AI 的 Little Tech 议程

There have been these big institutional players in D.C. and in the state capitals for a very long time. There wasn't anyone who was actually advocating on behalf of the startups and entrepreneurs, the smaller builders in the space.

They're trying to build models that might compete with Microsoft, OpenAI, Meta, or Google. What are the regulatory frameworks that would actually work for those companies, as opposed to making that competition even more difficult than it already is?

“Regulate use, do not regulate development” somehow gets interpreted as “do not regulate.”

I actually can't think of a single example across the portfolio in which we are arguing for zero regulation.

Collin and Matt, welcome to the podcast.

Thanks so much. Thanks for having us.

There's a lot we want to get into around AI policy, but first I want us to take a step back and reflect a little bit. We publicly announced the Little Tech Agenda in July of last year, and there's been a lot that's happened since. Why don't we first take a step back, Collin, and talk about what the Little Tech Agenda is and how it came to be at the firm?

Yeah, I mean, look, I give a ton of credit to Mark and Ben for having the vision on this. I think when I first started here, we began advocating on behalf of technology interests and technology policy. And I think what we realized was that there had been these big institutional players in D.C. and in the state capitals for a very long time. Some of them have done a lot of really good work on behalf of the entire tech community, but there wasn't anyone specific who was actually advocating on behalf of what I think we call Little Tech—which, in my mind, is the startups and entrepreneurs, the smaller builders in the space.

Well, they're not always 100% aligned with what's going on with the big tech folks. That's not necessarily always a bad thing or a good thing, but I think that was the whole impetus of this. How are we going to think about positioning ourselves in D.C. and in the state capitals in terms of our advocacy on these issues? How do we differentiate from the big tech folks, who come with certain degrees of baggage?

Yeah. Small from the left and the right, right? And the smallest of the small. That was really the basic impetus of this.

For me, it was actually almost a recruiting vehicle. When it came out in July, I wasn't yet at the firm; I started in November. When I first read the agenda, it transformed the way that I looked at the rooms I would sit in where there would be policy conversations. All of a sudden, you could see essentially an empty seat, and Little Tech wasn't there.

There would be conversations where people would say, “In this proposal, we want to add this disclosure requirement,” and then we'd have companies do a little bit more and a little bit more. When you've read the Little Tech agenda, all of a sudden you start thinking, “How is this going to work for all the people who aren't in the room?”

For me, the question, thinking about coming into this role at the firm, was: Is this a voice? Is this a part of the community I want to advocate for and think about? When you start looking at the policy debate from a Little Tech perspective and see how many of the conversations don't include a Little Tech perspective, it was very compelling to think about how I could advocate for this part of the internet ecosystem.

Right. Why don't you outline some of the pillars of the agenda, or some of the things that we focus the most on, and maybe how it differentiates from Big Tech more broadly?

Yeah, just from a firm perspective, obviously we're verticalized. We all live and breathe this, and I think that's been very, very competitive for us on the business side. But I also think it's very competitive for us on the policy side, too.

Obviously, Matt leads our AI vertical and is our AI policy lead. We have a huge crypto effort. We have a major effort around American Dynamism, which is defense procurement reform—something that the United States has needed forever and ever. We have other colleagues who work on the bio and health team, and they're fighting on behalf of everything from FDA reform to PBMs. There's a whole vertical there that they're working on.

We're working a lot on fintech-related issues and then, just like classic tech-related internet entrepreneurs coming up. What does that relate to? There are a lot of tax issues that come along with it. And, of course, there are the venture-specific things that we have to deal with.

But I try to think about this from a basic point of view: If you're a small builder, what are the things that should differentiate you from someone who's a trillion-dollar company with hundreds of thousands of employees? If you're 5 people and you're in a garage—

Yeah.

How are you supposed to be able to comply with the same things that are built for 1,000-person compliance teams? It's just not the same thing.

Right. And there are categories and categories and categories that Matt and I are dealing with on a regular basis.

It's made my job really hard in certain ways since I started at the firm, because the kinds of partners that you want within our portfolio often don't exist. A lot of the companies don't have a general counsel, a head of policy, or a head of communications.

The kinds of people who typically sit at companies thinking all day about what a state is doing in AI policy or what a federal agency is doing in terms of rulemaking aren't at startups that are just a couple of people and engineers trying really hard to build products.

Those companies face this incredibly daunting challenge. It seems so daunting for someone like me, who's nontechnical and has never worked at a startup. If they're trying to build models that might compete with Microsoft, OpenAI, Meta, or Google, that is unbelievably challenging.

In AI, you have to have data and compute. There's been a lot written about the cost of AI talent recently. It's incredibly daunting. The question that Collin and I talk about all the time is: For those companies, what are the regulatory frameworks that would actually work for them, as opposed to making that competition even more difficult than it already is?

Yeah.

One of the principles I've heard you guys hammer home is that we want a market that's competitive, where startups can compete. We don't want a monopoly. We don't even want oligopolies—a cartel-like system.

That doesn't mean no regulation, because, as we've seen, that could be destabilizing, too. But it means smart regulation that enables that competition in the first place.

Yeah. One of the things that has surprised me about venture is the time horizon that we operate in. Our firm and our funds operate on 10-year cycles. We're not looking to spike an AI market tomorrow and have a good year, a good 6 months, or a good 2 years. We're looking to create vibrant, healthy ecosystems that result in long-run benefits for people, and long-run financial benefits for our investors and for us.

That means having a regulatory environment that facilitates healthy, good, safe products. It doesn't mean that if people have scammy, problematic experiences with AI products, or if they think AI is bad for democracy or is corroding their communities, that's somehow good for us. That's not in our financial interest.

That really animates the core component of the agenda, which is not trying to strip away all regulation, but instead focusing on regulation that will actually protect people. We think there are ways to do that without making it harder for startups to compete.

Yeah. To Matt's good point, I walk into a lot of lawmaker offices and genuinely say—it sounds like I'm pitching my book, but I genuinely mean it—our interests are aligned with the interests of the United States of America.

The people we're funding are on the cutting edge. They're the people who are going to build the companies that drive jobs, national security capabilities, and the economy. We want to see them build over a long time horizon, and that is exactly how we should be building policy in the United States.

Of course, half the offices I walk into are like, “All right, great. Get that guy out of here.”

99.9% of the people we talk to think that all we want is no regulation. And yet, despite both of us writing and speaking extensively about the importance of good governance for creating the kinds of markets that we want to create, that perception persists.

Collin can speak more to it in crypto. I've learned a lot from our crypto practice, because the idea there is that you really need to separate good actors from bad actors and ensure that you take account of the differences.

And it's true in AI as well. If we don't have safe AI tools—if there is absolutely no governance—that's not going to create a long-run, healthy ecosystem that's going to be good for us and for people throughout the country.

I actually can't think of a single example across the portfolio in which we are arguing for zero regulation.

The core component of our AI policy framework, which was developed before my time—I wish I could take credit, and I can't—is focused on regulating harmful use, not on regulating development. And that sentence, “Regulate use, do not regulate development,” is somehow interpreted as “Do not regulate.” People just omit, for some reason, the part about focusing on regulating harmful use.

And that, in our view, is robust and expansive and leaves lots of room for policymakers to take steps that we think are actually really effective in protecting people. So regulating use means regulating when people use AI to violate consumer protection law, when they use AI in a way that violates civil rights law at the state and federal level, or when they violate state or federal criminal law. There's an enormous amount of action there for lawmakers to seize on, and we really want that to be an active component of the governance agenda that we're proposing. For some reason, it's all passed over, and the focus is just on “Don't regulate development.” I don't exactly understand why that ends up being the case.

Easy headline. There's been a lot that's happened in AI policy, and I want to get to it, but first, perhaps, Matt, you can trace the evolution a bit over the last few years. I believe there was a time when we were pattern-matching with social media regulation a bit. Why don't you trace some of the biggest inflection points, or the debates over the last few years, and we'll get to today?

I think we have to play a little bit of history. I want to get to a point that I think is the really critical point of what we're all facing here. For me, from a policy and government affairs perspective, this conversation started in early 2023. That was sort of the starting gun. It puttered along and became more and more real over time.

But in the fall of 2023—so, almost exactly to the day 2 years ago—there was a series of Senate hearings in which some major CEOs from the AI space came and testified. I think the message that folks heard was, one, we need and want to be regulated, which I think remains true today. That's obviously what Collin and I are working on on a regular basis.

But I think included in some of that testimony was a lot of speculation about the industry that led to, and absolutely jump-started, this whole huge wave of conversation around the rise of the Terminator. You know: “Go hug your families, because we're all going to be dead in 5 years.” And that spooked Capitol Hill. They absolutely freaked out about it.

And look, rightfully so. You have these really important, powerful people who are building this really important, powerful thing, and they're coming in and telling you that everyone's going to die in 5 years, right? That's a scary thing for people to hear. And, oh, by the way, they want to be regulated. That starting gun, I think, moved us at hyperspeed into this conversation around: How do we lock this down? How do we regulate it very, very quickly?

I think that led to the Biden executive order, which we have publicly denounced in certain categories. That executive order led to a lot of the conversation that I think we're having in the states, a lot of the bad bills that we've seen come through the states, and a number of federal proposals that have not been very well thought through either.

And look, I think people were sitting around saying, “Oh, well, was it just some testimony from these CEOs that did this?” And the answer is no. From my point of view, the effective altruism community, for 10 years, backed by large sums of money, was very, very effective at influencing think tanks and nonprofit organizations in D.C. and the state capitals to push us in a direction where people are very fearful about the technology. That has significantly shaped the conversation we're having throughout D.C. and the state capitals and on a global stage.

You know, the EU enacting the EU AI Act—we're public on that. There are a lot of very, very problematic provisions in there. This banner of safetyism came from this 10-year head start that these guys have had.

When I think about the history, I always have to smirk—or smile, or try to laugh it off—when people write these articles about the fact that the AI industry is pumping all this money into the system. Certainly, I'm not suggesting that there's no money going into the system. We're obviously active on the political and policy side; we're not hiding that.

But it is dwarfed by the amount of money that has been spent and is being spent over a 10-year window. And candidly, the reason that Matt and I have jobs is because we're playing catch-up. We're here to try and make sure that people understand what is actually going on in this conversation and to be a counterforce to this group of people and this idea—this ideology—that has been here for a long period of time. So that's kind of the briefer on this.

Yeah. I mean, companies, I think, were ready to consider some policy frameworks that were probably really going to be challenging for the AI sector in the long run. I think that's because I was at Meta, then Facebook, starting in 2011 and through 2019.

After 2016, there was aggressive criticism of tech companies. The general framing was, “You're not being responsible, and regulation needs to catch up. Governance of social media is behind where the products are.” Whatever you think about that, that was really the strong view in the ecosystem: The lack of governance has allowed problematic things to happen.

And so, when AI was starting to accelerate and you had certain prevailing political interests driving the conversation, companies rushed to the table. I think it was a group of 3, 5, or 7 companies who went into the White House and negotiated voluntary commitments. Yeah.

I mean, we don't even have to make the argument about the importance of representing Little Tech when you see that there is a set of companies that negotiated an arrangement for what it would look like to build AI at the frontier, with all current developers who weren't those companies and all future startups not represented at the table. I think that's why we started to think about the value of having more dedicated support around AI policy, because clearly the views of Little Tech companies aren't represented in the conversation.

Yeah. Well, let me just add one thing to this.

It's Mark and Ben's story. They've told it many times. I was in the meeting as well, and everything they've said has been 100% true and accurate. But there was a prevailing view among very, very powerful people in the previous administration that only 2 or 3 major companies would be able to compete in the AI landscape.

Because that was the case, they needed to be basically locked down and put into this incredibly restrictive policy and regulatory view. That was going to be kind of like an entity that was an arm of the government. I think that was the most alarming thing we had heard from the administration, on top of an incredibly alarming series of events that happened on the crypto side, including seeming to want to eradicate it off the face of the planet.

I think that all led to the position that we're in now, and certainly to Matt's hiring and us building out the team.

That narrative is clearly a very alarming—maybe the most alarming—version of this. But even since I've been in this role, I've heard other versions of it where people will say, “Oh, don't worry about this framework. It just applies to 3 or 5 companies, or it just applies to 5 to 7 companies.” And I think they mean that to provide comfort to us: “Oh, this isn't going to cover a lot of startups.”

But the view of the AI market where there are only a small number of companies building at the frontier is not the vision for the market that we have. We want it to be competitive and diverse at the frontier. The policy ideas that were coming out of the period Collin's talking about were dramatically different from where they are today, in a way that I think some people have even lost sight of exactly where we were a couple of years ago.

There were ideas being proposed not just by the government but by industry to require a license to build frontier AI tools and for them to be regulated like nuclear energy.

Which would be historic for software development.

Yeah. Right. Unprecedented. Yeah. And for it to be regulated like nuclear energy, with an international-level, nuclear-style regulatory regime to govern it. And we've moved—no matter what you think about the right level of governance, there are not a lot of people now who are saying what we need is a licensing regime where you literally apply for permission from the government to build the tool. But that wasn't that far in the rearview mirror.

Yeah. And look, we're also talking about bans on open source. We're still kicking around that idea at the state level. And look, for us who live and breathe the tech stuff on a daily basis, this sounds insane and crazy. But let me just make it a little bit more real, right?

The nuclear policy in the United States has yielded 2 or 3 new nuclear power plants in a 50-year period since these organizations were started. And look, some people are pro-nuclear, some people are anti-nuclear. I don't want to get into that debate. The point, though, is that that was not the intended policy of the United States of America. That was the effect of putting together this agency and what has come from that.

And I think, look—

Had we done the same thing in AI in that period of time, then you don't have the medical advancements, you don't have the breakthroughs, you don't have all of the things that come from this that are incredible. But beyond that, we lose to China—full stop. You lose to China, and then our greatest national security threat becomes the one who has the most powerful technology in the world, right?

And I think the early concern on open source was that we would be somehow giving it to China, but then we've seen with DeepSeek, et cetera, that they just have it anyway.

Yeah.

Yeah. Exactly. Right. Exactly. The idea that we could lock this down—I think Mark and Ben have talked about this. I think they've debunked that a number of times.

Yeah. Just to understand: For the previous administration, what was their calculus? Was it that they were true believers in the fears? Was it that there was some sort of political benefit to having the views that they had, especially on the crypto side? I don't understand what the constituency for an anti-crypto stance is. How do you make sense of the players, or the motives or motivations? I just want to understand the calculus there.

Yeah. You know, I think that's a really hard one to answer, and I'm not sure I can pretend to be completely in their minds. I think there are a couple of different competing forces here. One is: What are the constituencies that support that administration? What are the constituencies that support that side of the aisle?

And I think that, especially over the last 10 to 15 years, there has been a very, very heavy focus on consumer safety, which I think is a very important thing. We're obviously in alignment on that. I think everyone should be in alignment: We have to protect consumers, and we have to be able to protect the American public.

But I think that a lot of that conversation has been weaponized. I think that it is a big-time moneymaker. I think a lot of these groups either get backing from very, very wealthy special interests, or they are doing small-dollar fundraising off of quick hits like, “AI is coming for your jobs. Donate $5, and we'll make sure that we take care of this in Washington for you.” And that's a pretty easy manipulation tactic. It's used by a bunch of people, but—

But I think that that held very true, right? And I think the other thing here is that personnel is policy. It's the old saying: personnel is policy. And I think a lot of the individuals who were in very senior decision-making roles within that White House and that administration came from this sort of consumer protection background. That was their constituency.

They were put in this position to come after private enterprise—that was the goal. There's this whole idea out there, I think among some of those folks, that, as Senator Warren has proposed many times, if you're not going after and getting people on a regular basis in the private sector, then you're not working hard enough.

And I just think that that is probably the second thing. And the third is that we're at this very weird moment where being a builder and being in private enterprise is a bad thing to some policymakers. You're not doing good because you're earning a profit, and they certainly won't say that, but the activities and the things that they're doing are 100% aligned with that type of idea. So I think that's the basic crux of it.

I think the things that motivated that approach were done in good faith. And I think it's what you alluded to earlier: I don't share this view, but there are a lot of people who believe that social media is poorly regulated and that, because policymakers were asleep at the wheel, we woke up at some point—I don't know, sometime in the 2014 to 2018 period—and realized that we had technology that we thought was actually not good for our society.

And I think that, whether or not you think that that's true, that has been a widely held view. It's a view held on the right and on the left. It's a bipartisan view. And so I think when this new technology came on the scene, this was a do-over opportunity for policymakers, right? We can get this right when we didn't get the last thing right. And so I understand that motivation. It makes a lot of sense.

I think the thing that we strongly feel is that the set of policy ideas that came out of that good-faith belief were not the right policy ideas to either protect consumers or lead to a competitive AI market. Many of the politicians who were pushing concepts that would have really put a stranglehold, I think, on AI startups would have led to more monopolization of a market that already tends toward monopoly because of the high barriers to entry.

Those politicians, 3 years before, had been talking about how problematic it was that there wasn't more competition in social media. And then all of a sudden, they're behind a licensing regime, which is not—I don't think there's much economic evidence that licensing is pro-competitive. It typically is the opposite. The disagreement is less with the core feeling—we want to protect people from harmful uses of this technology—and more with the policy concepts that came out of that feeling, which we think would have been disruptive in a problematic way to the future of the AI market.

Yeah. Right. Anecdotally, it seemed from afar that some of the concerns early on were almost to match social media—around disinformation or even DEI concerns. And then people were trying to make sure the models were compatible with the speech regime at the time, but then it kind of shifted to: Oh, wait, are there more existential concerns around jobs? Or is AI even like nukes, in the sense of people doing harm or AI itself doing harm? But it seemed to escalate a bit, and maybe aligned with that testimony that you alluded to.

I experienced it as feeling like the goalposts always move. And one of the things that I started asking people when I was really trying to settle into this regulate-use-not-development policy position is: What do we miss? If we regulate use primarily using existing law, what are the things that we miss? And I haven't gotten very many clear answers to that.

You can't do illegal things in the universe, and you also can't use AI to do illegal things. And typically, when people list out the set of things that they're most concerned about with AI, they are typically things that are covered by existing law. Not probably exclusively, but primarily. And so that at least seems like a good starting point.

Some of the other issues that I think are understandably ones that we should be concerned about have a range of different considerations associated with them. If you're concerned about misinformation or speech that you think might not be true or might be problematic, there are significant constraints on the government's ability to regulate that. The First Amendment imposes pretty stringent restrictions, and I think for very good reason, because you don't want the government to dictate the speech preferences or policies of private speech platforms, for the most part.

And so those issues might be concerns, but they're not necessarily areas where you want the government to step in and take strong action. There are things that we should probably do as a society to try to address those issues, but government regulation maybe isn't the primary one. And again, in most of the things that people are most concerned about—real use of the technology for clear, cognizable, real-world harm—existing law typically covers it.

I have a theory on this. I think everything that Matt just said is spot on.

But you know, then you’re kind of sitting around, scratching your head. It’s like, okay, well, if use covers it and there hasn’t been a very fair rebuttal as to why use is not enough in terms of focus on the policy and regulatory side, what’s the answer?

I think we’re experiencing this pattern on the crypto side, too, which is that we’re having a very spirited debate on how to regulate these tokens and how to launch a token in the United States. Is it a security, or is it a commodity? This is an age-old debate that’s plagued traditional securities laws for years, but also certainly the crypto industry.

What we have found is that there are a number of people who have entered this debate who are actually trying to get at the underlying securities laws. They want to reform securities laws. They don’t want to reform crypto laws that involve securities.

And this is their only venue by which they can enter that conversation, because we’re not having—there’s no will from Congress or policymakers to go and overhaul the securities laws right now. It’s just not there. But what is moving is crypto.

So there are all these people who are now trying to enter this debate and saying, “Oh, we should relook at this.” Well, this doesn’t have anything to do with it. We shouldn’t be entering this conversation. Yet they’re still pushing, right? And that’s kind of muddying the water.

I think a very similar thing is happening on the AI side, which is that there are a number of members of Congress who feel like, “Well, we missed it on the 1996 Telecommunications Act. We didn’t do good enough around then, so we need to rewrite the wrongs” through the venue of an AI policy conversation.

Because if you think about it, assuming that use doesn’t go far enough for someone—and this is the same conversation that we’re having in California right now, or in Colorado right now—if use does not go far enough, okay, well, then it would be really simple if you could have a privacy conversation around this. If you could have an online content moderation conversation or an algorithmic bias conversation around it, you could do all of that and wedge it through AI.

Assuming AI is actually going to be the thing that we all think it’s going to be, now you’ve put basically a regulatory funnel on the other side. You’ve put a mesh screen where everything has to run through AI, and therefore it runs through this regulatory proposal you put together.

Yeah. The thing that I’ve really been wrestling with in the last few weeks is whether those kinds of regimes are actually helpful in addressing the harm that they purport to want to address. Colorado is a really good example. There are all these bills that have been introduced at the state level. Colorado is the only one that’s passed so far that set up this regime where you basically have to decide: Are you doing a high-risk use of AI or a low-risk use of AI?

And this would be for startups that don’t have a general counsel, don’t have a head of policy, and can’t hire an outside law firm to figure it out. You have high-risk and low-risk, and then if you’re high-risk, you have to do a bunch of stuff. Usually, that means impact assessments and sometimes auditing your technology to try to anticipate whether there’s going to be bias in your model in some form.

Maybe an impact assessment helps you figure that out a little bit, but it’s probably not going to eliminate bias entirely. It certainly isn’t going to end racism in our society.

Colorado’s governor and attorney general have put pressure on the legislature to roll back this law because they think it’s going to be problematic for AI in Colorado. There was just a special session there to consider various alternatives. One of the alternatives introduced proposed codifying that the use of AI to violate Colorado’s anti-discrimination statute is illegal.

That’s consistent with the regulate-harmful-use framing that we’ve talked about. Instead of having this amorphous process where maybe you address bias in some form, maybe you don’t, this goes straight at it. It’s not a bank shot. It goes straight at it: If someone uses AI in a way that violates anti-discrimination law, that would be illegal and could be prosecuted; the attorney general could enforce it.

I still don’t understand why that approach is somehow less compelling than this complex administrative paperwork approach. I think it’s kind of the reason that Collin’s describing, which is that people want a different bite at the apple of bias, I suppose. But it’s not clear to me that it’s actually the best way to effectuate the outcomes that you want, as opposed to just criminalizing or creating civil penalties for the harm that you can see clearly.

It’s also—I mean, in policymaking and bill writing, it’s really easy to come up with bad ideas.

Yeah.

I think that’s part of the reason that people think that we are anti-governance. When we—I mean, Collin, again, he lived this history. I’m coming in late to it, but as we were ramping up our policy apparatus, these were the ideas in the ecosystem: licensing, nuclear-style regulation, FLOPs-threshold-based disclosures, really complicated transparency regimes, impact assessments, and audits. Those are a bunch of ideas that we think are not going to help protect people and are going to make it really hard for low-resource startups.

And so we’ve been trying to say, “No, no, no, don’t do that.” And so that sounds like deregulation, but for whatever reason, it’s been hard so far to shift toward, “Here’s another set of ideas that we think would be compelling in actually protecting people and creating stronger AI markets.”

Right now, we don’t see terrorists or criminals being aided 10,000x with AI in performing terrorism or crime. When I ask people, “What are you truly scared about? Give me a concrete scenario,” people will say, “Oh, what about bioterrorism or something? Or what about cybersecurity theft or something?” We seem very far away from that.

Is there any amount of development in the next few years—any amount of breakthroughs—where you might say, “Oh, maybe use isn’t enough”? Or do we think that that will always be enough?

I think it’s conceivable. And I think we’ve been open about that: We think existing law is a good place to start. It’s probably not where we end. Martin Casado, one of our general partners, wrote a great piece on marginal risk in AI, basically saying that when there’s incremental additional risk, we should look for policy to address that risk.

The situation you’re describing might be that. I think what you’re getting at is a really important question about significant potential harms that we don’t yet contemplate. We get asked often about our regulate-use, not-regulate-development framework. Are you just saying that we should address issues after they occur?

I understand why that’s a concern. There might be future harms. And wouldn’t it be nice if we could prevent them in advance? But that is how our legal system is designed.

Typically, when you talk to people about ways that you could try to address potential criminal activity or other legal violations ex ante before they occur, that’s really scary to people. Eric, what if we just learned a lot of information about you and then predicted the likelihood that you might do something unlawful in the future? And if we think it exceeds a certain threshold, then we’re going to go and try to take action against you before you’ve done it so that we can prevent future crime.

You’re laughing because it’s laughable. We don’t want a kind of ex ante surveillance, both because it feels invasive, but also because it often is ineffective. We might run some test that shows that maybe you’re predisposed to some kind of criminal activity, but we don’t know until you’ve done it whether you’re actually going to do it.

That kind of approach, again, is motivated by a really valid concern and a valid desire to prevent harm. What if we could prevent harm before it occurs? The challenge is that the regulatory framework probably won’t do that. It probably won’t have the effect of preventing harm. And there are all these costs associated with it, mainly, from our perspective, inhibiting startup activity.

Yeah. Mark once told me on a podcast—he told me this joke: A man goes to the government and says, “I go to the government because I have this big problem.” Now I get a lot of regulation; now I have 2 problems.

Okay, let’s talk about the state of AI policy today. There’s a lot that’s happened in the last few months with the moratorium and the action plan. What are some of the things that we’re excited about right now and looking forward to, and some of the things we’re less excited about right now? Why don’t we give a breakdown of where we’re at right now?

So, I think, given what Collin described about where things were a couple of years ago, it’s great to see the federal government—certainly the executive branch, but not just the executive branch.

I think this is true in Congress, across both aisles, with support for frameworks that we think are much better for little tech. So we're trying to identify areas where regulatory burden outweighs value and where we can rightsize regulation to make it easier for AI startups. As Collin said, support for open source—we were in a really different place on that a couple of years ago. Now it seems like there's much more consensus, again, actually, at the end of the last administration and in the current administration, around the value of open source for competition and innovation.

National AI Action Plan also had great stuff in it about thinking through the balance between the federal government and state governments, which is something that we've done a lot of thinking about. There's an important role for each, but we think the federal government should really lead regulation of AI development. States should police harmful conduct within their borders, and I think there's stuff in the action plan that would try to ensure those respective roles.

There's also a lot of stuff in the action plan that wasn't really talked about much. It wasn't the headline-grabbing stuff, but I thought it was incredibly compelling in terms of, again, trying to create a future for AI that just works better for more people. A really good example is the stuff on worker retraining, which focused on different programs that could help workers if they're displaced as a result of AI, as well as monitoring AI markets and labor markets to make sure that we understand when there are significant labor disruptions.

So I think it gets at a point that you were alluding to a couple of minutes ago: What happens when there's something really disruptive in the future? Can you predict with certainty that there won't be this crazy, disruptive thing? No, we can't. There might be significant labor disruption.

Others at the firm have talked extensively about how there are typically worries about labor disruptions when there's new technology introduced. Typically, there are increases in productivity that end up being good for labor overall. We think that's the direction of travel, but you never know. We can't predict it with certainty.

So I think it's a really strong step to try to just monitor labor markets to see what the disruption might look like, so that we're set up to take strong policy action in the future.

Can I just say one thing about the AI Action Plan?

Sure.

I don't want to juxtapose what we saw under the Biden administration, which was an incredible amount of activity, with the incredible amount of activity under the Trump administration. But I view these executive orders and plans that come out from an administration as very, very important. Some of them have real policy: They direct the agencies to do things, to come out with reports, and then take on rulemakings and things like that.

But from an AI Action Plan perspective, for me, it was so significant because I think it turned the conversation on its head. Before, it was, “We have to only focus on safety, with a splash of innovation.”

Yeah.

Right. And that dynamic and that shift of rhetoric is incredibly important, because what that does is signal to the rest of the world—to other governments—that this is the position of the United States and will be the position for the next 3½ years. This is the position of the United States to Congress.

So when Congress is looking at potentially taking up pieces of legislation or taking actions, or even holding committee hearings—which, for the broad base of what we're talking about, are fairly insignificant—all of that is kept in mind. Now the conversation has shifted significantly, and that is really, really important.

Speaking of winning, Collin, I'm curious for your thoughts on AI policy vis-à-vis China, whether it's export controls or any other issues we care about.

Yeah, I mean, look, first and foremost, we've talked about it already: We have to win, right? I think that is at the main thrust of a lot of what we're doing here and a lot of the way that we think about this from a firm perspective.

I think, first, it's making sure that the founders and builders can build appropriately, with appropriate safeguards and an appropriate regulatory structure. The second is: How do we win and make sure that America is the place where AI is probably the most functional and foundational vis-à-vis China?

I think there has been a long conversation about the AI Diffusion Rule that came out from the Biden administration, specifically on export controls. Many, I think, panned that proposal. A lot of people suggested it was probably too restrictive and wasn't the right way to think about things.

We have spent most of our time—Matt, leading this effort, has spent most of his time—specifically focused on how we are regulating the underlying models and how we are regulating, hopefully, the use of these models, versus specifically focusing on the export-control piece.

What I will say, though, is that some of the proposals that came out from the Biden administration, some of the proposals we've seen at the state level, and some of the proposals we've seen at the congressional level from a federal standpoint that dealt specifically with export controls on models themselves are very concerning.

We're still having this conversation. There's a policy set that's been kicked around for a while called the outbound investment policy, which is basically about how much U.S. money from the private sector is flowing into Chinese companies. I think it's a very noble, laudable concept, and we're super supportive of it. We are a very America-first organization here; we're investing primarily in American companies and American founders. So we're very supportive of it.

But when you edge into the idea that we might inadvertently ban U.S. open-source models from being exported across the country, by definition of open source, there are no walls around these types of things. So that's one of the areas that we've been very, very focused on.

I think it's obviously very important to make sure that we don't have these very powerful technologies—U.S.-made technologies—in the hands of our Chinese counterparts, the PLA, and the CCP, using this against us. But I also think that we need to make sure that we're not extending too far and limiting the power of open-source technologies to be the platform around the world.

The final point that I'd make here is that we do ultimately and fundamentally have a decision to make as the U.S.: Do we want people using U.S. products across the world, which helps for a whole bunch of different reasons, certainly including soft power from a national security perspective? Or do we want people to use Chinese products?

The more that we lock down American products, obviously, the more the Chinese will enter those markets and take a land grab in that space.

Why don't you get into more of what happened with the moratorium and the fallout that ensued?

I think this one's a bit complicated. There was a perception about the moratorium when it came out that it would have prohibited all state laws from existing for a 10-year window. Obviously, that's a long period of time. I'm not sure we would necessarily completely agree with that policy stance.

That, from our point of view, is a misinterpretation, for a whole bunch of different reasons, of what the language actually said. But sometimes in D.C.—a lot of times in D.C.—perception is reality, and that kind of took hold.

I also think that there are strong competing forces, like we've discussed, from the doomer crowd or the safety crowd, that were very, very anti-moratorium and had used all of their tentacles that they've spread out over the last decade to try to move in and kill this. I think they also were successful in leveraging some other industries to try to come in and also move forward to try to kill this thing.

By virtue of the vehicle—the underlying procedural vehicle, this reconciliation package that it was moving in—it was a partisan exercise. It was going to be Republicans and Democrats, and that was that. There was nothing, even a prominent AI policy, that was going to be dropped into a reconciliation package that was ever going to drag Democratic votes over it, because it was such a big Christmas tree-style thing that had all kinds of tax-reform positions, et cetera.

If you're in one of those situations, the margins on the votes become very, very small. So all it took was 1 or 2 Republican senators hitching their wagon to some of these ideas that were out there to tank this thing, right?

I think that's going to be a situation that you're going to fight in any sort of political, policy, or legislative outcome, or any sort of issue that you're going to be running within Congress. But I think, more so than anything, we heard this repeatedly from a whole bunch of different people—and this is what we've also experienced—the industry was just not organized well enough.

That's not just the industry; it's also the people who care about this thing who aren't actually industry stakeholders. The stakeholders who were in favor of some level of moratorium or some level of preemption were just not organized. I think that was both an eye-opening moment and an important moment, because what we have done in the preceding 3 or 4 months since this thing has gone down is take a long, hard look at what we need to do collectively, from a coalition, to be in a better position next time we're there.

What does that look like? First and foremost, it comes with writing, doing podcasts, and talking about these things—talking about the details of what's actually in these proposals and what they actually mean for states and the federal government—to make sure that we're fighting through the FUD that's coming through, because it's always going to be there. There's misrepresentation all over the field.

The second piece is let's all get on the same page, which I think we've worked very hard to do. Where we can find alignment, I think we've found that alignment between big, medium, and little. The third, and probably the most important, is what we're doing on the political advocacy side to make sure that we have the appropriate tools to push forward in a way that ensures America continues to lead and that we don't lose out in this race to China.

That's part of the reason that we recently announced our donation to the Leading the Future PAC, which will have several different entities underneath it. I think it's designed to be that political center of gravity in the space, and it will fight at the federal level and at the state and local levels. We're happy to be a part of it, and I expect there will be others that join this common-cause fight on the AI side.

If we could wave a wand, what would we like to be done at the state level? What would we like versus the federal level, and how should we think about that interplay compared to where we're at now?

The helpful answer here comes from the Constitution. The Constitution actually lays out a role for the federal government and a role for state governments. The federal government takes the lead in interstate commerce, so governing a national AI market and governing AI development, we think, is primarily Congress's role.

Sometimes when people say that, I think what other people hear, for some reason, is that states should do nothing. We have tried very hard to be deliberate in not saying that and making clear that states have an incredibly important role to play in policing harmful conduct within their jurisdictions.

Criminal law is a perfect example. There is some criminal law at the federal level, but the bulk of criminal law is at the state level. When you think about routine crimes, if you were going to prosecute a perpetrator, it's likely that would occur under state law. To the extent we want to take account of local activity where there's criminal conduct involved, and we want to make sure that the laws are robust enough to protect people from that activity, that's going to be primarily state law.

Oddly enough, as Collin is describing, this isn't the delineation that we started out with. There are a lot of state laws that have taken the approach—sometimes explicitly—that Congress hasn't acted, so we have a responsibility to act. That's true to some extent: states can act within their constitutional lane. Some of what states have done has gone outside that lane.

We just this week released a post on potential dormant Commerce Clause concerns associated with state laws. The basic idea is that there's a constitutional test that says states cannot excessively burden out-of-state commerce when that greatly exceeds the in-state, local benefits. Courts actually weigh that. There's a balancing test: Do the harms and costs to out-of-state activity significantly outweigh the benefits on the local side?

We think that, at least for some of the proposals that have been introduced, it's likely that they won't. The benefits are somewhat diminished relative to what the proponents think they are, and the costs are significant. The cost of a developer in Washington state complying with a law in California or a law in New York is going to be significant.

Our hope is not that the dormant Commerce Clause ends up serving a function that makes it hard for states to enact laws, but actually just serves as a guidepost for states around the kinds of laws that they might introduce. I think it pushes in the direction that's consistent with our agenda, which is to take an active role in legislating and enforcing laws that are focused on harmful use.

Looking at the next 6 months to a year, what are the issues that we're most focused on or thinking about that are going to be playing a role in the conversation?

I think it's first and foremost some level of federal preemption. I want to be very specific about this. We're not talking about preempting all state law. We're talking about making sure that we have a federal framework specifically for this model regulation and, hopefully, how models can be used.

I think that's going to be so critical, because just like any other technology, no technology can live under a 50-state patchwork. That's been the biggest issue that we've been fighting over the last year and a half or so.

I think there are some other policy sets beyond that that can kick into workforce training. I think there are some literacy things that should be coming up. Obviously, there's a huge, robust conversation around data centers and energy that will be really important.

Above all, I think most of our time and energy will be focused on trying to have some level of federal standard here to try and drive the dividing line between the federal and state governments, which I think Matt has already done a ton of great work on.

I think this is just a super exciting policy moment for AI. The last couple of years, I think, there have been a bunch of ideas that have been proposed, and for the reasons that we've discussed, we think those ideas fall short both in terms of protecting consumers and ensuring that there's a robust startup ecosystem.

Most of those laws, I think, have actually not succeeded in passing. There were a number of laws introduced at the state level in this past year's legislative sessions that we thought had a strong likelihood of passing, and to date, none of them have passed.

Collin has also been building out the expertise, skill set, and capacity on his team. We just hired Kevin McKinley to lead our work in state policy, and he will help us take a real affirmative position in the legislative sessions ahead on what might actually be AI policy that's good for startups.

Instead of being in the position of saying no because we're starting late and with one hand behind our back, I think we're in a position to really try to articulate and advance a proactive agenda in AI that's compelling. I think Collin hit the main parts of it: ensuring proper roles for the federal and state governments, focusing on regulating harmful use, not development.

There are specific things that you can do there in terms of increasing capacity in enforcement agencies, making clear that AI is not a defense to claims brought under existing criminal or civil law, and providing technical training for government officials to make sure that they can identify and prosecute cases where AI is used in a harmful way.

Then there's all this infrastructure and talent stuff that Collin's describing: worker retraining and AI literacy. We've also given some thought to the idea, articulated by a number of lawmakers and included in the National AI Action Plan, of creating a central resource housed in the federal government—and you could also do it in state governments—that lowers some of the barriers to entry for startups, such as compute costs and data access.

We think that's really compelling in terms of ensuring that startups can compete. That idea, like many of these, is bipartisan. It's been supported by the current administration, and it was supported by leading Democrats over the last couple of years. That's the kind of thing we're hoping will get some traction in policy circles when we have the room and position to really advocate for an affirmative agenda.

We're not always in 100% alignment with other people in the industry, and I think that's true across big, medium, and little. There are other consumer advocacy groups that obviously feel differently about these things. For the most part, the industry is generally aligned on some level of a federal standard here and understands that the thing that won't work is a 50-state patchwork.

Yeah, and I think that's super important, because for the first time, you actually have this sort of alignment. If you have that sort of alignment, that's momentum that you can use to push things over the finish line and get something done.

I think the Trump administration, to its credit, has also been incredibly supportive of this idea.

That's an incredibly important point. One criticism usually raised, sort of implicitly, is, "Hey, you're the little guys, but often you align with the big guys. Aren't you just in favor of a deregulatory agenda that works for big tech?" One of the things that's really extraordinary about the little tech agenda is that it's really nonpartisan and doesn't take a position on big versus little.

It basically says, “Here’s the agenda, and when you agree with us, we’ll support you, and when you disagree with us, we’ll oppose you.” That’s not a party line. It’s not Big Tech versus Little Tech.

I think what we saw in the phase that Collin was referring to, initially, in the recent set of AI policy was a phase of divergence between Big Tech and Little Tech on the licensing regime. Big Tech was sort of pushing it; Little Tech was concerned about it. Then there was a period of convergence, and I think if you look at the National AI Action Plan comments across a range of different providers, as Collin’s saying, a lot of them had some core similarities.

Lots of large companies have advocated for federal preemption. We don’t oppose that just because big companies are advocating for it; we think that’s good for startups. I think it’s possible, and I’m curious—I mean, Collin really understands this in a way that I don’t—how the political chips will fall. I think it’s possible we’re in a period of some divergence.

And one thing that we hear repeatedly, which is sort of funny, is people will bring us stuff and they’ll say, “Industry agrees with this, so we expect you to agree. You can’t—the industry’s already agreed. You can’t disagree.” And we say, “The big parts of the industry have agreed, but sometimes we agree with them, and sometimes we have different views.”

When we disagree, it’s not because we’re trying to blow up a policy process or make it difficult for lawmakers who are trying to move something forward. It’s because when we’re looking at it, we’re looking at it through this particular lens. I think—I hope it’s not the case—but I think there might be more fracturing in the months ahead.

Yeah, I agree with you on that. And by “people,” he means lawmakers, just to be specific. Yes, that’s a great place to wrap. Collin, Matt, thanks so much for coming on the podcast.

Thanks very much.