Lawfare Daily: Adam Thierer on the Bipartisan House Task Force on AI’s Report

[Intro]

Adam Thierer: Are we going to have preemptive prophylactic types of regulations for AI that tries to figure out everything that could happen inside the black box beforehand? Or are we just going to say, look, we're going to see how these things play out and we're going to sort of roll with the punches and figure out after the fact utilizing a diverse toolkit? That very much informed this House AI task force report, that vision.

Kevin Frazier: It's the Lawfare Podcast. I'm Kevin Frazier, senior research fellow in the Constitutional Studies Program at the University of Texas at Austin and a Tarbell fellow at Lawfare joined by Adam Thierer, senior fellow for the Technology and Innovation Team at R Street.

Adam Thierer: At the end of the day, this paper, this report has opened the door to like open source being a key part of the American AI story. And I think that's a huge plus.

Kevin Frazier: Today we're talking about a lengthy detailed report issued by the Bipartisan House Task Force on AI.

[Main Podcast]

Let's start with the basics, Adam, who was behind this monster of a report? 273 pages, 66 key findings, 85 recommendations, who the heck authored this and what was their mandate originally?

Adam Thierer: Yeah, well, this report came together because the House AI Task Force was formed earlier this year to study a wide range of issues that were of concern to members in the House after the Senate had taken the lead.

Pretty unusual for the Senate to take the lead before the House, but Senator Chuck Schumer and a couple of other key senators had a bipartisan effort on the Senate side. And there were so called AI Insight Forums and other things. And so the House kind of got overtaken by events a little bit.

The Speaker's office and the Minority Leader's office got together and said, okay, let's do our own thing. And so they brought together a bipartisan team led by Representative Jay Obernolte and Representative Ted Lieu, both of California. And they got started in February.

And there were a variety of closed door sessions. Another unusual part about this is a series of testimonies including all the members of the task force, and I was lucky enough to be part of one of them and many colleagues that were parts of others. And it was a really interesting and more detailed discussion of AI policy that followed with lawmakers sitting for, in many cases, several hours to listen and have debates.

And it was a really interesting experience that yielded this beefy report, as you say. I mean, it's one of the biggest reports I've seen. I've been covering tech policy for 33 years now. This is one of the biggest technology policy reports I've ever seen Congress issue. So that gives you a feel for the gravity of this issue in Congress today.

Kevin Frazier: Well, we're very thankful to you for diving into the morass that is the report. And one point I want to additionally flesh out is why having Representative Lieu and Representative Obernolte lead this task force was so important.

What's kind of their distinguishing factors as folks who are authoritative voices on this policy topic that may counteract the traditional narrative that Congress doesn't know what it's doing. It doesn't have any expertise. What's unique about these two co-chairs?

Adam Thierer: Sure. Well, that's a great question.

First of all, and most obviously, they are from California. California is a pretty important state when it comes to AI and digital technology. That's the most obvious thing. But more importantly, these are both two skilled individuals in this department these are people that have put a lot of serious thought into technology policy writ large.

In Mr. Obernolte's case, as the chair of this, he is the only member of Congress with a degree in artificial intelligence. And he's also a successful video game entrepreneur. He's one of the most interesting characters I've ever met in Congress. You go to meet him in his office and he literally has a collection of his favorite old arcade games from the 80s. And he’s literally challenged me to competitions in front of some of these games that we're both old enough to know.

Kevin Frazier: Critically, did you win?

Adam Thierer: I didn't, we didn't play. We're like, we should talk about policy, right? We shouldn't pay Galaga all day long in your office, Congressman. And he's like, yeah, you're probably right.

Kevin Frazier: Ah, dang next time.

Adam Thierer: That's a unique character right there. And so here's someone with real world business experience as a successful entrepreneur and a degree in the topic. That's a winning alignment of skill sets. And so he did a wonderful job. Chairing this along with Ted Lieu. I have to say, I think they really did bring out the best in sort of the bipartisan policy making effort that we had hoped for.

It proved that once again to me, that we can have bipartisanship in tech policy in the United States. This was a key feature of technology policy in the 1990s that people have sort of forgotten about in our hyperpartisan age today. But back in the 90s, tech policy was very much a bipartisan thing. The Clinton and Gore administration worked very closely with the Republican Congress to bring about many of the policy successes that gave us the internet and the digital revolution.

And I'd like to think that what Representatives Obernolte and Lieu have done here is sort of rekindled that flame a little bit. And although there's a lot more to chew on here, and there's a lot of, you know, ambiguous language and aspirational statements in this report. It's still really good to see that a lot of lawmakers can get on the same page on some key issues.

Kevin Frazier: And flagging for our listeners, you've done extensive writing in this space and you've paid particular attention to statements from representative Obernolte on this topic. And you've highlighted that he, for example, has called for AI regulations to be based on quote, the values of freedom and entrepreneurship, end quote.

And I'm wondering if you can give us a sense, looking at the report, knowing Representative Obernolte's general framing, how do you see that sort of value statement appear in the report itself? And then we can get into the substance soon. But just at a high level, how is this the sort of document you would have expected from Representative Obernolte?

Adam Thierer: Yes, indeed. Well, Representative Obernolte has had a lot to say about AI policy leading up to this report and has written some really interesting things about it, including a journal article for the Ripon Forum that he talked about the role of Congress and regulating AI.

And that's where that quote appears that you mentioned, where he identified the importance of having that sort of a vision of freedom and entrepreneurialism. And then also making sure we didn't have government control of technology or what he called the anti democratization of knowledge. And I like that phrase.

That's very much in line, again, with the old bipartisan vision from the Clinton/Gore days in the 1990s. And so I love that. But more importantly Representative Obernolte has gotten into the weeds of policy. And he's talked about sort of like the key principles that should guide the policymaking process. And the most important of them I wrote about in a piece that I, I called, ‘The Most Important Principle for AI regulation.’

And I highlighted how Obernolte has repeatedly said that we have to make sure that we don't get too hyper focused on the actual process or models behind algorithmic technologies and AI systems, and we should focus more on outputs and outcomes, real world consequences and results of these systems. He said, because you can get too lost trying to get obsessed about, like, what's going on inside the black box, and sometimes you won't even be able to figure out preemptively exactly where the harm might lie.

And so he talks about this as sort of a defining feature of like good AI law, like making sure we focus on, you know, the actual outputs and not the input side of things. And I think that's crucial. I really do think is, again, as I said, in my piece, that's the most important principle. That's what we're really fighting over.

Are we going to have preemptive prophylactic types of regulations for AI that tries to figure out everything that could happen inside the black box beforehand, or are we just going to say, look, we're going to see how these things play out, and we're going to sort of roll with the punches and figure out after the fact, utilizing a diverse toolkit? That very much informed this House AI Task Force report, that vision.

Because thirdly, the thing that he has stressed again and again is we need flexibility. We need incremental flexibility, agility, incrementalism, it comes up again and again in his speeches, in his interviews, and that has played out in this report in a really big way.

Kevin Frazier: And before we get into the weeds, and I promise we're almost there, I just have an aversion for pulling weeds because it was a household chore when I was growing up. We'll get to the weeds soon.

One thing I'd love for you to flesh out in just slightly more detail, too, would be why, from your perspective, it is so important to follow that Gore/Clinton approach to digital innovation and innovation writ large in comparison to what we've seen, for example, be applied generally as a regulatory framework in the EU where innovation perhaps isn't as rampant, isn't as evident as a result of some of the regulatory choices there.

Adam Thierer: Yes, absolutely. Thanks for asking that question because this has been the focus of my life's work. In fact, I've written a couple of books about this issue and talked about this quote unquote conflict divisions, if you will, in the transatlantic sense between the U.S. and the EU on digital technology policy over the last 25 years.

When economists and political scientists like myself look at the real world and try to study it, we really love when we can find so called natural real world experiments where two jurisdictions that were maybe similarly situated a certain period of time decide to end up utilizing different policy paradigms to govern a new emerging technology or a new sector. And that happened in the 1990s.

The United States went one direction under the Clinton/Gore administration through the so called Framework for Global Electronic Commerce and a variety of resulting laws and policies. And the European Union went in a very different direction with very heavy handed top down data directives and a wide variety of rulemakings that continue to this day.

And so you've got rules like the GDPR, the DMA, the DSA, and now the EU AI Act and a whole bunch of policies and regulations in between. The United States took a far more incremental bottom up approach. And I would argue that these two paradigmatic approaches to tech policy had some important real world ramifications that still have continuing lessons in the AI world.

Today in, you know, in the world 19 of the 25 largest digital technology companies by market capitalization are U.S. headquartered companies. Only one is in Europe. 48 of the 100 largest digital employers in the world are U.S. headquartered companies. There's not a huge number in Europe. You can take a look at investment flows and talent flows and so many other metrics on this.

I could go on all day about this and I've got all this research. The situation is not good for Europe right now. They've really hollowed out their digital technology sector. And whenever I'm in front of students or audiences, I often challenge them with a simple test, like name me a single leading digital technology innovator headquartered in the European Union today.

And the best they usually got for me is Spotify. I said, well, I'll give you that one. But you know, and there are a few others. I admit there are a few others, but there's not many. And the fact that it's so hard for us to name those leaders and the fact there's such a lack of them in the digital technology space, it has to tell us something about policy.

And I believe it tells us that, again, the Clinton/Gore vision worked, that this sort of more bottom up incremental, you know, take it as it comes kind of approach made a great deal of sense as opposed to the more top down precautionary principle based approach of the EU that attempted to solve every problem before it even happened.

And I fear that's the direction, you know, we're having the debate right now in AI policy in the United States at the state level, as we'll get to in a moment. But the fact of the matter is that the United States steered a different direction and it had positive consequences for innovation and for speech.

I want to be clear. It wasn't just about commerce. It's about speech too. We have more speech platforms and speech opportunities than any nation in the world because we got policy right.

Kevin Frazier: And as you and I have talked about, we've seen that the incoming Trump administration seems to be leaning into this approach of putting innovation at the forefront of AI policy.

There's a long discourse we could have about the extent to which the Biden administration appeared to be more focused on some of these ethical concerns or concerns about discrimination and an urge to want to get into some of those internal processes you were talking about earlier.

So now turning to the actual report and having that sort of background framework in mind and emphasis on innovation, we've got 66 key findings, 85 recommendations. We're not gonna march through all of those as much as I, I think listeners would love to know the ins and outs of this full report. I'd love for your kind of expert take on which of those findings, which of those recommendations really stand out to you as warranting additional attention.

Adam Thierer: Yeah, I'd be happy to, but let me start, Kevin, by taking just a little bit of a step back because I want to remind people where we stood in the AI policy debate in this Congress. Just even 18 months ago, at the beginning of the 118th Congress, we started a pretty vigorous debate about AI policy following the launch of ChatGPT and interest was intense.

And I would argue there was a lot of hysterics. There were people who really kind of lost their minds at first about this stuff. And I'll just take you back to May 16th of 2023. Senate Judiciary Committee held a hearing as Sam Altman at it, Gary Marcus, and a representative from IBM. It's the Senate Judiciary Committee.

In this committee meeting, which happened at the same time that the Pause AI letter was gaining many signatories, you may recall saying we should just stop AI period, full stop, whatever that means, it was never fully explained, but there were a lot of people signing onto it, right? Including Elon Musk.

Kevin Frazier: The basis was this concern for listeners who perhaps had hazy memories from May 2023, the overriding concern of the folks signing onto that letter was existential risks or catastrophic risks posed by AI, which have broadly informed this sort of AI safety movement which we've talked about previously on the pod. But just for folks who hadn't read the letter recently, but Adam, please continue.

Adam Thierer: That's exactly right. So it was very much dominated by that concern, sort of like Terminator-esque concerns.

And so, unsurprisingly at this May 2023 Senate Judiciary hearing, we heard references to AI as a new atom bomb, and one senator famously said that when we think about AI, we should just start from the premise that it wants to kill us. And then there were proposals put on the table for comprehensive AI licensing, just broad licensing; didn't even specify like by sector or anything else.

It just said broad based general purpose AI licensing by a new regulatory body that would be resembling a so called FDA for algorithms, whatever that is. It got worse from there, where we went into a discussion of like how we might bring it all under global control from some new global regulatory body. Or at least America would, you know, somehow become compliant with everything the Europeans wanted.

I mean, it went on from there. And it was just an astonishing thing to witness. And I wrote about it that day, and I said, you know, it's hard for me to believe this conversation can get much worse from here. The only place we had to go that was left was full blown nationalization of all algorithmic systems and high powered computing. And believe it or not, there were some people that were putting on that table, that idea on the table in academic circles. But let's flash forward now, 19 months.

Kevin Frazier: Well, and just to pause there before the 19 months. For folks to hear your take on the AI safety debate, the folks who were urging senators and encouraging senators to consider those risks were saying, yes, perhaps we agree that the worst case scenarios we're hypothesizing may have very low odds of occurrence. But because those risks could end humanity, that's why they weren't particular attention. What's your pushback there in terms of just that epistemic approach to understanding policy?

Adam Thierer: Sure. Yeah. Well, we got way out ahead of ourselves with hypothetical Chicken Little scenarios about like how, you know, the AI was going to result in the Terminator and come and kill us all.

And it resulted in California pushing for a major comprehensive law, SB 1047, that resulted in one of the most historic technology policy debates in my lifetime. In fact, I don't think I've seen in, except I'd have to go back to the crypto wars of the nineties, I don't think I've ever seen such an interesting, strange bedfellows collection of people for and against SB1047.

It was really, really intense. And California passed that law. Which would have had a pretty interesting restrictive control regime for so called frontier AI models as measured by a certain computational power, and then even a new regulatory bureaucracy to oversee it. But then Governor Newsom vetoed that bill in California.

And so the debate that was started at the federal level kind of shifted to the state level, but really played out in Sacramento. And, you know, again, many of us, including me, pushed back aggressively and said, like this, we are getting way ahead of ourselves. These are hypothetical concerns. They have no basis in reality.

People have been watching, you know, too much sci-fi and Black Mirror episodes, you know, in the evenings. And we need to step back and have a more cautious approach to this, especially because we pointed out how devastating that would be to certain types of algorithmic innovation, not just for smaller players and others, but more specifically from open source players.

And so that was an incredibly intense debate, but one that's now been kind of paused a little bit after Newsom's veto. And we could get into this, but this has shifted the dynamic over the so called X risk, existential risk debate back to the federal level. And the so called AI Safety Institutes that are being formed across the globe, including here in the U.S., and efforts to have different approaches to it at the federal level besides flat bans or flat restrictions.

But it could play out in many, many ways, including some ways that aren't really specified in this new House AI Task Force, which doesn't say a lot about some of the export control debates and some of the other things about cap, capping computational power or something. So there's a lot left ambiguous about like where this task force comes down on AI safety. And so that's one side of the debate.

And then there's different types of debates like AI discrimination or ethics and bias. That's another bucket. And then there's sort of sectoral battles or targeted, you know, concerns that we could talk about as well, but those are the major sort of like political, you know, demarcations or buckets that I use to describe the way AI policy is playing out in the United States right now.

Kevin Frazier: Looking at the report itself, you mentioned that even at its incredible length, it doesn't touch on everything. Of its recommendations and findings, what were your big takeaways from its focus? What should we glean? What might it be suggesting about Congress's focus come 2025?

Adam Thierer: Yeah, absolutely. So first of all, let's talk about tone and general recommendations. It gets the tone is balanced. The tone is sober. It is not the tone that we heard from that. May 2023 Judiciary Committee hearing that I just discussed. It's very different and that is important.

Words matter. Entrepreneurs in the market get signals from the way that policy makers talk and it also spooks investors as well. And so we have a very much more responsible approach here. I'd argue the adults have entered the conversation. And they've come up with a more balanced approach.

They've also come up with an approach that repeatedly stresses flexibility. The term flexible or flexibility appears over 20 times in this report. And terms like agile, incremental, things like that, they appear multiple times as well. And so that is what I'll call a return to normalcy. Like this is the way American technology policy for digital issues has worked generally for the last 25 years. Incremental, agile, flexible, bottom up, a lot of stress on things like best practices, multi stakeholderism, you know, collaborative efforts.

But it's really this sort of more rolling with the punches kind of approach to policy that I will admit freely is very, very messy because all decentralization is messy. And that frustrates people. People love more broad, you know, silver bullet solutions up front, like, how do we solve this problem?

And this report very maturely admits and even uses the term humility when talking about it that we don't have all the answers. And Obernolte said, leading up to the report, I'll just quote it, that this is not going to be one 3,000 page AI bill like the European Union passed last year, and then we're all done with it. He said, instead, he says, you know, America will need to have a more flexible and responsive approach.

And that is what this house AI task force is recommended. A more incremental flexible approach. So that at the highest level is in and of itself an important policy approach. It's not a specific recommendation, but it informs all the recommendations you see from there on out, which very much have a sort of on one hand, on the other kind of approach.

It doesn't have this bright line like, and we therefore decree that this is how we will solve all issues involving AI and copyright, or AI and national security, or whatever. It says on the one hand, there's this issue. On the other hand, there's this. But generally speaking, comes out more in the direction of freedom, flexibility, you know, pro-innovation kind of approaches. So that's important.

Kevin Frazier: Yeah, and what's really telling about this report in comparison to the Senate's Bipartisan AI Working Group and their policy roadmap. That policy roadmap was about eight pages of what could have been a Harvard Kennedy School memo, and I can say that I've been to policy school where there just wasn't a ton of substance, right?

We weren't even able to really detect what the general tone would be, what the signal would be from the Senate working group. Here though now, I think with this bipartisan group of folks who are well versed in AI, who have a stake in AI's continued exploration and research and development, this really does seem to me to send a signal. And so looking at those specific recommendations, are there any that stand out to you that you think are going to be particularly top of mind for Congress going forward?

Adam Thierer: Yeah, well, first of all, at the highest level, it stresses the need to treat different types of AI and different types of sectors differently, and that we should take more of a sectoral risk-based approach. And a lot of people use those phrases, but this report actually puts some meat on the bones of it.

It's easy to say risk based approach, but it actually says right up front in the preface that the key principle would be identifying AI issue novelty. And really specifying, like, is this a truly new AI capability that we've never been able to ever seen before and never could never handle using existing law? Or is this really just an existing issue that the nature of it has changed because AI entered the picture? And you know, you look at those two buckets, you say, well, what were they talking about?

Well, on the first thing, like something that's AI’s totally different, really quite new. I mean, you know, nonconsensual deep fakes utilizing AI is something that has really become something policymakers are concerned about. And I totally get it. There's some new capabilities there that transcend old laws and capabilities that probably are going to require some novel approaches. And we're starting to see them. And that makes a certain amount of sense.

On the other hand, what's an existing issue that has been changed significantly by AI? Well, how about AI ML in medical devices? Or AI in like autonomous systems like driverless cars or drones. I mean, you know, a driverless car is in one sense a computer on wheels and that's new, but at the same time, it's still something on wheels. It's a car. It's something we've known about and regulated for a hundred years.

And the FDA has been off and running doing computerized medicine or digital health for now the better part of 15 to 20 years. It's not all that new. And this report finally gets serious about breaking things down to smaller components.

And this has been my primary recommendation and all my testimonies to Congress, all the papers I've written. The only way we're going to get anywhere on AI policy and governance as a nation is if we break it down into smaller components. And we take a building block approach of like, okay, how can we utilize these existing standards, laws, regulations, court-based systems, and then how do we supplement them as needed.

This report really finally starts to think about that in a serious way. Doesn't answer all the questions, admits that it doesn't know how in some cases, and in other cases just dodges a really sticky wicket and just says we're going to pass on that one. It's very clear where they did that in some chapters that we'll get to in a second, but I think at a high level that's a great framing.

I think this is what we've always needed. You know, we've eschewed the idea of like, you know, oh, here's the simple solution. Here's the silver bullet. That just doesn't exist. And, you know, the people that were coming up with really far reaching and I would argue radical solutions out of the gates, again, going back to that old hearing, you know, sweeping broad based, you know, let's license AI.

You know, what does that mean? What does that mean? They never unpacked it for us. This report is willing to sweat the details. These people behind it and the staff who put enormous hours into it. I mean, they actually spent some time thinking through those, you know, devilish details, and that's good. We've made progress here in that sense.

Kevin Frazier: I think a lot of listeners and especially the folks in the AI world who may be checking out this report are going to do a control F quickly for open source. So what are some of the takeaways for you from the report with respect to open source? We've seen that this is one of the hottest topics in AI debate.

In part because the folks who are on the AI safety community side would say things like, hey, perhaps open source models as a result of facilitating the spread of models that we don't know their full capacity, what they may be capable of. That's a really grave concern, especially from a national security perspective.

On the other side, we may see folks who are leaning into innovation and democratization of this technology and saying, open source is the way to go. And that's the way we make sure these benefits are realized across the U.S. So this remains one of those sticky wicket issues you pointed out. Does the report give us some clarity on how representatives Lieu and Obernolte are thinking about this?

Adam Thierer: Well, it does say some important things and you're absolutely right, Kevin. This has become a really interesting part of the broader AI policy wars. And it, it again has created interesting sort of strange bedfellow coalitions.

And in a recent piece that I wrote about the Trump administration and their AI approach. I pointed out that one of the tensions in the coming Trump administration may be between those conservatives who take a very old fashioned approach to open source being a potential national security vulnerability or threat versus the newer MAGA conservatives who think of it as a great way to inject more competition and choice into a world dominated by what they regard as the evils of big tech.

And that is a really interesting development. Because in the past, when I spent 10 years working at Heritage in the 1990s and ran their first technology, digital technology program. Back then I supported open source, but I didn't have a lot of friends who did in the institution. But now Heritage and other conservators are vociferously behind, like, supporting open source.

And so where does this report come out? Well, the report says, very interestingly enough, that Open AI models encourage innovation and competition, and there is currently limited evidence that open models should be restricted. It goes into a little bit more detail about that and has some generic recommendations about we should continue to study its vulnerabilities and specifically how it could create, you know, dangerous capabilities, whether they be chemical, biological, radiological, nuclear, and so on and so forth.

But at the end of the day, this paper this report has opened the door to like open source being a key part of the American AI story. And I think that's a huge plus. The Biden administration, in my opinion, is pretty good on this as well. The Biden NTIA within the Department of Commerce issued a major report on this issue that I thought was excellent and really seriously evaluated, you know, the so called marginal risk question of what's going on with open model weight systems and said, like, look, you know, it depends is the answer.

And it's really hard, but we have to be careful about casting too wide of a net for open systems or else you'll crush all of these beneficial forms of innovation and competition. So this tension is going to continue to play out. But the important point is that this report did not only did it not foreclose open systems, it kept the door wide open and kind of encouraged them. And I'm extraordinarily pleased with that result.

Kevin Frazier: So we're talking just 24 hours after the release of the report. So, number one, thank you for completing a very difficult assignment. I'm a mean professor for assigning 273 pages of reading in one night.

Number two, I think that leaves us with a bit of speculation we have to engage in, in terms of identifying, what are the aspects of the report that are going to be kind of top of mind for the public and for policymakers going forward? You've mentioned there are a couple key chapters that you think stand out. What are some of those that you think warrant specific diving into?

Adam Thierer: Absolutely. So there's a couple of areas here where I think there's a lot of consensus, and I think they're going to be priority issues in 2025, both in Congress and in the Trump administration. The first on that list, in my opinion, would be energy uses in data centers, which is a very interesting chapter in the report.

And they do something in this report that's really, really crucial in my opinion. They make it very clear that there is a symbiotic relationship between the success of AI systems and the ongoing success of our ability to diversify our energy portfolio and grow our energy, you know, sources as a nation.

Then secondly, they link that to geopolitical strength and security. And the report says, direct quote, AI is critical to U.S. economic interests and national security and maintaining a sufficiently robust power grid is a necessity. And then says again in another bullet point, continued U.S. innovation in AI requires innovations in the energy sector.

So these things are now going to go hand in hand. We're going to be having conversations about energy and AI policy at the same time. And in my recent report that I mentioned about, you know, AI and the Trump administration, I think this is a top line winner for the Trump team. They've been talking about energy diversification and the importance of making sure that basically we don't become Europe and become subservient on other nations for our power. And, you know, we've got to make sure we can continue to power all our important, you know, new sectors and innovations, especially AI and data centers.

So this is going to move forward. Now, there's going to be a little bit of attention, I think, on the Democratic side about, you know, what kind of power generation are we talking about? Is it too much power consumption? What are the, what's the climate footprint look like? So on and so forth. But the report basically also says, look. I mean, AI can be part of that solution as well. It can help us diversify and, you know, clean up our system and find alternative sources.

So I think that's going to be a compelling narrative in 2025. It's essentially the linkage of two technological revolutions. The revolution in alternative and new energy sources, and hopefully a renaissance rebirth for nuclear power in particular, but then also the continuing computational. So that's one area I think, a lot of agreement.

Kevin Frazier: I'm not a betting man, although I'm going to ask you to make some predictions at the end of this pod. So hang on to your hat for that one.

But thinking about issues where we can see the stars align or strange bedfellows develop, this does seem like the top of mind for me, when you think through democrats in rural states, for example, this is a huge win if you can bring home an energy project while also talking about innovation, national security, defense, right.

Conservatives likewise can pay maybe less attention, hey, this is known as a sustainable resource or what have you, solar, wind, hydro, but instead emphasizing that innovation. So I do think this is maybe one of those areas where we can see some consensus emerge. So I'm keen to keep the optimism going. What's another area we're seeing?

Adam Thierer: Yeah, you're absolutely right about that though. And another thing I should have mentioned is that also plays into the discussion in this report about technology R&D.

And there's a little bit of an industrial policy component to this that I won't get into too much, but the bottom line is that, you know, that's a love fest. Everybody loves spending money in their districts and like, you know, spending more money on new projects, whether it be energy projects, data centers, whatever else. So I think that's going to be something we're going to see action on.

The other thing I'll mention here, just very briefly, there's a couple of key sectors that you hear lawmakers when they're having AI hearings and other events and doing public speeches about these issues. They always come back to a couple of key sectors.

One of them is healthcare and the other is agriculture. And that makes sense because first of all, in agriculture, everybody, almost every lawmaker has a soft spot in their heart for the old agrarian lifestyle and farmers and everything else. And they're very, very excited for what AI and robotics and computational technologies might mean for, like, improving, you know, the farming system, the agricultural system.

I won't go into the details about that, but there's a pretty substantial chapter in this report about agriculture and about how, quote, AI driven precision agriculture could enhance farm productivity and natural resource management.

Kevin Frazier: I just Jefferson is rolling over his grave thinking about where did all of my human farmers go? They got replaced by AI. This is a problem, but I'll leave discussion of the Federalist Papers and anti-federalism.

Adam Thierer: Just as an aside, I come from a family of farmers in Midwestern Illinois and, you know, when I go back to see the family farm these days, I see uncles and friends out on tractors that are completely robotic. You know, computerized systems with their GPS linked and that are sort of, sort of plowing the fields for them, as they sort of sit back listening to satellite radio and an air conditioned cab of their tractors drinking a beer. It's like the world has already changed. This revolution is upon us.

Kevin Frazier: This is going to change country music tremendously. All references to horses are gone.

Adam Thierer: Right, right. So that's going to be an area that the question really only there is like, how does that manifest itself? What is Congress going to do? You know, I'm not sure what it means in terms of policy, except for like, okay. encourage more, you know, AI and robotics on farms. Maybe they spend some more money on it, I'm not sure.

But healthcare is the really big one. And on healthcare, you hear a lot of lawmakers talk about what, here's the first bullet point from the report on it. Quote, AI's use in healthcare can potentially reduce administrative burdens and speed up drug development and critical diagnoses. This is a crucial thing. And you know, they all, you know, pretty much what every congressman and woman holds, they really all want to serve forever, a little longer life.

And like, if AI can make them serve that longer life and serve long they love it. They love it. But they also love it because quite practically speaking, it's an important way to potentially not just reduce administrative burdens, but administrative costs associated with a healthcare system that is, you know, really out of control in terms of those costs.

And there's already been a lot of literature on this and a lot of it cited in the report. And I think Congress is very eager and trying to figure out how to use more pilot programs or like injecting these sorts of experiments within the Medicare/Medicaid system as well as, you know, insurance more broadly. So I think that's a common theme that everybody agrees with. I mean, there's a lot of love there.

It's a little bit more controversial when you get to some other sectors. There is a major section on FinTech in this report. There's some recommendations in it, but it's a little bit more generic. And I think at the margin there, they were, you know, trying to be careful about what they said because I think that does divide some members more than ag and healthcare does in terms of AI's role.

Kevin Frazier: Before we, we have to unfortunately draw this fascinating conversation to a close, I'm keen to hear more about one of the blind spots you identified in the report. So you said in your write up, you know, this is a great example of returning to an emphasis on agile, flexible, incremental regulation.

But you flagged one big issue, and that was, in your opinion, insufficient attention to the fact that we have this patchwork of state laws developing that could really hinder AI innovation, in your opinion. If I were to phone a friend right now, I'd call David Rubenstein, and he'd say, Adam just doesn't get it, we need AI federalism.

This idea he's coined of making sure that we should lean into states and celebrate states as a sort of laboratory of democracy approach. They can test ideas, we can identify best approaches, and then Congress should take action. You seem to be of a different mindset, and to the extent I represented David accurately, how would you counter that idea that now's the time for AI federalism?

Adam Thierer: Well, just in response to the ghost of David that haunts the skull, I'll just say this, that the first of my dreadfully boring ten books was on federalism and interstate commerce in 1998. And I really openly struggled with the questions about how it's applicable to various types of emerging technologies.

Because it's legitimately hard to know in some cases what we even mean by interstate commerce and the world of digital bits and algorithms, right? You know, generally speaking, these things don't stop at state borders, and they really shouldn't, and I want to see speech and commerce flow as freely as possible.

I mean, that's what makes America really great. And really, in the internet revolution, we proved it. We had a national framework. In fact, the Clinton Gore document was called the Framework for Global Electronic Commerce. And there was an acknowledgement at that time in the Telecommunications Act of 96, in Section 230, in the Internet Tax Freedom Act, and a whole variety of other things, that we needed a national vision for this new global medium and resource. And we got it.

But what's changed in the subsequent 25 plus years now is that the states have very aggressively asserted themselves. And not only just asserted themselves, but actually got out in front on digital technology policymaking in the United States. And they have led with two arguments.

One is that, well, Congress has just become a dysfunctional mess and they're not doing their jobs anymore, so we'll do it. What everyone wants to think about that's a leading argument by many of the state officials pushing for state AI bills. The second argument that they make is basically like, you know, well, we already have all these existing, you know, capabilities and laws and, you know, why can't we apply them to AI? And that's a more legitimate argument in my view.

But the problem is that the way they're applying those laws and what they're saying is that we should apply them more preemptively and prophylactically. That we should basically take and adopt a mini EU AI act model state by state and have a bunch of sort of preemptive rules that say you have to run your algorithm through some sort of a screening process or, you know, the so called algorithmic impact assessments and new bureaucracy will be set up.

And this is what I've argued in my work, sort of guilty until proven innocent model of digital policy. It's sort of like prior restraint for digital bits and algorithms until you get the blessing of some authority to move forward. That's the European model. I don't like it. And I think it's really problematic in particular for small businesses, which in the report is a major focus that we didn't discuss.

But basically, the report goes out of its way again in a very bipartisan way to say we need more quote unquote little tech competition to fend off big tech and like too much concentration. Well, you sure as hell ain't going to get it if you've got a patchwork of 700 plus laws, as we have today, moving in the states that are contradictory and highly costly with compliance costs that only the largest technology companies and their legal shops can afford to deal with.

And so we have to talk seriously about the two sides of federalism. And yes, I get it, states rights is important and they should have some flexibility. But not unlimited flexibility. Congress needs to establish, you know, assert its role and establish the fact that this is a national marketplace. And that we do have legitimate interstate commerce and speech concerns here that transcends state borders.

They at least need to play that role and put a little bit of fear of God in the hearts of the states to say, don't overstep your bounds as you do these things. But I'd like to see them go further and actually start to do some serious preemption, which is what I testified in front of this House AI Task Force on.

But I'll tell you this, as I wrote last night about the report, I got a very icy reception. There's just not much of an appetite for preempting in Congress these days. In fact, of the over 100 AI bills that were pending in the 118th Congress, and I read pretty much all of them, I think, I didn't see preemption mentioned in one of them.

And when I talked that day to lawmakers directly about, like, why we needed this, and we needed it to be more like ceiling preemption as opposed to floor preemption, I didn't get many takers. I got Representative Obernolte. He's made this a priority from day one. He wants some sort of preemption. And even Representative Lieu, the Democratic, you know, co-chair, he kind of gets this. But not many other people were interested and there's no laws proposing.

And so we're about to witness 2025 is going to be the year that I predicted that we're going to see the mother of all technocratic regulatory patchworks in the United States. And I don't see Congress doing much to stop it, and the report has very little to say. And it's mushy, mealy mouthed stuff about like, well, we should just look into this more. Literally, it recommends a study.

The lazy out in Congress, when you know they've reached a point of, you know, an impasse where they just can't go in for this, like, let's do a study on this issue. That's chickening out is another way to put it.

Kevin Frazier: You can always study more things. Well, before we sadly have to let you go, you've made one prediction.

Now I have to ask for one more. Speaking of icy receptions, your Indiana Hoosiers are going to be traveling to South Bend, Indiana for the first round of the college football playoff. For those who don't follow Adam and I on X, we've exchanged a few tweets, or whatever we call them now, about my Oregon Ducks, the greatest team known to man, and Adam's Hoosiers.

And so Adam, how are you feeling about the Hoosiers taking on the Fighting Irish? Do they have any chance?

Adam Thierer: I'm feeling great about it, except that Notre Dame's got God on their side, and I'm not sure the Indiana Hoosiers do, but I'll, I'm still rooting for them hard. I think they've got a great chance, but after that, Georgia awaits.

I'm pretty scared about that day, so we'll see. So if Georgia and Oregon get matched up at the end of this, I'll bet you on that. I'll take Georgia over your Ducks.

Kevin Frazier: Oh, geez. Okay. You've heard it here, folks. With that, we'll have to let you go. Thanks again for coming on, Adam.

Adam Thierer: Thanks so much for having me. I enjoyed it.

Kevin Frazier: The Lawfare Podcast is produced in cooperation with the Brookings Institution. You can get ad-free versions of this and other Lawfare podcasts by becoming a Lawfare material supporter through our website, lawfaremedia.org/support. You'll also get access to special events and other content available only to our supporters.

Please rate and review us wherever you get your podcasts. Look out for our other podcasts including Rational Security, Chatter, Allies, and the Aftermath, our latest Lawfare Presents podcast series on the government's response to January 6th. Check out our written work at lawfaremedia.org. The podcast is edited by Jen Patja. Our theme song is from Alibi Music. As always, thank you for listening.