How AI Becomes a Political Crisis

**Jordan Schneider** (0:00)
The political economy of AI. Things are getting weird, it seems. We have models on the loose, open weights being dropped from the, I don't know, heavenly temple. We have data center protests fanning out across the land. Just a lot of salience, I think, in a way that feels really different, perhaps, than it did even six months ago. To discuss, we have the wonderful Anton Leicht, a fellow at Carnegie and author of the excellent Substack Threading the Needle in the Attic today to discuss. Welcome to ChinaTalk, Anton.

**Anton Leicht** (0:43)
Thanks for having me.

**Jordan Schneider** (0:45)
So am I right? Things are getting weird, we're coming close-ish?

**Anton Leicht** (0:49)
Well, yeah, I think there's like three or four or five different trends happening all at the same time, and it just seems like they're all sort of compounding and things are getting real. Like parts of US government are sort of waking up to the cyber implications, plus there is, I think, a lot of incentives leading into the midterms to just like find very visible symbols of all the things that I think people on the populist flanks on both sides care about, and turns out AI is maybe one of them. And then AI is also just getting adopted very quickly and is actually becoming a real and somewhat salient thing to people very quickly. And I think these all just like compound to a point where just like a lot more like newspaper pages and media that is just like full of AI news, and it just feels like it's all getting very big.
I think it's still unclear whether this all sort of comes together into like one big AI's big thing or whether these are actually three separate stories that will die three separate ways. But yeah, it's happening. Yeah.

**Jordan Schneider** (1:34)
So I guess by this is happening, I am defining this is happening as this becomes a salient national issue that people campaign on and vote on. I remember in the 2024, oh my God, Trump Biden debate where one of the things I was like looking for instead of senility was how much are they going to talk about AI? And like export controls came up once, but this was basically not a thing. It was something on the horizon that people were being aware of. So we can look at all the different streams that you mentioned, the labor impact, the stock market, the inflation, cyber risk, whatever. And it seems reasonable to assume that at least in the next three months, we are going to pick up steam, not dial it down, in which case, politicians are going to want to say they're going to do something about it. So then the question is like, what?

**Anton Leicht** (2:40)
Yeah. I mean, I think they don't really know just yet. And I think there's, I wouldn't be surprised if midterms were still fairly quiet on this, just because people are kind of still hiding. And a lot of policymakers on both sides just don't really know, A, just like which way the wind is blowing, and B, what the specific policy asks here should be. So I think they'll just let the midterms happen, let some of the discussion unfold. And then I think once they do see how salient the issue is, and like post-midterm polling, and once they realize, well, people really cared about this, and that race, and that race. Once they have a view of what the political spending looked like, and what the public opinion looked like, I think then they'll go into like 27, into the next Congress, just like realizing, well, we absolutely have to do something about this. We can't go into another election cycle having done nothing about this. And I think that's going to be sentiment of both sides. And I think then we'll probably be in a window where like actual policy that actually gets like sort of some, some amount of national attention actually happens. Yeah. Okay.

**Jordan Schneider** (3:32)
All right. Well, let's make this podcast proactive. Let's give people an agenda that they could, if not campaign around, at least perhaps just like, maybe make the world and futures a better place. We've done plenty of talk about export control policy. We're going to set that to a side for today and talk more about the domestic dynamic. Should we start with labor impacts? Where do you want to begin?

**Anton Leicht** (3:58)
Yeah, we can do labor impacts first.
That seems like realistically one of the more salient, salient policies of areas. And I think with the labor market specifically, there's two questions. Is there, are there things that actually need doing? And then are there things that just need doing for political reasons, because there is going to be some amount of salient, some amount of political need for some action on this, and then we'll have to find the least bad policy that still deals with that? Because I think there are things we should just substantively, definitely do, right? We should be better at knowing what is actually happening in the labor market as AI affects it, right? We have very, very fuzzy, very spotty data on this. It's not very good. It's not very comprehensive. A lot of the most important and most interesting data is inside the labs. They have very, very good statistic on usage and diffusion and the gap between how much is being used for more augmentative purposes versus more automation heavy workloads. This is all super interesting data that is extremely useful to make any sort of responsive policy. But we don't really have it. It's just labs that have it. So I think finding some way for governments to actually get access to this data and make decisions based on that seems obviously extremely useful.