How AI Regulation Could Help the Public—and Investors Too

0:00This episode is supported by Anthropic, the team behind Claude. Every business wants to move faster. Few can afford to be wrong. That's why the brands you trust also trust Claude. See what's next at claude.ai.

0:17Hi, Miriam.

0:18Telis Demos: Hi, Telis. Today, we're going to step back and talk about something that I think really looms over a lot of the conversations we've had about AI, artificial intelligence, which is what does the regulatory picture look like in the future? And there's a few different ways that I think we've talked about this. You have obviously the national security human existence sort of considerations of what - Oh, that little thing. What do we need to do to make sure that AI doesn't run away and kill us all? And then we have, I think what you've talked about in the show in the past is like, you know, the way that the AI companies sometimes talk about working together to address these things, OK, that sounds that maybe that's reassuring to investors, but it also sounds a bit collusive.

1:02Telis Demos: And, you know, what antitrust concerns are raised by that sort of behavior? Yeah, I mean, I was very puzzled when the two CEOs of Anthropic and OpenAI came out and said, we need to slow down after this risk of ending humanity was raised. Because I was thinking, what are investors supposed to think about this? Is it marketing? Is this them saying our models are so powerful, look what they might be able to do? Is it a true desire to say, like, we need some rules of the road here? Or is it them signaling to each other in the way that very disciplined competitors do? If you slow down, I'll slow down.

1:43And then we can both, you know, maybe benefit from that. So to help investors think about this, because I think this is an important input into any kind of long term, you know, decision you might be making about whether to go, you know, all in on AI, whether to hedge some of your bets.

1:59Telis Demos: We've brought in an expert on this, somebody who used to work at the Department of Justice, Jonathan Cantor, who we have a pretty wide ranging conversation with about how he thinks AI regulation might shape up in the near future. Before we get to that, though, I think we should talk about some other AI related news that occurred this past week. SpaceX reported earnings and it actually reported a profit, which was a new thing. And driving that was the AI business, which, as we talked about on previous episodes of this show, is a huge component of the future potential valuation of this company. If you looked at the way that, you know, SpaceX was determining its own potential TAM, total addressable market, the AI business was a huge portion of that.

2:47And so the fact that it grew a lot, thanks in part to this merger with XAI, I think was a positive thing for investors. The big news, though, was that Musk wants to merge Tesla and SpaceX. And because, you know, he wants it, it seems likely that that's, you know, going to happen. And what are the implications for that? We're not so sure. Maybe we'll have a future guest on this show come and talk about that.

3:12Telis Demos: And of course, there's the Elon Musk-ness of it, right? Oh, we need artificial intelligence informing Tesla, which is now going to make flying cars that are going to fly to Mars and blah, blah, blah. Right. Like there's a vision element of it. And of course, I'm sure there's a personal financial element in it for him as well. But there's also the bigger companies can generally raise more money to invest in AI. And so, you know, a company that goes from, you know, a trillion dollars to whatever the combined entity would be, you know, a three trillion dollar company just has that much more financial might to throw around in this.

3:52Telis Demos: And, you know, XAI has not generally been the leader thus far in frontier models and things like that. So from that point of view, you know, it's in some ways maybe a defensive move. Yeah, I mean, and SpaceX is already spending a ton on CapEx, like all of these companies are. CapEx exploded to nearly$20 billion, up from$6.9 billion in the same period a year earlier. So it's already marketing plans to sell new debt. And you've been writing a lot about these AI bonds that companies are issuing. Can you talk a little bit more about what's going on there? Yeah, actually, it's funny when you said AI's been in the news this past week and raised like a sort of positive development.

4:33Telis Demos: I was thinking about all the negative things that have happened in AI sort of markets over the last couple of weeks. You've had CoreWeave. You've seen their stock decline and their bonds also take a hit. You had a couple of IPOs of companies in that space. Sometimes they're called neoclouds. I mean, these are companies that are developing kind of cloud computing businesses really aimed at AI applications. Some of those terms are just like beyond me. I'm like, OK, someday maybe I'll learn what a neocloud is. Well, you know, you kind of say that cheekily, but like I think that's a little bit of what is going on in the market because the credit markets, the debt markets that have evolved around this stuff are all pretty new.

5:13Telis Demos: You know, for a long time, if you wanted to build something, you'd go to the bank and get a loan for it. You generally wouldn't go to the bond market because the bond market needs to see something very real before it's willing to give all their money to you. But, you know, around a year ago, companies started actually going to the bond market to raise money to go out and construct a data center. And that's because they were already these huge tech behemoths that we knew and trusted, from a financial standpoint at least, to have the wherewithal to support this debt. That's part of it. But also they had kind of tapped out some of the banks on their lending appetite for this stuff.

5:43Telis Demos: So they needed to go to the debt market. And then since then, you've also had the emergence of the market for chip financing, right? Oh, not only do you need to build a data center, you need to buy all these chips. So you take out a loan to buy the chips, and the loan is secured by the chips. And then you've got the data center, which has contracts with these companies, and it's secured by those contracts. But all these things are relatively new. And I don't know if the market really has a firm sense of how to price these things. And I think there are a lot of very smart, clever credit investors who are looking at each of these pieces.

6:16Telis Demos: And this is kind of what I was writing about. And they're looking at each of these different layers of financing and say, OK, this is a loan, but it's secured by a contract. This is a loan and secured by a data center. Not all AI bonds are created equal. Yes. Yeah. So I think part of what we're seeing is part of it is, you know, maybe there's just kind of bearishness about, oh, maybe the market for these things is not going to be quite as big as we thought in the near future. Or maybe these companies are over levered. They borrow too much. But I think there's also just volatility that comes from a market that is so, so large, but also so nascent.

6:48Telis Demos: I mean, it's like a giant, I don't know if this sounds like, it's like a giant baby just sort of like hanging out. No one really knows what to do with it. Feed me more. It's cute. We all wanted to grow up, but it's also like gigantic and it's eating everything in the house. Like it's an unwieldy thing. And so part of what's happening is the market just kind of working this through. Speaking of markets still figuring things out, I want to pivot back to AI regulation, the topic of our conversation today. And, you know, obviously data centers, as we talked about in our last week's episode, are, you know, a big focus politically.

7:24And there are efforts by states to go and renegotiate those agreements. And so the states have become very involved in that and making sure that their constituents are getting more out of these data center agreements. You know, then there are also some states that are taking steps to actually try to regulate the models like these companies like Anthropic and OpenAI. And that's interesting, too, because the states can often be, you know, the place where some of this legislation begins. We also have some lawmakers in Washington who are starting to talk about this.

7:57Telis Demos: I think that's the bigger question mark at this point is the federal level. What's going to happen there? Yeah. And our guest today, Jonathan Cantor, has a pretty clear framework for how he thinks lawmakers should begin to regulate AI. And as we said, he was previously the Justice Department's assistant attorney general for the Antitrust Division. And he's currently a law professor at Washington University in St. Louis and a tech policy professor at Carnegie Mellon. It's a really interesting conversation. And we have that coming up in a bit.

8:52Welcome to the show, Jonathan. It's great to be here. So a recent WSJ survey found that 57 percent of Americans want the U.S. government to regulate AI immediately. And, you know, we've heard major candidates from both parties talking about this. So would you say that it is inevitable at this point? It's inevitable, but it's not imminent. Explain. Okay. So right now we are in a world where it's up to Congress. It is on the shoulders of Congress. The Supreme Court has made it more difficult for regulatory agencies to impose rules. There's something called the major questions doctrine that the Supreme Court pretty much created out of thin air.

9:33And that requires major questions to be handled by Congress and legislation, not by regulation or other kinds of regulatory rulemaking. So at the time when we need Congress the most, we have it the least because Congress is effectively dysfunctional and incapable of imposing any kind of meaningful rules of the road. Think about this. We are in 2026, 25 plus years into the Internet era. We still do not have comprehensive national privacy legislation. And that's not for lack of interest, either by the public or from Congress. So one of the solutions or one of the prerequisites to effective oversight, effective rules, effective regulation in the era of AI is going to require a functional Congress.

10:26And in a world of the major questions doctrine, in a world of gerrymandering, in a world of Citizens United, it is going to be extremely difficult to do. And so that is why we're in a situation where, yes, it's inevitable, has to be. This technology is too transformative for it not to have some sort of rules, but it is probably not imminent because at least for the next couple of years, we don't have an effective legislative body that has to be the source of those rules.

10:55Telis Demos: So you mentioned privacy legislation. Why is that a pillar of what AI regulation would look like? Why privacy? So it's a rear view mirror issue to some degree. It's relevant today and it's relevant going forward. But the reason I mention it is because think about how much data companies on the Internet collect every day. Think about how much personal information that they store and use. Think about how many privacy issues are implicated by social media, by targeted advertising. And we still have no comprehensive legislation. We're still largely relying on the FTC Act, which was created over 100 years ago.

11:35To regulate the Internet. Exactly. Yeah. Which was the technology of yesteryear now. Now we're on to a new technology. And again, if it were for a lack of consensus, that would be one thing. But there is a consensus largely that we need something, yet we still can't do it.

11:53Telis Demos: But what is different about AI that the existing rules, right, the things that protect the privacy of my information today when I'm sending information out to a bank or, you know, sharing information with in my email and things like that. What is different about AI that requires a different framework than the things that legislate or regulate the Internet today? So with the exception of certain kinds of regulated industries like banking, health care, we don't have privacy rules. And so the rules are extremely loose. We don't really have rules to regulate the Internet is kind of what you're saying.

12:28Exactly. If anything, we have rules not to regulate the Internet. So Section 230 in immunity actually gave the Internet companies a free pass with respect to lots of rules and made it easier for them to innovate without the concerns about liability. And I think we've learned the hard way that the results are not necessarily positive for the public. And so I think there's a sense this time around that AI is so even more potentially transformational than the Internet and that we need something in place to protect the public. And there is a sense of urgency coming from the public, coming from the companies themselves.

13:05We don't necessarily have a fair pathway to get there. You mentioned Citizens United as being a reason why Congress is ineffectual. Can you explain what that decision did and why it's kind of tied Congress's hands? So Citizens United basically opened the floodgates to money in politics. It made it so that companies can essentially pour so much money into campaigns that the members of Congress have largely become more responsive to fundraising from companies and industries with national interest than the votes of their local jurisdictions and constituencies. And so it's altered the framework of politics.

13:47The same time gerrymandering, which essentially carves out districts by partisans, by party, has made it so that all politics, which used to be local, is now national. And in order to get elected, you sort of have to run to the party's extremes rather than the needs of your local district, because that's the only way you can get access to the money and the political support you need. So we have a situation where, again, Congress is dysfunctional and it's not completely organic. We've let it get that way as a society. And if we really want to tackle the AI issue, I'm afraid that at least to some degree we're going to have to tackle the political dysfunction as well.

14:30Telis Demos: But you are hearing both parties. You know, we were just talking about, you know, the governor of Pennsylvania, the governor of Texas, right? Different ends of the political spectrum, but they both kind of shifted to be sort of delivering a, you know, anti-data center message. It seems to be a winning political argument in the midterms. What are the complexities of the politics of AI? So let's piece this apart. I do think the states, the laboratories of democracy provide a pathway here. But interestingly, if you go rewind about a year, there was legislation in Congress to create a moratorium, a preemption for 10 years for states to impose any rules and regulations on AI.

15:12Wow. I don't think I followed that. Thankfully, it didn't pass, but it almost did. And so now we have the states stepping forward. And here's, I think, the data centers in large part are a proxy for the broader concern and lack of trust that the public has with respect to tech companies. And this goes back to the Internet era, social media and all these areas where they were promised a rose garden, but they got social media addiction. And so part of what's happening here is the public is saying, hey, I think I'm being conned. I think these big data centers look big, but they don't necessarily employ people.

15:54I challenge people all the time to think of a person that they know who's actually worked in a data center. And I don't think yet I've gotten a response for somebody outside the tech industry. So they're not providing the jobs that people thought they would be. Exactly. And so what's happening, and I think the public is picking up on this, is you have these big boxes that consume lots of natural resources that potentially tax the grid but don't employ people. And then the wealth from the products and services go back to another area. And so I know there's a lot of tendency politically to say, well, these are like the railroads and the electrical grid and our telecommunications network.

16:35And to some degree, there's some truth to that. But think about this. When you built out roads, it allowed trucks and cars and freight to get to a local community so people can work there, allowed them to travel back and forth to their jobs. Same thing with railroads. Telecommunications network made those places more habitable. And so really, if we do need, and it's an if, but if we actually need these data centers in order to build out our infrastructure, then in my view, they should be accompanied by a commitment to those local markets by the tenants of those data centers, not just to share the tax revenue, but to actually employ people, move your accounting function, move your legal function, build office buildings that bring people to work every day, bring coders there, invest in those communities.

17:21alongside the location of the data center. Those are the kinds of ways that will show greater commitment to those communities and I think will probably make states a little bit more hospitable. But right now, I think there is a great reluctance to invite these data centers in, in part because I think the public feels it's really a symptom of the K-shaped economy. They feel like the burden will be shouldered by the local community, but the benefit will go somewhere else. So if you could wave your magic wand and somehow make Congress functional, where do you think lawmakers should begin when it comes to regulating AI without curbing innovation and growth?

18:02Yes. So I would start by establishing duties of care and duties of loyalty. Necessity is the mother of all invention, right? And so we want companies to have to build products that are safe and secure. And so establishing, one, duties of care, safety and security, privacy, creator's rights, a duty to ship a product that not only works, but that doesn't harm people, embed those. Then create duties of loyalty. We talk about agents, right, autonomous bots that work on your behalf. Who are they working for? Great question. And so if there's a duty of loyalty that if you're my agent, you need to just like if you're my lawyer, my doctor, you have an obligation to me as the as the customer, as the patient, as the client.

18:50And I think establishing that same kind of duty of loyalty for for AI agents will be necessary. And then the third pillar of that is our consequences that exceed the benefits. Right. And this is absolutely critical, because if you just have fines 10 or 15 years later, after lots of litigation, you're not going to get it very far. And so the consequences of violating those duties of care or those duties of loyalty need to be significant enough that shareholders demand that companies comply. And so that's everything from big fines to when I gave a speech when I talked about robot jail, right, where you actually take the products off the market and essentially say these agents or these products, if they're violating their duties of care or duties of loyalty, can no longer be in service, at least for a period of time or perhaps forever.

19:45Robot jail. I love that. Yeah. And so these are the kinds of things you want to start with. But there's more.

19:51Telis Demos: Well, so you mentioned being responsible for the output of your product, right? In the case of AI, though, when you have agents that are thinking autonomously to some extent, how do you hold a company responsible for something they create that then has some unintended consequence that they maybe realistically would never have envisioned or envisioned users using it in a certain way? Can those things actually coexist? Absolutely. Dog bites the neighbor. Okay. Right? The owner of the dog is responsible. Somebody slips on your sidewalk out in front of your house, right? You had an obligation to shovel it, right?

20:33If you're building something and putting it out in the market, then you have an obligation to make sure that that product is safe and secure. You have a duty of care, right? Just like the owner of a dog or a property owner with their walkway in front. This is embedded in our legal system. It's workable. Yeah, there are going to be questions about who's liable.

20:51Telis Demos: And it sounds rather untenable for an AI company, right, to say that, you know, we're going to put out this agent, that the whole intention of it is that it will sort of grow and learn and behave in new and hopefully useful ways. But if it does something wrong, I mean, would anybody feel comfortable putting an agent out there under that legal standard? We do have employees all the time. Right. And so if you work at a company, you get trained. Right. You can't break the law. You can't bribe officials. You have to maintain books that are accurate. You have to obey employment rules, right? We do that all the time.

21:26And then when you create compliance mechanisms to make sure that your employees adhere to those rules and regulations because that keeps the parent company out of the soup, right? In a world where there's going to be autonomy, we should have the same kind of principles that flow through. And guess what? If the consequences are significant enough, companies will innovate their way around the problem. That's what we want. We don't want to build their products for them as a government. We want to say you should go off and build them yourselves, but innovate around safety, security, loyalty and all of these principles.

21:57These are engineering problems and they are solvable. So, I mean, thinking about the Internet, which was, as we've discussed, we didn't really regulate all that much. I mean, those companies that were not very regulated grew to become some of the biggest companies in our market. Would they have become so big? Would they be so valuable had we regulated them? I argue that they would be more valuable and they would be more successful and they would be more useful. Imagine if we had social media that was built for child safety, that had mental health concerns baked into its design as opposed to just engagement.

22:36Necessity is the mother of invention. If these principles, it's easier to mold clay than it is to chisel marble. We're at the outset. We're at the formative phases of these technologies. If we build them to be better and safer and more responsible, not only will they be more useful, but they will be more innovative and they will be more profitable. So I think I can just hear like the technologists saying, oh, Congress is not equipped to be writing these rules of the road. They don't understand the technology. They didn't understand the Internet and they don't understand AI. How are these lawmakers going to go out and write the rules of AI?

23:17What do you say to this idea of a knowledge gap between lawmakers and the technology? It's another solvable problem. So when I arrived at the Department of Justice in 2021, one of the first things I did was announce that I was going to build a data science unit, which we did in 2022. And this is before ChatGPT was a household name. And what we did is we hired in PhD data scientists, technologists, people who understood how the technology works so that when we had to start investigating, we had the team to do it. So we had a major investigation into a company that used algorithms to set rental rates.

23:54And so rather than using the traditional tools for investigation, we were able to subpoena the algorithm and have our data scientists do a linear review of the code, figure out exactly what was happening under the hood, and then bring a case that has so far turned out to be successful. These are the kinds of things we need to make our government functional in this era. But I should say that's not necessary to get started, right? The basics, which is ship a safe product, ship a product that respects creators' rights, ship a product that respects privacy, ship a product that doesn't break into someone else's servers.

24:31Creating those basic duties of care is something that Congress can do today with its current range of expertise. But we are going to have to upgrade the technical capability and tools inside our government for a modern era.

24:45Telis Demos: All right, hold that thought. We're going to take a quick break. When we come back, more with Jonathan Cantor.

24:58The Alternatives universe is vast. Knowing where to focus matters. From private equity and private credit to hedge funds and real assets, Morgan Stanley Investment Management helps advisors see the solutions clearly. With more than 40 years of alternatives experience and specialized global expertise, we address liquidity, growth, and diversification with conviction backed by capability. Explore Morgan Stanley Investment Management's approach to alternatives at morganstanley.com slash im slash alternatives.

25:31Telis Demos: try and put yourself in the shoes of of an investor in one of these companies or you know i guess we will all be confronted with that if and when a company like anthropic goes public would you want to own a company that has this potential liability in front of it do you think that that's something that will dramatically impact the valuation and investor interest in these companies? No, not at all. Companies live with liability exposure, right? You're the Wall Street Journal. If you print something that is slanderous or libelous, you're responsible. But we also have a very well-defined and, you know, decades and decades of what protects us and what doesn't, how cases have been decided.

26:12Telis Demos: This would be pretty unknown at this point. We have to start somewhere, right? And if we don't start now, it's going to be a lot harder to do later. Like I said, it's easier to mold clay than it is to chisel marble. And so let's start with some of the basics. Build a safe product that doesn't harm people. Build a safe product that doesn't steal. Build a safe product that respects privacy. And then we can go from there. But you got to start somewhere. And the first principles, if you don't know where you want to go, you can't get there. Right. And a lot of the debates right now over speed limits, how fast should we go?

26:47How slow should we go? But if you don't know where you're going, what's the point of how fast or slow you're going? Speaking of that and how fast and slow, when investors hear Anthropic and OpenAI's leaders saying we need to slow down and they're both saying it in unison, what are we to make of that? Is that market signaling on their part? Is that disingenuous or is that a real cry for regulation? I think it's all of the above. So I believe they believe that their products can be quite harmful and they want to put products on the market that are safe. Great. The reality is, and I don't say this in a way that suggests there's anything wrong, right?

27:32These companies are optimizing for shareholder value, right? That is the market reality for for-profit businesses. And that's fine. That's our system.

27:40Telis Demos: Well, that's certainly how many shareholders would want them to behave. Exactly. And so and so how they're legally obligated to behave. Yeah. Right. So, right. Exactly. You know, there are certain things that the public and the government should have responsibility for. The end of the day, they have shareholder obligations. And so what you want to do is you want to create duties such that those fulfilling those shareholder obligations requires safe and secure products and failure to do so would create significant consequences for shareholders such that they will hold their companies accountable.

28:07This will make the products better, they're still going to be wildly successful. They're still going to make money hand over fist. There's still going to be plenty of opportunity for great returns. I think what we're hearing, including from the shareholders, is just tell us where the lines on the road are. Tell us where the stop signs and the traffic lights are, and then make the consequence significant enough so that we can adhere to them. But when I hear that they want to slow down, there's a degree of suspicion, which is if you really think you're building something harmful, you don't have an obligation to ship it, right?

28:37If Boeing has a door flying off midair, it doesn't need Airbus to ground its planes in order to figure out how to fix its problem, right? And so I think we want, my view is we want companies competing to develop safe and secure products, and we want to give them the incentive to do that. And I think that is good for shareholder returns. It's good for the public. And it's going to be necessary if we want products that are safe and secure. I think we've learned our lesson on the first generation of the internet that just some rules are necessary so that we bake into the development of the products some safety.

29:19Telis Demos: So we've been talking about regulation a lot from kind of the user's point of view. I think there's a couple other areas, which is in order to say that we have a fair competitive market? Because at the end of the day, too, these are products. We want them to be priced appropriately, right? If we're all using them in our daily lives. We at the Wall Street Journal have written a lot about how these companies are very intermeshed, right? We have a company that sells you the chips, will finance you to buy the chips, and then they also have a relationship with the user who will be using your compute services and things like that.

29:52Telis Demos: These circularity, financial circularity, does any of this raise antitrust concerns for you that these companies are are essentially all kind of rowing in the same direction? Yes. Right? Of course it does. I mean, this is why the antitrust laws exist. It reminds you a lot of, you know, the trusts from antitrust were these organizations that had their, you know, kind of ownership interests all over the economy. And that circular ownership was actually part of what led to the creation of the antitrust laws. And so we should be concerned from a competition and from an economic resiliency perspective perspective when we see all this circular financing, when we see these minority interests.

30:33I mean, if you take the dots and you start connecting among various different companies, you realize they all own part of each other. That is a bit concerning. I think we want to make sure that we have free and fair competition so that companies are going to duke it out on the marketplace. And once you start compromising your interest or have conflicts of interest, Just it can create some messiness. So we've talked about competition among U.S. companies, but we haven't talked about a thing that the Trump administration brings up a lot, which is competition with China. How how could regulating our own A.I.

31:09industry hold us back from competing with China? Is that a fear? Because that's something that's talked about a lot. Yeah, it's a red herring in my view. So the best way to compete against China is to keep our people safe and encourage our companies to build safe and secure products. Let's be real about it. China imposes all sorts of rules and regulations on how people in its country use. It's a very regulated society. Yes. We are less regulated, which is fine. But let's not pretend that, you know, we need to tolerate our companies in the United States shipping harmful products in order to compete against China.

31:46To what end? There are very real issues, let me say, though, with respect to national security and defense and cyber. That should be on the shoulders of the government, right? We need to rethink our defense infrastructure so that we are suited to address that.

32:04Telis Demos: We've been talking about competition at the relatively prosaic level of people's financial lives and even their privacy. You know, how can we regulate things that are not even necessarily being sold as products, but just sort of being built in a lab somewhere that we worry could get out and have horrific consequences? What kind of regulatory framework would be sort of precedential or possible here? Yeah. I mean, you can look, for example, with some of the things like nuclear technology or some of the more advanced pharmaceutical development where you have to make sure that products are safe and secure and we have review structures in place.

32:42Those are models that we can deploy potentially. And it's going to be tough to figure it out. And so I'm not suggesting that simply by imposing a duty of care and duty of loyalty that we're going to solve all those problems. I think we're going to need, you know, that's where we're going to need, you know, agencies that have greater skill and authority. I think sometimes this sort of doomsday scenarios of these things are going to be like the Terminator and they're going to kill us. That's distracting. Right. The reality is like these products can do very harmful things. Right. And they are very potent and they have the capability to do a lot of harm.

33:19they're not going to rise up and kill us tomorrow. Like, I think we need to, like, take a little bit of a step back.

33:24Telis Demos: I think that might distract us from the much more realistic, imminent threats to our privacy or our, you know, personalization. Yeah. The reality is, like, let's figure out how to engineer into these questions and problems and solve them with a dose of reality today. If we make everything about this sort of science fiction movie-like problem, it will just create paralysis. And I think I worry that that's what's happening. We need to go back to first principles. What are the basics, like safe, secure products that don't violate the law, that don't break in, that respect creators' rights, that respect privacy, that have a duty of loyalty to the person using it?

34:06You can start with some of those basics. It may not solve all of our problems, but at least it's a step in the right direction. Jonathan Cantor, thank you so much for joining us. This has been a fascinating conversation. Awesome. Thank you.

34:20Telis Demos: And that's everything you need to know to take on your week. This show is produced by Alexis Moore and Michael LaVelle. Michael LaVelle is our sound designer. He also wrote our theme music. Aisha Al-Muslim is our development producer. Chris Zinsley is our deputy editor. Leetal Malad is our senior director of shows. And Samantha Hennig is the Wall Street Journal's head of multimedia. For even more, head to WSJ.com. I'm Talis Demos. And I'm Miriam Gottfried. Until next time. Happy to have him. I don't like it when it's empty because it's so tempting. And now it's like suddenly I'm like parched.

35:01This episode is supported by Anthropic, the team behind Claude. Every business wants to move faster. Few can afford to be wrong. That's why nine of the Fortune 10 run on Claude for the work they want to get right. Safety and accuracy are not just line items for Anthropic. That's their whole mission. See what's next at Claude.ai.