Maybe we can also take, Andrew, the FSB perspective that you've been working on on this topic. Please.
Yeah, and I just wanted to come back actually to a point that Olayemi made and a point Kristalina made, which is this point about — I mean, you made the point about the labor market, and Kristalina made a point about fiscal policy. I don't — obviously I don't comment on fiscal policy as a central bank governor, but I think this is just about within my scope.
I mean, again, if you look back on history, you know, the impact of big general purpose technology changes on the labor market is actually not consistent. So, you know, the first one in Britain, the classic industrial revolution around the steam engine, actually reduced the labor share. So it substituted unskilled jobs for skilled jobs. The second one, which is sort of, you know, much more to do with electrification, if you think about Henry Ford and so on, actually increased the labor share because it created skilled jobs. So you can't quite generalize about this point.
I think it's absolutely right to say this is a critical issue, and I think the fiscal policy issue Kristalina is rightly pointing to does have a — you know, this question about what's going to be the labor share and what's going to be the profit share obviously has a bearing on that question about what the outcome's going to be.
So on financial stability, obviously there is a very important financial stability aspect to this. I mean, it has been — I can tell you, you know, we really sort of, particularly when Frontier AI first became aware of just what its potential was, back in —
April, when we were in Washington, when we had the last of these sessions actually — it was that week, I remember — and it has been a very big impact since then. I mean, we certainly in Britain have been working very closely with our banks on what the implications of this are for cyber risk, for instance. You know, how it's changed the whole risk. It's really pushed it up. It's already very high at the league table. It's pushed it even further up the league table. You know, I think when you talk to people like OpenAI and Anthropic, they will tell you they were surprised. I mean, they were surprised by the power of the recursive learning in these models. So, as they put it, you know, they draw these sort of curves that go exponential and they, you know, they say we've moved upwards in towards the vertical part of the exponential curve with these models. And you know, we were talking about Mythos back in April. Mythos is old technology now. So you know, so I mean, it is a major issue. I think so, what do we do? Well, I think there are two aspects to this. One of them is not by any means limited to financial stability, but I think it's the foundation that we have to depend on, and it is this: I think we need stronger model testing standards. But we have to define obviously what we're trying to get out of these model testing standards. And I think we need them both pre-release and post-release because the models move on post-release. I think model testing is a lot harder for open-weight models than it is for the sort of the Mythos-type models, and that we've got to tackle that question as well. And I think we do need — we've got the AI security institutes in the UK, which came out of the Bletchley Park conference a few years ago. It's a very, I think, a very good start. It's a very skilled organization, and I hope we can see these develop. And I think, you know, I mean, other countries have got them, and I hope this works because again, I think they have to work with the industry. The industry has to be heavily involved in these model testing standards. But going back to my point about public trust, I think you need, you know, things like the AISI in the UK to, in a sense, act not only as really sort of coordinators of standards, but also I think to give the public some confidence about how these things are being done. That's important. And I think that's more important. I mean, there's a lot of talk about regulation, but you've got to work out how and why with regulation. I think model testing is more important. Then we come into the financial sector, and again it's about how AI is going to be used in the financial sector, how the models are going to be used, and how we meet the cyber risk threat. In other words, you know, these models arm the offense; we've got to arm the defense as well. And we can do that. There's no question. I mean, there's a lot of work going on at the moment, but it's absolutely urgent and it's going to go on. One other issue I would also say that I think we've got to address with AI, and that is what I call the agentic issue, in the sense that you can create what I would call economic agents, but you can't create legal agents in — well, not most of our systems anyway. So the question of responsibility and therefore liability and therefore risk management is very, very relevant here, and I worry a bit that people could get misled about that.
Sure. Thank you very much. Again, I think there are peculiarities here. Without a doubt, increase in digitalization, AI brings better access to finance for emerging market economies, without a question of doubt. And in addition to that, of course, cross-border payments — the whole payment system is deepened and potentially strengthened. So I think that is a good thing. However, I think that for emerging economies, there's an aspect of this that really ties in with the extent of your macros. And let me even share an example that I went through about three years ago. We discovered — and this was at a time when our currency was under a significant amount of pressure, and I had just taken on the responsibility of the central bank governor — and we discovered that a lot of this pressure was coming from virtual assets, largely because there's a shortage of foreign exchange. And the premium between the official and the outside rate was about 60% at that time, and it was a no-brainer really for those who were operating in that sector, you know, to get access to foreign exchange. They basically went on digital currencies, and a lot of foreign exchange as a result — it accelerated the pace of devaluation. Now fast forward, we discovered that when we were able to undertake the critical reforms, close the gaps, and ensure that you had a more functioning market, that those inefficiencies began to disappear, with the result that it's not that that market doesn't exist, but the reason for having it centered around illiquidity of markets and all that went away. So I think it's critically important that for emerging economies, you've got to put your house in order and ensure that your monetary policies are targeted in such a way that you get stability in your currencies and these inefficiencies that attract this kind of behavior are reduced or eliminated. And as I said, certainly in our case, that has made a big difference towards managing the instability that we had prior to getting the macros in place.
I am all for strong fundamentals. I want to walk back to the question of impact of AI on labor share, and to be clear that I'm not making a prediction that the share would drop. We don't know. In fact, the current analysis we do shows surprisingly that the net impact on employment — this is a study in the United States — the net impact on employment is positive, but the structure of employment significantly changes, with a small segment of highly paid, highly skilled people who augment themselves with AI and are much more productive, a shrinking part of routine jobs, and expanding employment in services. Again, because people feel richer, they go out and spend to restaurants, to spas, to travel, and that increases demand for services. But these are lower-paid jobs. So what is going to happen in employment? As Andrew said, this is something we need to monitor carefully, and right now we don't quite know whether there would be more augmentation, where would be the creation of jobs we cannot even think of, like let's say a house manager on Mars if we get there. So we don't know, but the reason I'm bringing it up is because what we do know is that we need to figure out what kind of fiscal policy we would have to put in place, and we know that when it comes down to fiscal matters, they're incredibly politically sensitive. It's a very difficult arena. Now, forward to the question of financial stability, AI and financial stability. There are three parts of the answer. The first one is the risk of withdrawal of investment in AI that could lead to the economy slowing down or even going into recession, and the financial stability risks coming from it. This is a very real risk. We are seeing valuations up, and we are seeing an economy that is standing on one leg. What is driving growth is the AI value chain. So that is one risk, and I would claim that this is the kind of risk we understand well because we've kind of been there. We have experience. We know how to prevent. We know how to manage. The second kind of risk is one that we also know, but never in that magnitude, and it is cyber security risks. They are not new. We have been dealing with them for quite some time. But they may come on steroids, and that we don't quite know what to do. And I would put as a separate category, although it is cyber security related, catastrophic events. I'm not saying that is going to happen, but we need to think of rogue agents, whether they are with or without human help, that could completely destroy segments of the economy. Of course, we think of the financial sector first. But if we have a catastrophic event that is health events, that could lead to panic with all the implications from it. I would tell you my own journey on the topic of AI. I started with a great deal of enthusiasm. Why? Because we are so desperate for more productivity and higher growth. And here is our study at the Fund saying we can get half a percentage growth per year. What a beautiful thing. Then I got into all the stories about potential catastrophic risks. And so from here, from enthusiasm, I dropped down to saying, 'Oh my god, this is really scary.' And then I realized that — I mean, this is the phrase I'm now using — is love it, hate it, fear it. It is here. So I came to my current status, which is calm. It's here. Let's see how we can cope with it.
No, it's interesting, Kristalina, on this issue of the uncertainty of the impact of AI on the labor share. We are about to publish the BIS survey on household expectations related to the implications of AI, and I think it's very interesting the outcome. I think I can anticipate the headlines. It's an international survey, more than 20 countries. The first issue is that households are very positive on the impact of AI on growth, on demand. And well, probably this explains also the fact that they are investing in stock markets, in shares not related to AI, and of course they are already enjoying the wealth effects coming from this. The second is that they are also very positive on the capacity of AI to improve their own productivity. But at the same time, they also have concerns about its impact on their own employer. And by the way, there are also very interesting differences across groups. For example, young people and less educated people have more concerns than other groups on the impact of AI on employment. So the older generation is more comfortable with AI. Yes. Is it coming back to the discussion yesterday on intergenerational effects? Is it that they're more comfortable because they know what is coming, or because they have no idea what is coming? I don't know how star races work, but it's not very different to what you are observing in the labor market, because what is happening is that firing has not increased, at least in many, many countries, but entry has been affected, in particular of young people. So maybe they are just responding to what they observe in the labor market.
Okay. So let's move to what international cooperation can do on these challenges that you've been emphasizing. Maybe we can start again, Andrew, with your FSB hat this time. What are the priorities for you that...
Well, let's broaden it out a bit because international cooperation clearly does have an important role. First of all, obviously, this technology knows no borders, to state the obvious. Secondly, I mean, if we can point out a bit to digital technology generally, you know, we're all involved in this. I mean, there is a very large amount of work going on on payments, and this is a very important area. It's a very important area where I think we can exploit the benefits of digital technology. Let me give probably the leading example, which a lot of work by the FSB on, which is cross-border payments. I mean, I sometimes characterize this as the land that time forgot, in the sense that, you know, frankly, they've stayed quite well behind what I'd call the leading edge. They're quite slow. They're quite expensive and slow, and they are ripe, I think, for change. And again, to come back to my point about arming the defense and arming the offense — I mean, I'm not sure the analog actually quite works as well in this context, but it's one where we need to work with the financial institutions, and we are, to build the benefits of this technology into the payment system. And one very obvious area where the defense can work here is fraud. There is too high a level of fraud in the payments world, and the offense is getting too many tools in its hands to pull it off, and that's going to get worse. I mean, AI will be used. I mean, I'm afraid it's a fact of life. So we have an opportunity, I think. In the world we've been in with the old technology, frankly, we're trying to bolt things on to the existing systems to make them more robust. We're now in a world of designing, you know, using this technology to really change payment systems, to design new payment systems, to change the technological form of money, not the economic form of money, with things like tokenized deposits. And for me, one of the key things where we need — we've got international cooperation, it's essential and it's happening — is that we build, we use the technology and build the anti-fraud mechanisms in from the start, because I think it's absolutely essential. We've got the opportunity to do it. So we must do it.
Yeah, I couldn't agree more with what Andrew has said, really, and I think it's so important to do this early. So that the horse doesn't bolt away. And I think that is particularly so. It's so everywhere in the world, but I think it's particularly so in the emerging markets. And I believe that that engagement can be both at a supranational level — FATF, Basel Committee and the likes of that — and it can also be with the private sector, those who bring about these innovations, to ensure that there's symmetry with respect to standards and that also there's an opportunity to train and to build capacity before it's too late.
