The Blind Spot in U.S.-China AI Competition with Paul Haenle and Terah Lyons

[Music] Welcome to "Free Risky Business," the economic security podcast from the Center for a New American Security in Washington, D.C. I'm Emily Gilkruz. And I'm Jeff Gertz. On today's episode, we're joined by Paul Hamley and Tara Lyons of JP Morgan Chase to get the private sector perspective on how G-Economy Competition is reshaping the investment landscape. We talk about the state of U.S.-China AI Competition, the most recent offer over Chinese open week. And why policy makers should pay attention to AI ecosystems rather than any single models benchmark scores? We also discussed JP Morgan Chase's new security and resiliency initiative and the challenges of investing in strategic sectors where markets, left to their own devices, may fail to deliver outcomes that meet U.S. national and economic security needs. And just a quick note and disclosure before we get into this conversation. Except from this I've asked, JP Morgan is a financial supporter of CNAS work. As with all of our supporters, we maintain strict intellectual independence and editorial discretion in our research friends, including our podcast. More information on our intellectual independence policy can be found on our website CNAS.org. Guys, please, you can reach us at dmsbbismas.cns.org. This podcast is supported in part by Amazon and through general support to CNAS Energy Economic and Security Program. Alright, so Paul and Tara, welcome to Dersky Business. Thanks for having us. So we wanted to bring you both on. So you work at JP Morgan Chase working on teams covering geopolitics and AI. These are two topics that here at Dersky Business we spend a lot of time thinking about. So maybe just a sort of background to guys into the conversation. Can you tell us a little bit about what your respective teams do and maybe kind of particularly think about, you know, what's unique working on these issues within a major bank relative to kind of government or think tank. Sort of more of the often policy guests we have on that have that kind of poke policy angle. What's the private sector angle here that's really relevant. Paul, do you want to kick us off? Happy to. Happy to. Well, thanks for having us on Emily and Jeff really appreciate the opportunity. As you mentioned the geopolitics and AI side, I sort of am in the geopolitics side. I lead JP Morgan's Asia Pacific policy and strategic competitiveness team within the center for geopolitics. We were closely with colleagues and Tara's team. We were closely with colleagues in the global government relations team and across our firm security and resiliency initiative, which we can talk about. Because many of the issues that we focus on, whether it's technology competition or trade supply chains industrial policy, they all sit at the intersection of geopolitics public policy and business. And I guess to answer your question, Jeff, I think unlike government where I've served. Where you, you know, helping to make policy or in a think tank where I also spent 14 years with the Carnegie endowment. Where you're researching and recommending policy. Our role is really to help the firm and our clients understand where policy is heading. You know, where are geopolitical developments heading that are going to reshape markets and what those changes mean for business strategy and we engage colleagues within the firm on that. And then of course our clients outside of the firm. And for my part, actually, Paul and I started within two weeks at each other. That's right. And our teams were kind of formed as sister teams in large part because at the time, almost three years ago. And the firm came to the realization that these issues were going to be sort of critical path interests to not just our business as a company as Paul is mentioning, but to all of the clients and communities that we serve. So our team is a little bit of a hybrid sort of experiment. So I would say within the firm, I sit on both the management team of our, our team did an analytics office organization, which is the central team that is a spanned line of business. And globally, all the firms work in AI data operationalization kind of putting all that to work for all of our business and also leadership team of our global government relations organization that Paul mentioned. He also works really closely with. So we do work across kind of corporate responsibility, traditional GR external affairs, public policy. And on the sort of on the internal side, all of the work that the firm does really trying to make AI effective in use for our own business. And we are, you know, we're significant player in that regard. We don't necessarily get classified as a tech company, but we spend almost $20 billion a year on technology, a significant share of which is focused on these sort of AI and data products oriented objectives, of course, and in software more broadly. And we are actively deploying over 400 use cases in production of AI across the firm. So it really touches almost every dimension of the work that we do is a business. And of course, externally is of significant interest to clients kind of as I mentioned as well. So that's a little bit more about what we're doing over here. And if I could just to kind of come full circle with what Terragis said, I would say AI is probably one of the clearest examples of where technology, geopolitics and business strategy intersect, which is why I think, you know, as Terragis suggested it kind of naturally brought our teams together. We came in around the same time, you know, we were, you know, we were two new capacities that were built within the firm because neither the policy story nor the business story is complete on its own. And I think that's why the collaboration that we have is so critical. So I want to pick up on a one thing that you talked about, which is, you know, big, big multinational companies, big banks have always had kind of an eye on geo politics, right, it can affect the bottom line. One thing we're following a lot at CNAS and everything happening in Iran, I've had a huge shifts in the market. And I have an angle on this and I hope you get into in the conversation, which is that there might be opportunities here, right, like you're a bank, you're allocating capital. How do we think about that role of capital allocation investments in this very shifting, fractured geopolitical environment that can be difficult, right, clients will need to navigate that will need to navigate trade barriers, and they're asked, but maybe there's opportunities that open up as well. And so, you know, kind of curious, your initial thoughts on that, I hope you get into that in the conversation too, how the bank is thinking about the role of like some of the things that policymakers want to do are going to cost a lot of money that's kind of come from somewhere. Yeah, so I mean, I guess I would start Emily by just saying, you know, if you listen to our CEO Jamie Dheim, and he often talks about geo politics being one of the biggest risks that face the global economy. And I think that's right. And so today, geo politics is not something that companies just watch from the sidelines. It is really shaping investment decisions, supply chains, technology strategy, capital allocation, you know, all of those issues on a daily basis. And so we, you know, created the center for geo politics really to help clients understand where policy is heading, what geopolitical developments mean for markets, and how those changes might affect their businesses. I also mentioned the security and resiliency initiative where, you know, we're facilitating 1.5 trillion dollars over a 10 year period, you know, to helping, you know, build capacity and critical supply chains, you know, build economic resilience for our country and help the defense industrial base. You know, what is central to those, I think, are public private partnerships where, you know, it may take some government intervention, but to reduce risk or to create demand, but ultimately, you want to crowd in private capital to keep those industrial sectors moving over the long term, and I think we can really help there as well. Sorry Jeffrey, I was just going to add one quick layer to what Paul mentioned and saying that I think the other object here is like one thing I really appreciate about working at JP Morgan, one of the things that drew me out of the public and social sectors to your earlier question, I'm late to work in the private sector. And I think we in particular, but banks more broadly and financial institutions for large are used to sort of being in the arena because they have to be. And so, you know, for all the reasons that Paul just conveyed, and I think we're doing a lot to sort of lead with our resources really, you know, act. Action behind our words through the SRI and other initiatives that Paul mentioned, but it's also just become an operating imperative for organizations. to be involved in policy discussions of this sort. And I think in the AI realm in particular, that it's never true, it has never been trueer than in the non-topic area, which I know we're going to talk a lot more about. So I just wanted to overlay that as well. Great. Let's get into it all. We'll kind of come back to SRI issues in a moment. But first, one of the reasons we want to bring both of you on is that GPU and Morgan Chase just put out this big new report on US China AI competition. So we want to kind of dig into that begin with. So the report that came out argues that the US China AI auditions becoming one of the defining geopolitical dynamics of the 21st century. You know, Paul, I'm sorry with you and then go over Tara. But you kind of walk us through some of the big headline takeaways sort of why you wanted to explore what should our listeners kind of remember about it. Well, thanks, Jeff. And thanks for highlighting the report that in OTERRA and our team did collaboratively. I think the biggest takeaway, I guess I would say, is we're not asking the right question if we're simply asking who has the best AI model. That isn't fundamentally-- this, I think, is not fundamentally a model competition. It is a competition over which country can build the strongest AI ecosystem and then deploy AI most effectively across its economy. Ultimately, AI is not going to be valuable because you've built the best model. That is important, no doubt. But it will become really valuable if you can diffuse it across manufacturing, health care, finance, logistics, scientific research, and sort of the rest of the broader economy. The US has important strengths across that ecosystem, which we highlight in the report. China's comparative strengths increasingly lies in deploying AI rapidly and at scale as in several other industrial sectors. So I don't think the competition will be one by whoever builds the smartest model. It will be one, I think, and this is probably one of the key takeaways, by who can translate AI leadership into lasting economic strength, industrial competitiveness, and of course geopolitical influence. I would even go further than what Paul's describing and actually saying that this is-- I don't even necessarily think that this is a race, even necessarily. I think that there's sort of the binary US versus China escalation dominance sort of frame up, I think, for the wider competitive dynamics between the two states. In many ways, frame through the AI piece of this equation is sort of a false premise, I think, to begin with. And I think some of the nuance that Paul is overlaying are the many varied dimensions of measurement within that sort of dominance frame is part of what we're trying to bring to that conversation. So as he mentioned, wrote model competitiveness really is not the only thing at play here. We look at policy, we look at hardware, we look at financing, socioeconomics, energy, and security as different dynamics in that picture. And the other thing I'll say is I've been in the space for a long time and actually doing measurement in AI progress development for a very long time, including through work at the Stanford AI Index, which is now in its-- marching towards its ninth year, its eighth year publication this year, where we used to just basically have benchmark data on the most capable models. And now there's just so much more availability and richness in that picture, being demonstrated both by projects like that one and by the work that we're trying to do here through the report that we've conducted among other vehicles for that. Great, so maybe that's a good kind of thread to pull on. Take this point well that it's not just about who has the best AI model in a given time in that competition view. So we're recording this July 27th here in DC. At the moment, there's a lot of for you attention around the Moonshot Kimi K3 model that came out just last week. This idea of-- oh, has China now caught up or it's open weight models going to take on the US? And that's a question to you. Does the release of Kimi K3 change anything in your mind? Or is this, again, one small point in a much bigger, broader competitive dynamic? And we should not pay too much attention to the new Chinese model, hit a new benchmark, whatever kind of-- that's not actually what we should be watching. It's a good question. Oh, that's-- OK, go ahead. No, no, go ahead, Paul. No, no, no, please, please. I didn't know who was asking to. Well, maybe I'll kick us off. And then I'll pass it over to you because I know you've got stuff to say about this as well. But Jeff, I think what I'd say about that is that all of these signals, I would say, are worth paying attention to. I'd rarely dismiss any of these developments as being irrelevant to the broader conversation being had-- I think what's really notable about the Kimi development, in particular, in this case, is that it really opened the breach in terms of the conversation now being had about open-weight models versus closed. And the sort of relative competitiveness of those approaches from a technical perspective, but also the utility of those economically within this broader discussion and frame of business models emerging in the industry. There was a letter released last week spearheaded by large enterprises, including Nvidia, Microsoft, Meta, and others with CEOs calling for protection of what sort of widespread availability, of open-weight models, and the start at community and innovation economies, spectrum of stakeholders also really pushing for the same availability just by virtue of the scale of their use right now, especially by small and medium-sized organizations that really find them cost-competitive to close weight alternatives. So that's kind of-- that is backdrop to the broader issue space. Those are all very material developments, especially from a public policy perspective, where we concurrently have a Trump administration potentially weighing action on the availability of private sector use, so open-weight or foreign-open-weight models potentially. At the same time, I would also say that models like DeepSeek before Kimmy and now Kimmy's release reveal that the gap at the model level is actually narrower than benchmarks may have originally suggested. And so that's also, I think, material information, all of these conversations. China's really deep-enginering talent and their deep investments paid in this whole ecosystem in the last five years, especially, have really allowed them to produce highly capable low-cost models and to get really creative about the ways in which they do that, given some of the sort of paradigm progressions that we've seen take place with evolutions like the one that Kimmy represents. But again, I think coming out of all of that success isn't necessarily determined so much by who builds the best model, but probably by who deploys it most effectively. And that gets back to the conversation about the sort of-- the economic issues that hand here and some of the competitiveness dynamics that we're starting to see play out in the public policy conversation about accessibility of these models as options for, especially for private sector players. But Paul, go ahead. You were going to layer something as well. Yeah. Well, first, I was going to go back to your comment about it being a race is not the right sort of frame to look at this. And I take that point intellectually. It's harder for a guy who works in geopolitics on an issue that is clearly right now smack dab in the middle of the strategic competition lens between the US and China. And policymakers that I speak to in the US government certainly view it in this way. And I would say Chinese policymakers as well view it in this way. But I take the point intellectually that it's not necessarily the right frame to use. But if you-- like the Kimi example, caught the attention, obviously, of US policymakers for the reasons terror that you said, which is maybe the gap that we have over China in terms of our frontier models is not as wide as we thought. One thing I would add is from a China watcher perspective is Kimi came just as China hosted the World AI conference. Xi Jinping delivered his most substantive speech on AI to date. Now he did not focus his remarks primarily on AI about beating the United States. But there was an element of a strong element about positioning China as a global AI partner, especially for countries across the global South. And I hear I think this is an area where China is racing to build their capacity to get China's models out into the international community in particular developing countries. He emphasized open source AI. He emphasized international cooperation in his speech. And he emphasized helping other countries build AI capabilities. And that I think reinforces the idea. China is increasingly viewing AI, not simply as a domestic technology, but part of its broader geopolitical strategy. which again puts it right into the context of the strategic competition with the United States. I think there's something really interesting about this example as a sort of its own petri dish of sorts in terms of driving action and investment in this in the broader kind of environment and and something you had on there Paul related to open source emphasis I think is really a key here like just in the public attention and sort of public policy maker attention being driven by the kidney release and the aftermath of that that we just talked about. There's been a huge sort of swivel in the direction of focusing US investment and attention on driving more effectively competitive open way alternatives to Chinese counterparts and I think that's the perfect sort of it's really exemplar of this sort of intersectionality between commercial and market interest investment you know questions and the sort of geopolitical and political issues that we're talking about here are really driving a lot of this this sort of like the race dynamic if if there isn't a better way to put it and I think it's going to result in the US actually having explosive investment probably an open way alternatives to China on the basis of that need for competitiveness but it's a little too early to make that call so we'll see what happens. Let me pick up on that because I think it would be super interesting at this moment led to a huge surge of investment in US or you know allied democracies open-weat models and all of a sudden we've got a good competitive open-weight solution and you know companies aren't as attracted to the kind of Chinese or foreign cheaper yet good enough open-weat model maybe that would be great. There's another way this could attract though which is that you know we kind of do what we did on Huawei and we kind of just you know don't create that alternative and so the commercial incentive will continue to be to you know for a lot of use cases it's a good enough cheaper accessible model is going to make a lot of sense for companies even in the United States there's another way the policymakers could engage here and of course you're starting to hear even secretary Bethan talk about it which is to use the more restrictive toolkit right so should we sanction some of these open-weights mild developers should we ban them in the United States I think it's really hard to see how some of those policy measures would play out but if we're just banning it in the United States that is almost like exactly the Huawei playbook again where you know kind of the Chinese open-weight models will just become prevalent and diffuse all over the global South you know and so I'm curious maybe Paul for you as you look at this especially with your deep experience on China how much of this conversation around the open-weight models is it's looking at kind of all the mistakes we made in the Huawei case the right example and you know how do we kind of pivot away from that and try to get to something more productive like like Tara with that lining up a run faster let's make the right investments now yeah I think you're exactly right I think and I hope we're kind of entering a new phase of the conversation as you said the past several years a lot of the focus has been on restricting China's access to advance semiconductors and now you're hearing talk about you know potentially you know sanctions on you know Chinese AI models in the US you know I hope to see going forward the discussion increasingly shift toward how the US and its allies can you know maintain their lead you know kind of move to the offensive side of the competition in a more robust way I suspect the US will continue to try to you know keep its most advanced semiconductor technology out of China's hands you know even if policy makers allow for sales of kind of lower performance chips designed kind of specifically for the Chinese market like the determination they made on the H-200s and so I think maintaining a lead at that technological frontier will likely remain you know a central objective of the administration what is important I think equally important is how China is responding to all of this one of the things we hear consistently from Chinese officials you know kind of when you're when you're talking to them one-on-one or scholars in China for example is that US export controls have really only reinforced China's determination to become technologically self-sufficient and they often draw this you know analogy to the 1960s where you know China was cut off from technological support by the Soviet Union after the riff between the China and the Soviet Union and China had to respond by mobilizing the country to develop their their own nuclear weapons and space program you know whether or not that analogy is perfect or not I think it tells you something important about how China sees this challenge they see semiconductors and AI as strategic technologies that are going to require a long-term whole of country effort to master and so you know we'll have to see what the administration does in terms of the defensive side of the strategy with export controls and sanctions especially given you know we have this new constructive strategic stability framework you know will the administration for example you know moderate some of its competitive actions that's not clear what is clear however is the congress continues to show strong bipartisan interest in tightening export controls and preserving America's technological edge and so this will be space that we're going to need to watch carefully going forward I think it will help define kind of the next chapter here yeah oh god well let me let me maybe wait one comment and then welcome you to come in on this the tariff they're going to ask you your thoughts on this anyway which is you know obviously part of the question about how much China is catching up is the question of the distillation right and so there's certainly at least in DC a sense that this weed wouldn't be you know as narrow the us lead wouldn't be as narrow if we were able to solve the distillation problem you know and I think that's part of what's driving the the conversation around some of these more restrictive measures it just feels unfair and they're China's cheating and if we can crack down on distillation then that's going to help us extend our lead so curious if you have thoughts on on that piece of it and then welcome any other reactions that you have on that I mean I guess just a few observations this is more kind of a survey of the conversation happening in the field right now then any sort of normative claim about it but you know I think the concept of distillation as a technical sort of framework is itself really being hotly debated in the in the sort of computing field right now just on the on the basis of its sort of long standing use by everybody not just the Chinese to produce the types of sort of scaled capabilities that we have today right so I you know another dimension of this of course just to speak to all the elephants in the room is this sort of ongoing set of IP disputes that the labs themselves have with publishing houses and creators which is its own dimension of you know what might qualify as an intellectual property debate right now in the field and and the fact that from a legal basis it's not necessarily clear that distillation there is a sort of legal basis for claiming distillation is against anything other than maybe the terms of use you know conducted by these individual model developer organizations and their users so so I think there's there are a lot of dimensions to sort of clarify there that we're you know we're in the White Hot Center of kind of debating right now I would say I do think that the question of the security of the labs themselves is key and that I know has also been a conversation sector with respect to the work the administration has been trying to do to sort of balance innovation with national security objectives as these models have become increasingly capable and you know the other thing I would add to what Paul was mentioning is just that you know I I don't know that there's necessarily a I don't know that it has to be a mutually exclusive strategy for the US to pursue sort of limitations on on these foreign actors in you know in American companies use for example a foreign foreign model providers while also doing the types of sort of investments in the open-way ecosystem that we talked about potentially we're starting to see the American ecosystem turn their attention towards and I have a feeling that we're going to kind of I mean as with everything this administration is going to have to think about a multi sort of prong strategy that really considers these issues holistically even as we've seen in the realm of export controls on advanced summit conductors as Paul mentioned has been a really key focus not just of this administration but the last several presidents prior with respect to maintaining the sort of computing edge that America has traditionally sustained that itself isn't even enough and we've seen you know we've seen sort of the limitations of these measures as their own sort of vectors of action here in preserving American dominance it going you know as far back as the deep seek sort of developments earlier last year so that's that's just another thing to mention try not also just place for the long game the really really long game and you know this is something we mentioned in the report but Paul and I have lived to breathe this throughout our entire careers. I was in the White House during the Obama administration in 2016 when we released the first ever kind of country level AI strategy in that presidential term. And then in 2017 China released their sort of first of its kind AI strategic plan, which was called the new generation AI development plan. And it was on a multi-decadal sort of planning, you know, time horizon that they set a bunch of goals for themselves. And we're seeing them check almost every single box. So they're really kind of thinking logically about the stuff in a way that I think Washington is just going to have to increasingly start doing in order to maintain the type of advantage that we're talking about the US having traditionally had. Yeah, I want to pick up on that because it's a great point about like essentially what is the US industrial policy for AI? And there's a bunch of different levels to that, right? And there's not a lot of consistency in the US approach as administrations turnover. Right? We saw the Biden administration had a particular approach to this that was very heavy on investments and trying to ensure as much, you know, fabrication capacity for advanced tips as we could. Trump administration is taking a different approach. I think they still care about obviously the the ensuring of manufacturing, but using tariffs as a way to incentivize that, being very creative in terms of kind of the government role in some of these critical factors at times, you know, being very aggressive about taking stakes in companies and the like, right? So there's there's different industrial policy approaches, but then also there's different pieces of the overall AI ecosystem as you were just saying to her where like the investment in the government intervention maybe is more or less important, right? Like I'm not sure how much, you know, government industrial policy or dollars you need to encourage more data centers to be built like there's maybe a conversation around permitting and that sort of thing, whereas on open way it's going to be very different. And then to add another layer, and this is some of the work that that Jeff and I have been trying to push at, there's this whole question about what the kind of the commercialization and profitability strategy for AI companies looks at, depending on where they are in the stack. We literally had a report that came out that said, who will make money on AI? And like, again, like, you kind of know how the data centers are going to do it. How are all that like AI open-weight companies going to do it? How are the app developers going to do it? And so, you know, there's that's a big layer case if you look at it across all of those layers. And so as you're looking at this from, you know, bank perspective and capital allocation perspective, Tara, how do you, how do you even make sense of all of that complexity? That's really hard to. And I would probably be an investor on this thing if I had better answer to all of those questions. But I mean, Paul, I'm sure you also have some great thoughts to lend here. But, you know, I think this kind of harkens back to the beginning of this conversation wherein it is just become a set of requirements for organizations to think about upstream of them, what impacts might be had on the broader supply chain and the value chain on which they rely to think about sort of downstream consequences of policy development that may or may not be easy to anticipate, especially given the current environment and just how dynamic it is. And I think actually Emily, what you point to having worked on with Jeff in so far as the business model and this sort of the general economics of these sets of questions is really, really key. And I actually can't, I cannot remember working or living through a time where there has been what seems like so many jump balls with respect to those issues. It really feels like we're starting to see, we see entire sub industries sort of made and and broken, frankly, what feels like overnight given the pace of development coming out of the technology industry right now with every new model release. There are whole subcategories of products that you know never see the light of day. And frankly, like even grocery companies are having to pivot on a dime what their sort of entire corporate architecture and strategies look like based on the capabilities they're emerging right now. And it's just really hard to predict what all that is going to lead to. I think it's part, again, you know, just connect this back to the open weights debate playing out another interesting sort of component of it. And I think part of why it has become so visceral and symbolic of so many things to so many people in the tech industry right now is that there is a sort of debate about business model at the very heart of that, right? And so that you know, what I and Paul would probably describe as the geoeconomics of all these questions are just really they can't be underscored in their importance and sometimes are really hard to parse out and especially right now. Yeah, I would just add agree it's hard to parse Emily that Jeff that's why we pay attention to what you're working on in the geoeconomic space. Because we're constantly trying to learn we've learned a lot from our security and resiliency initiative which we announced in October and I mentioned and I think probably the biggest lesson we learn and this may not be this will not be a surprise to you Emily and Jeff, but it's not and it's not just AI if you look at you know, semiconductors or batteries or critical minerals or advanced manufacturing and effective industrial policy is really about as I said at the very beginning, you know, building long-term economic competitiveness and that requires an ecosystem approach. You know, it's not one policy. It's not one subsidy that's going to you know, drive this home. It's it's things, you know, topics that are, you know, can be a little bit less exciting like, you know, permitting reform or workforce development, infrastructure, you know, research and development, access to capital partnerships with allies and smart government policies are going to drive, you know, help private capital scale innovative companies. One thing that we know watching China is that they've done very well in terms of thinking about ecosystems rather than individual industries. If you look at EVs or batteries, even now AI as Tara has has talked about, China has brought multiple policy tools to bear simultaneously and it doesn't mean that the United States needs to, you know, copy China's model, but we can recognize I think importantly that long-term competitiveness will require that coordinated action across multiple parts of the economy. And so, you know, the government plays a role. They're going to help establish priorities and reduce friction, create the conditions for investment, but ultimately the long-term competitiveness will depend on private sector innovation, entrepreneurship and capital and really to have to work kind of hand in hand and that gets pretty complicated pretty quickly. And it should look different than each private ecosystem too. Sorry, sorry, already. No, exactly. Well, that's actually a good pivot. We've been talking a lot on sort of the US China AI competition. Obviously, there's also kind of tons of other countries in the world. They're going up with their own AI strategies. And frankly, a lot of them seem to be looking at both the United States and China right now and saying, we don't want to be dependent on either of these big countries. You know, we had the, kind of, mythos export controls a few weeks back where a lot of people kind of woke up to the right of, oh, what if we are using a USAI model and then is that support controlled and we're kind of cut off from the technology frontier here? So we see, you know, real push for sovereign AI in Europe and many other corners of the world kind of curious both what, you know, both of you would make of sovereign AI kind of in general and sort of what, you know, what would be a sort of legitimate kind of realistic sovereign AI push given poverty thing you're saying that you kind of, you need the whole ecosystem kind of, is there anyone else who can do that, frankly? But then also from, you know, the kind of investment banking lens on this, you know, where are the business opportunities in the sovereign AI push and how does JPMorgan think about that as an issue said? I can maybe kick us off and then over to you, Paul. I think on this one, what I would say is, so it depends on what we mean when we talk about sovereign AI. And there's a lot of different approaches. I mean, it feels like there's almost, there's a different approach for almost every state that's focused on this set of objectives right now, but what I'd say to your good set of questions, Jeff, is that, you know, I think there, every country right now is in in a hard spot, insofar as sort of delicately balancing, advancing their self-interest and making sure that they have resiliency and sort of durability in their own supply chains and their own ability to leverage and create contribute to the sort of construction of the broader AI value chain. And and at the same time, there's incredible concentration of certain parts of that value chain. You know, not just in the US or China, but in other sort of points of leverage in that that broader supply chain at an ecosystem level, as Paul's mentioning, whether you look at some of my conductors and the hardware supply chain there or in the model layer or otherwise, grid and electricity is another dimension here worth looking at. So, you know, what I would say is There's a huge opportunity for investment of country that are looking to get more involved in every aspect of that, and they should lean in there. But countries also have to be realistic to the fact that this is a borderless set of technologies. And there's a huge amount of inter-reliance between states and regions and allies as we think about constructing capabilities here. And I think this is another place where the sensitivity and the open-weights debate has really shone a light on some of these issues, right? Because part of what is at play there is, as you mentioned, this sort of belt and road-like approach that China's taking to trying to defuse their version of an AI supply chain and as many parts of the world as possible, which speaks to the competition they're having with the United States to do the same. But also the ability of powers everywhere to leverage the cutting edge, which if there is a sort of restriction of that, by the states that have production of that edge, a real issue of consequence there from a geopolitical perspective. So I know that doesn't offer really concrete answers, but I guess just a couple of reflections. And I note that there's really, I think the game is very far from over, we're still in the early enemies here. There's still a huge opportunity for countries of all sizes and scales to really kind of figure out how they're going to pick their version of the winners here and get involved. Yeah, I would just add, I agree with everything Tara said. I mean, she mentioned that China's belt and road approach to diffusing AI across the international community. I mean, the bottom line is, AI is becoming another form of strategic infrastructure. And I think, you know, I agree, it's, it countries want AI systems. They can trust, they can govern, they can adapt to their own laws. And so, you know, countries do not want to become wholly dependent on any single AI ecosystem. And so, just as countries think about trusted telecommunication networks or trusted supply chains, they're going to increasingly think about trusted AI ecosystems. That will be an important element of international partnerships going forward. Is this about competing with China? I think partly, but I think it is broader than that. Many countries just simply want options. Helping allies build trusted ecosystems will strengthen resilience, regardless of who's trying to balance against. But I think, you know, we're going to continue to see this play an important role, this idea of sovereign AI in countries across the world. China's open weighted models look more attractive today, especially after, you know, some of the models were restricted through export controls like Fable. But I suspect that as Chinese open weighted models become more capable, and we're already seeing reports about, you know, the Chinese government thinking about whether it wants to restrict access to its most capable open weighted model. So it may not be that the open weighted models continue to be unrestricted going forward. Can I make one more plug just as a wank here from a, from a public policy perspective? I think one thing that is under addressed in this conversation is often the fact that for middle powers, the biggest impediment to progress for them is going to be the risk of a highly fragmented sort of overlapping global standard system. And so, you know, as our own team, really thinking about trying to make sure that a bank like ours can operate across the 100 plus countries that we do business in every day, you know, we're really, our eyes very closely on the ball related to advocating for globally interoperable frameworks and tech neutral enforcement so that countries can really safely integrate AI strategies with an eye towards making sure the innovation isn't stifled and that there is, you know, you don't see data localization or interoperability challenges really hamper that progress of those goals. And I think sometimes those things seem at odds with one another, but you know, the way that we think about it increasingly is in really trying to emphasize the need for that, especially for, for the global economy in which global organizations like us all. You mentioned the security and resiliency initiatives. So this massive fun for investing in industries that are important for for economic and national security of the United States. And obviously that seems like a great idea. I am curious as you're thinking about the intersection of geopolitics and investments strategy, like a lot of the areas, a lot of the critical factors where we have these deep vulnerabilities in the US are because the economics of the sector don't really work if you're not a Chinese company. And the bank is thinking about trying to make investments in something like batteries or critical minerals or legacy chips. How do you get around what could potentially be kind of a massive investment barrier. Yeah, great, great question. First of all, just stepping back a little bit. Our security and resiliency initiative, I would say, reflects a single simple idea. And that is economic resilience and national competitiveness are becoming increasingly intertwined historically companies have focused primarily on efficiency. And today they're balancing efficiency with resilience, security and long term competitiveness. AI fits naturally into that because as we've been talking about it's not just another technology sector. It's becoming a foundational capability that touches manufacturing, healthcare, energy, logistics, finance and scientific discovery and more. And so, you know, financing these transactions is a huge challenge as you're implying. And whether you're talking about semiconductors or data centers or power generation, advanced manufacturing, private capital will be needed. And that's where the financial institutions can play an important role. But government is increasingly looking to private institutions, you know, where they're having to do because for because as you suggest, Emily, the market is not going to provide 100% of the solution. Right. Governments need to intervene through equity or off-take agreements or price floors to reduce the risk for private capital to create that long, longer term demand so that private capital then come in and be part of the responding to the challenges of building capacity in these critical sectors. And so that's where where markets don't provide 100% of that solution. Increasingly, we're going to need to look to public private partnerships. And I think you're we're beginning to see through initiatives like JP Morgan security and resiliency initiative, other financial institutions are beginning to play a role other companies as well. A greater alignment between government and the private sector in those sectors, Emily, that you talk about where the market's not going to provide the solution. And again, I think one of the themes of our work is that resilience is not simply reducing vulnerabilities. It's about creating up new opportunities for investment, innovation and long term growth as well. Great. Terry, anything you want to add on a sort of AI or tech sectors that are relevant to SRI? No, I mean, I think Paul pretty well covered it. I think the only thing to add is a point made obvious now by this entire conversation, I think, which is that all of these capabilities that we're talking about are deeply intertwined in a lot of the different sort of strategic pillars of that work. I think that's a recognition that we have that we've been leaning really hard into as a bank that's really striving to make the types of durable investments in long term American resiliency that Paul's describing here. Thank you both for joining us. It's been a fantastic conversation as listeners will know we like to kind of close out every episode by asking our guests for recommendations, either a book, a podcast, an article, you know, something that you'd encourage our listeners to go check out to learn more about everything we talked about today. So maybe Paul first then we'll close out Sarah. Sure. Thanks for the question Jeff. I guess let me if I can if I'm not breaking any rules, I'll recommend to the first one I thought about, which is probably already been recommended, which is why I came up with the second one is, you know, Chris Miller's chip war. It's a great introduction, you know, into why semiconductors have become so central to both economic competitiveness and national security. It's not exclusively about AI per se, but it does give I think much of the strategic context for understanding today's competition. And the second book I would recommend is break neck by Dan Wong, you know, we had Dan speak in May in Shanghai at our China global summit, which we do every year in Shanghai in May. And he does a really good job, I think capturing the speed of technological change in the broader dynamics shaping US China competition, including on AI. So that would be the second book. Okay. And for my part, this one is a little. self-serving, but I have one that is probably obscure enough that he listeners will have otherwise heard of it, which is an article that we as JPMorgan actually just published by Michael Semble is to use the Termin of our Market Investment Strategy team in our Assuming Wealth Management Division. His team does great research on the markets and our team as well as our security teams in the bank just cladred with him to write one called Patch McGetten that just came out a couple of weeks ago. We could spend probably entire podcast season talking about the sort of geo-security, national security dimensions of all these advanced AI capabilities that are emerging, which we didn't even really scrape the surface of, I realize, in this conversation, but I would be remiss not to mention that dimension of all of this with respect to US China and more broadly than that. It's a great 30-page survey of basically everything that we think is changing about the landscape right now, given the compression of attacker time to exploit, given just the proliferation of what effectively amount to state grade cyber capabilities now being put in the hands of everyday actors by a lot of these frontier models that have been released in the last five months or so. So I highly recommend it. It's relevant to a lot of what we discussed through today and come find me later from our discussion on any of those topics. Yeah, that sounds fantastic. I'm excited to go read it right now. So this has been great. Thanks for joining. Thanks for listening to The RISP Business, an economic security podcast from the Energy, Economic and Security Program at the Center for a New American Security. Have any questions or comments? Get in touch at TheRISP Business at cnas.org, the nerdyer, the better. And find us wherever you get your podcasts, every other Thursday.