Order. Order. Welcome to the Treasury Select Committee on Tuesday the 14th of July 2026. We're here today to consider the Bank of England's financial stability report. One of our regular sessions with the governor. And I'm delighted to welcome Andrew Bailey, the governor of the Bank of England, who's joined by Randall, from my left to right, Randall Crosner, who's an external member of the financial policy committee at the bank. And Nathaniel Benjamin, who is executive director for financial stability and strategy and risk at the Bank of England. And Steven Blythe, who's another external member of the financial policy committee at the bank. Governor, if I could just start with you, you told us when we spoke to you in January that the risks of financial stability had increased in 2025 for the obvious reasons, all the geopolitical events that we've all witnessed. Do you think they've increased further since then?
Yes, for a number of reasons actually, but I'll pick out two. One is obviously developments in the world economy and particularly developments in the Gulf and the Middle East conflict, which I don't think actually directly increases the sort of structural issues that we've identified in the past but obviously increases the risk of things crystallizing and also I think increases the impact when things do crystallize. So that's the first thing. The second thing, which I have written to you about in the last week, is I would say the big new development, which is the news on frontier AI and the cyber risk consequences of that. So I think unfortunately it's a sort of unambiguous yes to your question.
Well, we'll come to AI in more detail later. Now obviously the conflict in the Middle East, we keep having a ceasefire and then it's not a ceasefire. What's your prediction for how that's going to impact the long-term situation in the UK economy?
Well, if you'd asked me this question a week ago, I'd have probably given you a slightly different answer. I'd have said a week ago, look actually things, if you go back to the hearing we had on monetary policy a few weeks ago, I think I would have said, look, I think things are panning out at that stage probably at the lower end of the free scenarios that we had painted in the MPR. But I would have added the very strong caveat that it seems to me that the situation remained unstable, and the ceasefire was fragile. Sadly that's come to pass obviously. I think where I am now is that obviously that instability has come to pass. I don't feel qualified to judge how long this is going to go on for in terms of this resumption of hostilities and how it's going to end. But it underlines that this is going to be an unstable process for the foreseeable future. Now energy prices have come down. As of a week ago they were sitting a bit above where they were before the crisis. They're now somewhat further above but crude oil prices are well down on their peak. We are also seeing continued fairly soft evidence on the pass-through into UK prices, but we'll get another inflation number next week. The one other caveat I would add, which I don't think gets enough coverage, is that although crude oil prices have come down, the prices of refined products have not come down as much. So we're talking about things like gasoline and diesel, which of course are the end products. So those are the things that matter in terms of the end products, and one or two of the reasons for that in the global context are that there's a story around China and obviously a story around Russia. China has substantially reduced its imports of crude over the last couple of months. That's been interpreted as a sort of softening of the market and a reduced demand story. But they've also as far as we can tell reduced their exports of refined products substantially. So in a sense they haven't necessarily changed their balance and that is tightening the global market for refined products. And the other thing obviously, and this is good news in other respects, is that Russia's had a serious blow to its production capacity from what Ukraine has done. So I would make that caveat that we all spend our time quoting crude oil prices but you have to be a bit careful because I don't think refined product prices have come down as much. And one of the challenges is the damage to the infrastructure in the Middle East, which can take eight months minimum to get things back on track once there is a ceasefire that holds. What modeling is the bank doing to look at the impact on the UK economy in the short to medium term? I mean, we obviously talk a lot to people in the industry and to our counterparts in the Gulf area. I spend a lot of time talking to my counterparts. I think the story there is to take oil and gas separately. On the front, there hasn't been as much physical damage to the supply production infrastructure. Obviously the blockages in the Gulf are another matter. So I think the message we get from the Gulf States is that if they can ship the stuff out then they will resume production very rapidly, and they had resumed production rapidly up until last week. They also have got quite big plans for building new pipelines to get round the Strait of Hormuz to ports which are south of the strait, but that will take some time. On gas, the story we get from the Qataris, who are the people who matter in that market, is that about 83% of the supply is intact. The 17% that was hit by the Iranians in the attack some months ago will take longer to bring back. They too have big plans for increasing output of gas, but they're talking about 2029-2030 for that. So they are clearly very determined to get supply back onto the market, but obviously what we're seeing in the strait now is going to impact that.
Okay. So it's a long rocky period ahead. I'll come to some other witnesses on that in a moment, but we're about to have our seventh prime minister in 10 years. We've had a lot of change at the top there. So you know, we're talking about internationally, but domestically though, do you think those changes and all that upheaval is good for financial stability?
Well, I mean I suppose we are about the stability point.
No, but two things which are important. One, as the last few days have demonstrated, the big driver of moves in financial markets is actually not what happens in the UK. It's global. It's the global setting. What I would say however is that the evidence does indicate that the UK situation is supported by on our side monetary stability and financial stability, and on the other side the fiscal framework. And you did see a bit of movement in markets briefly on the fiscal framework but that came back quite quickly.
Or their best chance of maintaining financial stability in the UK?
Well, I think the overall message I would give is that the big issue is growth in the economy. So we can do all we can on financial stability. I do actually think there are signs of a very resilient financial system. We described them in the Financial Stability Report in two respects. One is that the core banking system is resilient. But also when you look at debt levels in both the household sector and the corporate sector, they are not stretched at the moment. So that gives us an underlying picture of resilience. What I think is much more challenging is that we've had low growth in the economy for the best part of probably 16 or 17 years. This is not a story about any one government. This is a general story. It's not a political story in that sense, but it is a story about UK growth. I've made a number of speeches on this subject. I think it's important because it's the critical structural issue. And as far as I will go into fiscal policy, what I have said is that if we did a simple naive counterfactual and said imagine that we'd had the growth rate for the last 17 years that we had for the preceding period, which was a potential and actual growth rate of around 2.5% a year, and we've been around 1.5% for the last 17, if you add that 1% back in with the same shocks, same fiscal policy for every government, the same shocks at COVID, Brexit, Ukraine, Gulf, where would we be today on the public sector debt level? It would be about 10 percentage points lower. So you haven't solved all problems but I think it would be a very different debate.
Okay, that's very helpful. And one of the things, there was a strong drive here in the UK for deregulation of the banking sector to promote growth. And when we meet with those financial institutions, they're keen to see that and constantly stress that we're a long way from 2008. Do you think that that emphasis on deregulation and growth poses a risk for financial stability because sometimes that's the tension that's put to us by other people that the bank is holding back growth because you don't want to see that deregulation. Is that a fair comment?
My overall view on this is we will not get growth if we don't have financial stability. I think we can point to episodes in history which support that view. So financial stability and monetary stability are key underpinnings of growth. Does that mean that every one of our rules in financial regulation is perfectly formed? No. Would it ever be the case? No, because the world always moves on. So do we support a process of review? Yes. I do think that where I differ from some of the arguments that are put is, let's take the banking system. The key to the banking system supporting the economy is to have a system that is earning returns which cover its cost of capital, so that they can invest their returns to support lending into the economy. I think after a long period of rebuilding capital in the banking system post the financial crisis, we are now in that situation. They earn more than their return on capital, and you can see that in the market numbers, in the price-to-book ratios. Therefore, they have a choice as to whether they reinvest earnings into their own business or pay them back to shareholders, and that choice should be made in the market. The reason I push back is that some of the debates I find suggest that we have a fixed amount of bank capital, so we must lighten regulation to allow it to be spread around further. It isn't fixed. A healthy banking system will earn its returns and they can be reinvested into the economy.
Okay. Well, they've heard that. Mr. Crosner, from your perspective, let's look at the Fed. Does a new Fed mean what's it mean really for financial stability globally and in the UK? Do you think there's going to be any changes as a result?
So Kevin Warsh and I went through our confirmation hearing together to join the Fed, and we sat next to each other for years during the global financial crisis. So he has been battle tested in financial stability and he's very aware of those issues. No one can guarantee that there will not be a financial crisis, but he is aware of and sensitive to those issues. There's a lot of change in financial regulation in the US and a new proposal related to the Basel III endgame which I think is providing much more effective regulation. It's very important not to confuse effective regulation with deregulation. We want to make sure that regulations are effective in achieving financial stability goals, but also that banks can intermediate. I think that's the key motivation for what's going on at the Fed now. And from a macro perspective, Kevin is going to be very much aware of those issues and will want to avoid a financial crisis during his time at the Fed. But sometimes events overtake what one wants. When I was at the Fed, I had studied a lot about the 1930s. Ben Bernanke had said a lot about the 1930s. I don't think any of us thought we would be in a 1930s-like crisis, but having studied that period, that was very helpful. Kevin as well as others who have been through that period will make sure we are best prepared to respond to minimize damage if there is some sort of shock that causes a crisis.
Thank you. That's very helpful from someone who's so close to Mr. Warsh. Jim Dixon.
Thank you, Dame Meg. Just looking at other risks that we're facing at the moment. We're just in the middle of the third wave of extreme weather of this year. So clearly climate change is front and center for all of us in thinking about how the country and the world is developing. Would you say that climate change is now a bigger risk than you'd previously estimated? I think when the governor came to us previously, you said it was one of a number of risks, but it had slightly receded in terms of the extent to which it was one of our primary risks. Would you say the recent focus on climate change has changed your view on that at all?
Well, I was making a relative point. I wasn't making a point about the absolute level of climate risk. I was making a relative point going back to the chair's first question. Given what we've seen on frontier AI over the last few months, I have to make that point again. I'm not a climate scientist myself, so I'm not going to venture an opinion on what we can conclude from what we've seen in the last few weeks, but Europe has had a very strong period of very hot weather. So I think it reminds us that there are very clear risks from climate.
And either the governor or other members of the panel, does the climate change we're seeing impact financial institutions in a way we should be taking notice of? Would anyone want to, Mr. Benjamin?
So the channels through which climate change affects financial stability haven't changed. They remain very present. There are risks to banks, to insurers, and to core markets from climate change. This hasn't changed. Part of our analysis is to gradually embed this analysis into the mainstream of how we test financial stability. We've still got strong teams of staff at the bank looking at those issues and understanding the analysis you're talking about.
Because I think the governor last time we spoke, you accepted there were now fewer staff working on those issues, and that might indicate a lesser emphasis on them, which was a bit concerning.
No, again, I make this point. We have to allocate our resources to reflect the relative risks. For instance, AI has come up the lead table very rapidly.
We'll come on to AI momentarily. Katherine West,
Thank you very much. Meg, on the point of AI, what assessment has the Bank of England done on the needs around water, particularly internationally, thinking countries like Malaysia where it's such a big part of the economy now but the physical needs around water that if not handled correctly could have an impact on economic stability? What assessments?
I'm not sure we're the experts on that one, to be honest.
There is one aspect. Thank you. In your organization. I'm not an expert either, but there is one aspect related to AI in which the availability of water supply does affect those developments. We've all heard about the very large increases in the infrastructure of data centers that are being planned to develop AI capabilities. Those large data centers will require water supply to cool down the production line. The needs for water are large, and one of the questions is whether the water infrastructure will be sufficient to meet the needs for those developments at the pace that is currently embedded in the expectations of the markets.
What assessment has the Bank of England done on the demand on the grid for AI in the UK and what impact that might have on financial stability if certain projects don't go ahead but AI is demanding that?
Yeah, can I make two points? I think it comes back to this question of growth. I do subscribe to the view that technology is a driver of growth. I made this point about lower growth over the last 17 years. I think technology is part of this. In the long run, technology is one of the drivers of growth. The next cycle is likely to be some combination of AI and robotics. So this question about grid capacity and data center capacity is relevant in terms of supporting growth. The other point, I have to be honest, I hadn't appreciated until going around the country and talking to people closer to this, that there is also a question of access to clean power supply. I'm using 'clean' in a slightly different context to the environmental point. It's the quality of the power supply, having a very high quality and continuous power supply. Some people have made the point that renewables on their own don't give that quality. They can be a big part of it, but they don't completely solve that problem. You need a base load of other power to ensure you have that clean power supply. So in thinking about data centers, there needs to be thinking about how you ensure that mix of power supply.
Thank you very much. John Glenn MP.
Yes. Can I turn to the decision to ease the leverage ratio by 0.2 percentage points? I want to understand the range of views that exist on this. I understand there appears to be a split decision, and the quote is 'most judged that the benefits of the proposed changes outweigh the costs.' It would be useful to understand what the range of views were and how meaningful that range is in terms of decisions. Let me start with you, governor.
Yes, I'm happy to do that. Just to put into context, the announcements we've made on the leverage ratio are really about cleaning up some issues around the leverage ratio. There was a particular issue where we'd made some changes back in 2019 which created an anomaly that was particularly affecting the large domestic banks. That's important because they are the ones who make the highest proportion of loans into the UK economy, and it had left them facing a higher leverage requirement than was justified. So cleaning that up was important. We've also increased the buffer usability in the leverage ratio, which is important because it's consistent with how we operate the risk-weighted ratio. The risk-weighted regime and the leverage regime are not additive; they are parallel. We've made some changes to the risk-weighted regime when we talked about them in December. These are changes in the parallel regime. The challenge we've got is that we have an issue with leverage in another context: leverage in financial markets. We've talked about this in previous hearings, particularly in government debt markets. What we've seen in recent months is also a growth in leverage in equity market leverage, in hedge fund leverage, and in exchange traded funds. That leverage is different but connected because the banking system is the ultimate provider of liquidity. The question we debated was whether it's better to tackle market leverage in the market regulation world or in the banking regulation world. My own view is that it's better to do it in the market world, but we've got to land that. We've put out some ideas and got mixed responses. We've included language in the announcement that means we will do further work before finalizing these changes, particularly to see whether they might increase market leverage risk. So I think these are sensible changes, but we have an outstanding question about what to do about market leverage.
Okay. So who can help me understand what a slightly complementary view to that is amongst the this morning?
So the aim of the change to the leverage ratio framework is to fix this idiosyncrasy that was a result of an increase in the countercyclical buffer in 2019. There was an offset in the risk-weighted stack of capital, but an increase in the countercyclical leverage buffer without an offset. That discrepancy disproportionately affected the domestically systemically important banks. There was unanimity on the committee that that was something to fix. There's also a broad view that we want to get the overall balance of risk-weighted and leverage capital requirements correct. The other view is that in addition to fixing the discrepancy for the D-SIBs, which makes sense and will support lending, there's a slight reduction in leverage ratio requirements for some other institutions that could potentially lead to further provision of leverage into markets. To the governor's point, I agree that there are mechanisms within markets, particularly haircuts or margining on repo, that would be very effective for providing stability. The leverage ratio won't fix that if you have no haircuts on a repo transaction. That's arbitrary leverage. So that was the balance in my view.
Yes, thank you. We had an excellent discussion that got us to a very good outcome. Exactly as the governor and Steven have said, the key motivation was to make the leverage ratio more effective and ensure it prevents excess leverage in the banking system. I also believe this is not the right tool to deal with market leverage. Things like haircuts or centrally cleared transactions are the best way to deal with that. What we are doing is trying to make sure that we have things that can be released in a crisis. A 20 basis point change in the leverage ratio does not seem material to me in changing the ability of banks to provide leverage. We have much better targeted instruments. Making the capital stack more releasable is helpful for financial stability and for supporting credit provision and growth. It will allow banks to reduce the management buffer they hold. When there is predictability about what is releasable, they hold less. The approach we took will help reduce that, making the leverage ratio more effective without affecting financial stability and not addressing market risk issues, which we'll do through separate means.
Can I just come on to the decision about lowering the capital requirement from 14% of risk-weighted assets to 13%? Because I think there are two broad sets of views. There are those who think the benefit goes to the banks and their shareholders or to the economy as a whole. But there's also significant dissent among banks over the actual net effect because of so many caveats that the notional 13% is never met, and people say if you look at the bolder measures in the US, there's still quite a margin between where we are and what is happening in other environments. How do you respond to those observations?
Well, I would disagree that the US has taken bolder measures. I don't think they have. The latest package they announced is to implement Basel 3.1, which we are doing as well. We've published comparisons showing their leverage ratio is actually higher for international banks. So I disagree with that point. I'd say the same about the euro area. Let me make two points. First, to reinforce what Randy said, we know banks maintain management buffers above our requirements. What we've done will enable them to have more confidence to reduce those buffers. Second, we included a diagram in the publication. We would like to see the buffer structure simplified. It's too complicated. We would like a single buffer and a single releasable buffer. The reason is that if we don't release buffers and tell them it's okay to go into them, they always ask if we will be prepared to take the buffer away when conditions change, accepting a rebuilding process later. We'd like to see that simplified with international agreement. We've been leaders in advocating for that. We'll keep working to get international consensus. We're not there yet, but that is the bold thing to do, and we are strong supporters of it.
But when banks look at their effective situation, they tell us they never get to that amount because of all the add-ins and overlays that occur. So whilst I understand the aspiration to simplify and have a single buffer, the reality is that because of their hesitancy around the implications, the net effect is they end up holding more than they feel they ought to. That is a persistent narrative they bring to us and members of this committee have had drummed into them at private meetings recently. So you present a very smooth version of where you'd like to go, but operationally it isn't working like that right now.
Can I just pick that argument apart for a moment? Why do they say that? There are at least three reasons. One, we've already covered: they maintain management buffers above our requirements. That is their choice, but we can simplify the structure to help them maintain smaller buffers. I think the big step would be moving to a single buffer with international agreement. Second, capital requirements differ from bank to bank partly because of the business they do and partly because sometimes they fail to demonstrate the characteristics we expect, so we put add-ons. That's us doing our job. If they can't manage risk adequately, supervisors can impose higher requirements. That's a reflection of the risk they take. Third, many still think 13% is too high. I want to come back to points I made earlier. We will have a strong banking system if they are well capitalized. They are lending into a strong economy and earning strong returns. The evidence shows they are in a strong position. There were complaints about how much they are earning. I've said they have to earn their cost of capital. They are doing that, which is a sign of resilience. They will never say that to you.
I'm afraid the odds of us getting them ever to say that yeah no it's not going to happen. But it it I'm afraid it is a fact. They are supported by that. If they have weak capital, they will not be supported in that sense.
What what what it feels like to them is that there'll be a move forward in one area and then for example, you know, Sam Woods uh did not include the exclusion of unencumbered guilts when looking at the regional uh the um leverage ratio denomination.
But that's not in Basel. No, Basel does not. Nobody does that. The US doesn't do that. I mean excluding guilts from the leverage ratio. I know they'd like it, but frankly that's not in Basel. would be contrary to Basel and nobody does that.
Thank you very much, Mr. Did Mr. Crosno, did you want to come in? You look like you were.
So, I think this is one of the reasons why we have stress tests to um and and make sure that they do have appropriate risk management and most banks do pass those stress tests now. So, it's not worth setting a barrier or a hurdle that is so high that is unattainable that no one can get an A in the class. Um, now most of the banks are getting an A in that. So they know how to achieve that and I think it's completely appropriate that if they are taking undue risks then they should have add-ons. If they have processes in place and sufficient capital and and uh appropriate risk management then they are free from that.
But the fact that major UK institutions have an effective requirement closer to 14.25% 25% rather than 13% in reality is really just a function of them not operationalizing risk management appropriately.
Well, it will come down when Basel 3.1 I mean part part of this is that Basel 3.1 will reduce that because when we went there work in progress when we went from 14 look when we went from 14 to 13 we were actually marrying up top down and bottom up. The FPC does top down PA does bottom up. um the the bottom up element is coming down consistent with 14 to 13 and a good part of that is parcel 3.1 and that's why the FPC felt it was sensible to to match the top down. C
could I ask you Mr. Crosno about this uh what the banks tell us about the apparent gap with the US that the governor um pushed back on. Um I mean what some of them have said to us is that there is a the UK banks now face a 190 basoint capital premium compared to the US competitors. Now you can dispute the the quantum of that but do you recognize anything anything like that or any meaningful difference between the US and the UK?
Many different ways to to calculate those uh those sorts of comparisons.
Yeah. And that's how you sort of get out of jail on this subject.
Yes. Yes. And also one of the challenges also is that just accounting is very different for like what what are considered bank assets and so that becomes um a challenge to do the international comparisons. Um I think you know we've been trying to undertake a review to and and obviously what we put into the uh our reports is looking at international comparisons. The UK is a major global financial center and it has to be competitive and I think we we very much take that into account. I'm very cognizant of that and bring that into uh into the discussions. U but then when we get into the specifics of where there's 190 basis points or minus 10 basis points uh that becomes much a bit more of an art than a science.
Do you think we'll ever get to a point where it will be apples and apples and you can make a reasonable comparison and we'll stop receiving all these representations from the banks. It might be worth if I mean that may want to come in because we published obviously in December a detailed comp I think the most detailed comparison.
Yeah. You you've looked at the feedback that we've had which is where they make these points to us.
That's right. And and there has been a range of feedback uh coming from different uh absolutely different uh different sides. Um and um so I think the fact is that with the package we have now uh set out this brings in line UK capital requirements uh with uh requirements in other jurisdictions and in a way that is consistent with the analysis that we did and consistent with the feedback we received on on this point. I think the other point um uh to note is often US banks I used to supervise US banks. banks hold also hold internal uh management buffer above their regulatory requirement actually much more uh than some of the uh some of the uh UK banks and yet that doesn't have a detriment uh to you know their earnings or or their or their profitability there's a sort of a
just very briefly um we also hear this in the US the international comparisons the US banks always feel that they are being unfairly um uh goldplated so I think it's a natural thing for domestic banks to go their domestic regulator and say, 'Look at the others. They get away with this. We don't get away with it.' So I think you would hear a very similar thing if you were sitting in the US Congress.
Sounds like managing teenagers. But that's a bit
No. Sorry. Yes. I I can ask one one more one more question around the net effect of this in terms of to the banks and the bank shareholders or to the economy as a whole. And in terms of you know the the narrative that they bring to us is that you know if they can actually get access to more money to to to lend to grow grow the economy yet we see you know banks doing quite well in these circumstances. Do you have any views then on how we should review the performance of banks financially and the impact on the economy and and what they do with them with the freedom they've got? Well, let me come back to what I said at the beginning. Um, I do not I mean I do push back at this what I sometimes call sort of lump of capital argument uh that goes on which is this idea that they've got a fixed amount of capital and if only we'd lighten the regulatory regimes they could sort of lend lend more out support the economy and everything would be good. Um because there isn't a lump of capital. I mean the fact is they earn uh you know their earnings they then have a decision to take about how much of their earnings you know they frankly distribute to their shareholders and how much they retain to to grow the business and obviously that's a process that they you know have to go through with their shareholders. Now I recognize we went through a period up until only about two or three years ago after the financial crisis where they were not earning their cost of capital and it was very hard to go to the shareholders and say you should leave this money in the business because it's a good thing because the shareholders say well no it's actually not a good thing because you're not earning the cost of capital I can take it somewhere else and find somebody who is but that's not true anymore they are more than earning that cost of capital and the best illustration of that is is the price to book ratios they now I think on average about 1.7 whereas they were below one. So one is the sort of the is is is is obviously the point where they're they're at cost of capital. So they're not in that position. So you know they can I think they are resilient. That's why as we talked about stress tests they pass stress tests. They can make choices now and the choice they can make is to support you know lending into the economy uh with with those earnings and you know that's what to my mind they should be doing and a resilient system will support that and it will support the economy.
I'm just going to bring in Mr. live then it wants to come in and we just we do need to wrap this section up because we we could spend forever discussing capital to the bank and I'll bring in Mr.
Thank I just wanted to bring attention to that we we've published um the feedback from a very broad consultation AC about this matter and not obviously not only from banks but from academics from rating agencies from bank analysts and bank investors and I think that is a very rich picture and I think it's it's clear that banks will all things being equal prefer to have less capital and less requirements I mean any investor or risk taker would would like that and I think that what we heard from academics and bank analysts in particular is their arguments that the capital should be higher and actually bring it down is not the right thing to do. So I think this is a very quite a balanced a balanced argument from a rating agencies and bank investors saying well you know we look carefully at the capital people are holding and we want to make sure that's appropriate in order to invest in these banks and I don't think their their view is that the the capital levels should be going down.
That's very helpful. Um Chris one brief question.
Sure. Um governor sounds slightly skeptical about the bank's position on this. Uh, of course, prior to 2008, banks were making exactly the same argument that, you know, you need to just deregulate, lower capital crimes, all will be fine. Economy will go gang busters in the process of the global economy. To what extent does that history inform your thinking now?
Well, it does a lot actually. So, I mean, I'm realistic enough to know that, you know, memories of the financial crisis are now disappearing sort of into rear view mirrors. Um some of us probably all of us probably have unfortunately you know undying memories of of it having had to deal with it. Um but they are disappearing and so you have seen this in history before you get these cycles and and I think we are seeing this where people now saying you know the response to the financial crisis was too great and too strong. It didn't feel like that at the time. I have to tell you when we were dealing with a system that was extremely weak uh and having to rebuild it. Um I get people saying, 'But you've solved that problem. You should move on.' And you say, 'Well, you know, yes, in the sense that we I think the system is now resilient. It's rebuilt.' But the idea that we therefore sort of relax the standards because it's all over. I mean, no, that's not the way it works. So, yeah, I I'm philosophical. I mean, this is this is a debate we have. And I just come back to this point. I'm not against, you know, saying I'm not going to defend all our rules as perfect. They're not. Um, but the central point is we will not have a strong economy without financial stability.
Thank you very much indeed. I'm just I'm actually going to move on because we could go on forever discussing this, but we do have other topics and on the subject of AI, which we we parked at the beginning. Um, uh, Governor, I'm going to bring in Bobby Dean.
Thank you, Chair. Yeah, Governor, I'd like to return to something you made a point out of highlighting in your opening remarks, which is about the sort of heightened risks around frontier AI.
Um, I think critics will say, 'I told you so.' Um, when the discussion started talking about AI a few years ago and it was kind of leveled down as a as another risk to manage, and now you're saying that it's grown much faster than you anticipated. So, what is it that you didn't foresee previously?
Well, it it's not just us. Uh I mean if you talk to the people who run you know anthropic um open AI they will tell you that the advance that they've seen in these models in the course of you know the last months has been much faster than they expected. Um I mean they draw these sort of what I call exponential curves um and they're saying we've now moved into this vertical sort of part of the exponential curve in terms of the advances of these models. So you know they too were taken by spite. So if you take Mythos which was the anthropic model which really sort of kicked this off back in well April really um and I I know I've talked to the I talked to Dario Amodia about this you know they were surprised by the advances this model made and when I talk about that the reason is that these models actually they are self-arning they do this sort of recursive learning process so it you know it's that it's the speed and effectiveness and and and the ability of these models has has has moved dramatically forward. forward and so when they're subjected to the sorts of tests sort of tests that the um AI security institute in the UK which is by the way absolutely worldleading on this subjects them to when they test the models and they say how how effectively and how quickly are these models solving and devel
I think we can all accept that the the reality has moved on much more quickly but we obviously had an inquiry into AI financial services and I feel that we heard evidence about the potential advances in AI and how they felt that the financial service sector was perhaps underprepared for the potential movements. Now the fact this happened more quickly than anticipated is one thing. Um but one of the things we heard from lots of regulators at the time was that actually our existing risk frameworks are are fine. We can be technology agnostic. Do you still hold that?
Well, I think can can I say I think there's two things here that we can and they need to they're both related to AI but they they need to be sort of kept separate. So on Frontier, we're talking about cyber. Very specifically. And to use a a sort of sports analogy for a moment, Frontier AI will strengthen the defense and the offense. So if you defense, if you're us, if you're the banks, firms, you can put frontier AI to to work to strengthen because it will find the flaws that have been buried in there for such a long time sometimes and you can fix them. uh unfortunately the off if the offense gets hold of it they will do the same thing to attack you and obviously the whole thing at the moment is is can we make sure that the defense gets such strengthened you know effectively before the offense and that's getting much more challenging the second thing which I think you're referring to is equally important but this is how banks use AI so we're not talking about cyber here we're talking about how they use it and the risks that that come with that that's equally important and I think you know you your report was very helpful and very good in this respect effect. We are very, you know, actively engaged on that subject too. I think the one banks are I think my my sense is banks are use we're all using AI a lot more. I mean that's a trism but you know that banks are using it a lot more where I think it's still more talk than actuality is is things like agentic trading systems having setting up your own agents. Um, and I think that's right because I think there are some very big questions still to be answered about how that world actually would work. For instance, I mean the one thing I'm not a lawyer but you know the thing that strikes me is if you have an agentic trader trading system as a bank and going back to the going back to Frontier AI for a moment. I mean what frontier AI demonstrates is that the recursive learning capacity of these systems means that they actually develop their own abilities and you know as as my colleague Sarah Breeden has said but she we get we we don't invent this we get this from you from the firms who are developing it they learn to cheat and they learn to lie. I'm I'm sad to say why where do they learn that from? Well, um, you know, it's partly inherent in them and partly because they're trained on a matter of past experiences and they cover things up. Now, if you've established an agent, you've got to be very clear where the legal liability rests in all of this. So, I think, you know, there are things that will, you know, need to be resolved before that world really develops.
Mr. Pedro, if I could turn to to you, and we might want to continue with the governor's World Cup analogy about attack v defense. um what are firms current access to these frontier models presumably they're on the defensive side of this equation uh so how is their defense
so I mean the the the financial sector generally uh has been putting a lot of focus on uh on on these things and what is happening at the moment is uh firms are in the process of testing those type of capabilities on their own system so the big risk that comes with frontier AI development is the speed at which weaknesses in IT estates can be identified and therefore what that means for patching and the speed at which software upgrades uh need to be made. So those are this is the work that is uh happening. Andrew mentioned earlier uh things that give the edge to the attackers versus things that give the edge to uh the defenders. One of the key aspects is the speed at which those capabilities become accessible to the attackers. they are typically accessible before to the defenders. So this buys them time in a sense to identify those weaknesses, those vulnerabilities and to prepare for much faster patching because another risk is that the moments when software are upgraded are also the moments where they are greater risk of disruption of it disruptions internally not because of an attack but because of the the effect of the upgrade itself. So this is the work that is happening within the banks and within large firms at the moment to uh get them prepared to this new uh state of the world.
We we've been talking about risks in the abstract. Could you describe to us what the impact would be if say the attack got there first to to a standard financial firm institution if there was
there are two ways uh cyber tends to manifest itself. one is those who want to disrupt uh uh systems and then often often there's a blackmail effectively and then the other is stealing stealing information. Um those are the two things that can go on. Um and as Matt was saying, I mean what what we're finding or what is being found is that you know flaws that have existed in very core pieces of software in in in the world and this by the way I mean fin systems are just a subset of this obviously it's it goes across the whole the whole the whole shooting match as it were uh that have often been there for a very long time can be discovered by these systems and can be discovered quite quickly. Now, I mean, I don't think I think the people who know much more about this than I do, I think at the moment, going back to the football analogy for a moment, it requires a lot for an offensive party to set this stuff up to do this. It's not, you don't, by the way, go home and get Miffos, plug it in, and off you go. It's rather more than that. But let me I mean, this is the second issue that we've got. Um, as I said, we're on this Yeah, we're now on this vertical exponential curve, they say. So that yeah these models have a shelf life of only about two or three months now actually and then they're replaced by something even more powerful. Um I know that you if you talk to the firms the developing firms they say there are going to be open opensource versions of these these things coming on on onto access which will mean that it's much harder to restrict access to them and then let's be honest you know where are other regimes in terms of developing their own and obviously China is the big case there you may be some months behind but they're not that far behind
bring in Mr. John Grady just one quick
briefly governor you mentioned legal responsibility for agentic models um I'm a recovering lawyer um if a is the bank and do you think we are clear about the who is responsible if for instance financial institution A uses an agentic model and it starts cheating and lying and as you describe briefly because I don't want to derail the whole hearing
well I'm not a lawyer so let So you know I think we it would be good to I mean and we are certainly doing this to get a you know to obvious to get the views of the legal world on this. But here's my understanding. Let's go back to sort of basic principal agent le law. If if I'm the principal, you're the agent. You're my agent. Um I think in the traditional sense I give you a remmit to act on my behalf. I am liable for what you do when you dispense my remitt. If you go outside my remit and start doing other things, you're liable. That works, of course, when you and I are both legal persona. In other words, you you could be held legally liable. I could be held legally, but obviously with an agentic AI model doesn't work because you're not a legal person. It's not a legal persona. Now, I think you know it's normally the case that in that case, so let's get going again. the agent does something outside the remit, the principal is liable where the agent's not a persona. I think well if that's the but I don't think you know this hasn't been tested much in the AI world but if that let's say that is the case then I think the principles the banks let's say are going to have to think very carefully about obviously therefore how they manage uh the use of agentic AI because if they're going to be responsible for everything it does and if there's this issue that it can go off pie then they've got to think that through before they start using this stuff the whole can of worms and I think it'll be useful one the limits of my knowledge I have to be honest with you
to come back to us in writing on that important point
yeah of course there's the third parties that are involved in as well the AI companies um and some of the major tech companies as well in terms of this I mean and if we move on to a slightly separate point which is the critical third parties uh regime and I think you there will be four brought into the regime now a concern that's been raised is has been how much oversight you're actually able to have over these firms when they when they kind of quite opaque about the technologies that they hold. Um, Mr. Benjamin, I don't I don't know if in your capacity, what what is your relationship with these providers and and how much do you get to look inside their black boxes to understand what the real risks are?
The the regime is coming in place uh now and our colleagues on the P will be the ones uh supervising uh them. So that that is going to that is going to to begin uh and we needed the regime to be in place first before before coming into that.
Are there these general concerns about working with these providers and and how much information they're willing to share about their systems in order to be able to understand them?
No, I mean if you if you're refer if you're referencing the four that have been nominated, no, I wouldn't say there are concerns at all. I mean we've had relationships with them. We've had discussions uh with them just for broader reasons uh in in the past and they're very very open uh to uh to those type of discussions. So I'm not worried about that. I
I guess my point is more about regulatory capacity though. I'm sure they're going to appear to be very open and will willing to cooperate. But does the bank have the the staff, the knowledge, the knowhow to be able to properly hold all their systems to account. So we have our colleagues in the PR have entire divisions of people who are specialist of operation resilience and who are well verssed in these in these technical topics.
Okay, I'm going to move on to a slightly different threat that's still related to AI and perhaps Mr. BL you can speak to us about this which is uh emer well growing concerns about the threat of an AI bubble and the impacts that that could have. Um the easy comparison is that of the dotcom bubble. Are the risks similar or worse?
I mean, it's it's a good good comparison in terms of uh the level of uh concentration and and heightened valuations and I think uh especially in equity markets. I think there are there are a number of elements here. First is that the continued growth in or increase in equity valuations especially in AI stocks has been um coincident with with very strong earnings. Um but the current valuations are predicated on a very optimistic future earnings path. So it's not clear that it's necessarily a a bubble. Uh but it is in our estimation and the bank's estimation based on a quite an optimistic path going forward. So if uh earnings don't um aren't realized along the optimistic path then there could be a significant significant repricing of of of AI equities. On that point, can I put to you I think it was JP Morgan's report where they they said something like $650 billion dollars worth of revenue per year would be needed to get a reasonable return. And they also remarked on some of the sort of physical limitations in ever being able to achieve that kind of revenue because of you know we talked about water supply earlier and data centers and other sort of physical infrastructure. Is that something you recognize? Do you think that actually as things stand the returns required are actually physically impossible to get at? I wouldn't say they're physically impossible, but we're very cognizant of the the risk both power cooling that we've mentioned before uh planning um you know actually being able to um actually get approval to build data centers. All these things are risks to a a positive earning story. And I think from a financial stability question, there are two two elements that are particularly concerning and or on the radar. One is that the funding for a lot of air expansion is now going form in debt form rather than uh equity form or from cash flow. So that you can see from some of the data the growth in 2026 has been extreme and obviously debt um if there's a popping of the bubble can have a lot more impactful uh downside and the second is there's there's a lot more leverage which we mentioned before but leverage on the long side in AI equities and that's both institutional and retail and that's both both of those are can be concerning. So institutionally a lot of uh levered funds hedge funds have been taking long positions in AI equities often funded either in derivative form or funded by prime brokerage and then on the retail side a lot of levered ETFs. So this means if there's a fall in equity prices then that can lead to a a negative spiral. So I think from a financial stability point of view market volatility is not per se a a financial uh stability concern but when it can be amplified rather than contained then it does. And I think that's those risks are growing on the AI side
and that change towards height and leverage positions has happened very quickly it seems over the last year or so. We got um information about Alphabet's position for example. It does that make you feel that a potential bubble burst is much closer than it was this time last year.
Um again it's hard to it's hard to predict hard to predict but I do think that the impact of a bubble burst would be more significant given that leverage. I mean you point the pace of change of leverage is remarkable. I mean the data that we see on the prime brokerage increase that is really that's that's remarkable. Uh levered ETFs um the pace of growth in that is remarkable. It's not as big um as in the prime brokerage but that pace and the pace of increase in debt issuance is is extraordinary. And so there are there many things we'd be thinking about. One is is this crowding out other debt issuance because there's so much debt issuance coming in AI. Is this actually now just increasing this levered leverage long to the AI story across the financial system? And if that levered long to the AI story across the financial system means greater interlinkages when the bubble bursts. So hard to say about the probability of of of a valuation uh reset.
Can I just ask you about the UK's specific position in relation to this? because obviously we don't have so many developed AI firms will still be exposed to any uh impact and and there's kind of two sides to the coin really. One is um the risks of that bubble bursting and what that means for the UK but also in the meantime what what's it doing to our growth prospects if so much capital is going in one direction?
Well, I mean as you say even if there was a if if there was a bursting of a bubble which is primarily not located in the UK the market effect is so large now that it clearly would wash over. So I think that's that's right. And of course we ran a stress exercise which which looked just at this question and it did uh reduce UK growth. Um now exactly how the effects would run through you would expect obviously some adjustment in UK macroeconomic policy which is another way of saying you know you might expect some adjustment in interest rates to reflect that. So how you play it through you'd have to decide but yes it would wash over on the UK growth side. Um I mean let me I just want let me start by adding two points to what Stephen was so interestingly saying because I think there's two other issues here on how can this sort of how can this go wrong. Um one is history would tell us with these big general purpose technology innovations that it takes longer for them to come through into productivity and growth than people think. Now this one may be a lot quicker because the p we were saying the pace of change in AI is clearly accelerated a lot but you you I mean you go back to something like Edison inventing electricity it took a lot longer you you go back to ICT the famous um Bob Solo comment you know you can see computers everywhere except in the productivity series uh and that's really one way of saying it takes time for these things to come through. The second thing is everybody is currently priced to win. um everybody sort of doesn't win in history. If you can remember the first company that actually got the biggest share of uh of of internet browsing, it wasn't Google, it was Netscape. Um so, you know,
us all the dates all governor, those of us who remember that.
Yeah, I'm sorry. Sorry, it's me. I don't know. Um, in terms of the UK, um, I I mean I think it's important first of all, you know, this is obviously a highly globalized market for these things. Um, I think it I think the UK does have real advantages in this area. I mean, I think there are real uh, you know, centers of expertise. is I think we've got a university system that is is is a very strong you know producer of of very strong uh out you know research and outputs. I think it's you know it's we have I think a very encouraging you state of affairs at the moment on tech startups. I mean Patrick Vance was saying to me the other day we think we've got the most ever in this country. I think the persistent question is actually the next stage. um you're much more expert than I am is how you go from startup to to scale up and that's that's an issue that remains a big issue in this country
MP you want to come in or not okay that's fine okay uh uh Chris MP okay fine you
thank you chair if you okay
thank you chair missa I want to go back to your comments about um leverage and the guilt markets and the issues there. Um the Financial Times reported earlier this month that there would be new reforms announced um this month um to the management of the guilt markets in particular to the use of haircuts um you in particular to the use of haircuts. Haircuts. Yes. Are those reforms still planned to be announced this month or what
