Hi and welcome to Ones and Twos, FP's economics podcast. Every week we take a couple data points, use them to try to explain the world. I'm Cameron Abati, FP's deputy editor with you in Berlin, Germany. Joining us as always, FP's economics columnist and Columbia University professor Adam Tooze this week in New York. Hi Adam.
So, we have one data point for you all this week. But first, a preview of our episode that's coming next week. So, in one week from today, our podcast will be dropping as usual at the usual time, but we thought we'd let you know already that we'll be discussing a specific book. It's already out. It is Francis Fukuyama's memoir. It's called In the Realm of the Last Man. We'll use it as an occasion to talk about Fukuyama as a thinker, his life, but we'll be placing it in the context of the book itself. So, if you're interested, it could be useful, interesting to pick up that book and read it ahead of time. It's called In the Realm of the Last Man. You can probably find it at your local bookstore. If you don't happen to read the book, I think the episode will be interesting in either case, so don't fret. But again, if you're interested, you can look it up. Otherwise, back to our data point. That is 109. That is the number of separate artificial intelligence laws that the United States in its various individual states has passed just this year as of July 1st, give or take. And if you extend that to the past 18 months, you'd actually apparently get some 250 laws aimed at regulating AI. But again, that's all been happening in the states. And more recently, these past couple weeks, momentum has been building for the federal government to join the fray to ostensibly stop a potential apocalypse as some AI industry players themselves have been warning about. Though, President Trump still seems intent on standing in the way of regulation. And all of this is really a symptom of a bigger set of questions because there are a lot of deep tensions underneath this entire concept of AI regulation. It puts a spotlight on questions of the structure of the AI marketplace, the efficacy of existing liability laws, and the relationship of apocalyptic scenarios to capitalism more generally. So, Adam, we thought we'd unpack some of this. And to start off with, in what ways could we expect or not expect liability law and criminal law to mitigate the dangers of AI as we do, say, with many other hazardous technological developments that we've seen, you know, essentially regulated under common law?
Yeah, this is a great way into this, I think, because I like it because it's somewhat deflationary, right? We're saying because there's one view on AI which says it's going to change the world completely and so we need a radically new regulatory regime that can also be a kind of holding action move because basically that's going to take ages to put in place whereas if you say well look there's already law there why can't we just use that we're pushing the immediacy of action as a possibility. Broadly speaking this could be done probably under negligence where an actor fails to anticipate, you know, foreseeable risk. Product liability could do this or maybe nuisance, professional malpractice, contract and consumer protection law, or there's lots of different angles you could go down. And the full ambiguity of this is brought out by the fact that, you know, after the top guys at OpenAI and Anthropic came out with these apocalyptic warnings and now they want to pace the frontier, not only did the Trump administration dissent, but Jensen Huang of Nvidia and Mark Zuckerberg of Meta also dissented from this position and they argued for saying, well, look, Jensen put it very very clearly, right? The market forces are already there. We don't need any new laws. We just don't need new regulations. If you build a product or a service and you're not confident in its functionality, capability or safety, then don't release it. You know, Jensen Huang says and Zuckerberg likewise, right? Every lab has the responsibility and incentive to move at the pace required to train its models safely and the ability to take its own actions to ensure that happens. Labs face significant liability if their models cause harm. So they have a strong incentive to prevent this as well. This is Mark Zuckerberg. And Zuckerberg after all has reason to say this because over the course of this year, we didn't do an episode about it, but it's really dramatic. Meta has come to a 16.7 billion dollar settlement with 29 states over issues to do with algorithmic manipulation of underage of teenaged consumers of its social media platforms. Right? So that's an acknowledgement by Zuckerberg no less that this law, the existing legal framework can actually do a lot of the work and his company has felt the force of this and in legal theory circles negligence is the way that people are going. So there's, you know, articles in the Yale Law Review like this is, you know, this is a high status frontier of legal thinking and though AI advocates will say things like well, you know, the thing about algorithms is they're not deterministic. So it's not true that if you put certain inputs in you're always going to get the same outputs and does this pose questions to negligence. The broader position I think is clearly going to dominate which is if you build a machine as unpredictable as that and put it out in the market with the potential for really dramatic consequences you're clearly negligent in doing so. So a short answer to your question is yes it's already been done with regard to Meta, with regard to, you know, base AI stuff, with regard to social media algos. Jensen Huang is on board with this view as well if, you know, I wouldn't release hardware that potentially could blow the world up because that clearly would be a product liability issue. The question is like is it going to work on the large scale with regards to AI. I think that's all the questions. Is it really going to be an effective way of dealing with, are you really going to be able to pinpoint causation? In the case of AI models, it's rather different from the mass effects of social media. How are you going to cope with the fact that the harms are probabilistic and they're tiny probabilities, right? This isn't like cancer causing cigarettes or something like that. It's, you know, a one in, well, they don't really know, but it's small percentages chance of blowing up the world. How does one really deal with that consequence? Furthermore, if you really do blow up civilization, will there be law courts to deal with the cases in the aftermath? Furthermore, the liabilities potentially are so huge that how would you rightsize them for corporate balance sheets? One of the thoughts here is that this is actually a play to get the corporate balance sheets de facto put on the balance sheets of the US government. So by pushing this kind of liability too far, you'd end up in that direction. And furthermore, the events that we're talking about are ones which have irreversibility and which can't really be properly compensated for. If we really are talking about extinction level events or even just really dramatic, you know, I don't know, security policy implications, then retrospective financial compensation isn't really going to do the job. And then there's the question of us even knowing whether anything's happened, right? Because the basis for really effective use of conventional law would be a robust reporting system that actually allowed us to establish what had happened. And right now there isn't really a really binding regime in place that requires the frontier labs to report every single instance of malfeasance, every single instance of breakout alarming behavior on the part of their agents. They do it essentially, you know, as a choice. And if you create legal liability, what does that do to their incentives to actually accurately report incidents when they happen? Because every time they do that, it establishes liability. And if you don't have then the cooperation, then you're going to need an enforcement mechanism and you're in the world of asymmetric information. How do you get in? So on the face of it, this looks like a way to go. As Meta has experienced, it can already bite. But when you keep pushing it down the line and you go to the risks associated with the frontier models, is it really going to work? I think is the question. And then at that point, the question of a new regulatory regime really becomes urgent.
Yeah, it does seem like several questions arise from this entire set of problems. I mean like among them is yeah how do they calculate the likelihood of an extinction level event. I mean I yeah I often come across this statistics that's cited as a 10% chance but who knows how exactly that is calculated. But then more broadly, I think there's an interesting question about how this apocalyptic potential that has always been part of this industry relates to the outsized value of the industry that everyone can see now at this point and can kind of trace in the stock markets. It strikes me that, you know, this alleged apocalyptic potential seems perversely to actually be fueling the value of these companies. I mean, it's even though there's the chance, a 10% chance of all value being eliminated entirely, that very potential seems to be an advertisement of the power behind this and thus the potential economic value. So I'm curious like how do those facts relate to one another?
I mean I think this is absolutely on point. I mean and I think we can take credit on the show over the last 18 months for having formulated this dilemma in various ways in our efforts to keep up with this story which obviously for non-specialists like ourselves is a kind of continuous struggle. But it has seemed to me for a while that we're on a kind of knife edge type problem here where, you know, scenario A is the technology delivers on its promises and dramatically transforms the world whether that's a catastrophic extinction level bioweapon type development or simply the promise to automate all white collar work including the making of podcasts which puts us all out of labor. But of course creates huge value for the labs and whoever else in the value chain. Maybe it's Nvidia that really gets the money or the data centers. In any case, someone in the AI value chain gets really rich and the condition of that however is spectacular not to say apocalyptic change in wider society or on the other hand it turns out that AI functionally is a bust. In fact, it's a much more modest level of innovation and transformation than the labs will have us believe, including the alarmists, for better and for worse, less dramatic. In which case, there needs to be a market reset because there's a massive overvaluation of the entire AI chain in the stock markets. And right now, it's driving 36% of all business investment in the United States. Spending is running at between 150 and 200 billion dollars a quarter in the hyperscalers. So were it to turn out that all of this in the end is not generating real value. It doesn't really need to do anything transformative. It's just not a business proposition. We're in deep water as well with potentially implications for financial stability increasingly as the big players get worked with private finance and in some cases even with big players on Wall Street with big banks. So we're in that kind of dilemma and this point that you're making speaks directly to this and you see it played out, right? So, as OpenAI and Anthropic bosses say we need to pace the frontier, SoftBank's shares sell off. They're a big stakeholder in OpenAI. Chip producer stock sells off and then the next day on Monday or Tuesday you see Jensen Huang of Nvidia which is at the core of this entire thing. I mean, it's the bank of AI and also the hardware provider, if you like, the kind of gold standard provider of the AI ecosystem because it's both a physical commodity and essentially a huge financial power leveraged on top of that. He comes out and says, oh, no, no, no. What are we talking about? Why are we even trying to slow this down? That doesn't make any sense. And, you know, on the one hand, you could say, well, he's maybe right. Maybe there's an exaggeration. Maybe this can just be handled by conventional product liability type rules. And on the other hand, of course, he's talking his own book because talk of regulating AI expansion is not good for Nvidia's business prospects. So whichever way you turn on this, we're really in a hall of mirrors.
