数字自由、AI监管与西方互联网之争｜a16z Show

AI is going to be more important, not less important, and so the proliferation of a Western AI stack needs to be a top priority for anyone who cares about freedom. The rules around AI are changing fast. There’s a lot of regulation abroad around digital safety and misinformation that is, in some ways, becoming the petri dish for lobbying groups and organizations in America to bring that here.

I think when we send signals in this policy domain, they should be signals that are consistent with free speech.

How can the U.S. government encourage the private sector to encourage free speech at their companies, even if they don’t have the ability to buy a company like Elon does and put his fingerprint on it?

Under Secretary Rogers, it is an honor to have you here. You’ve been, I would argue, one of the most vocal proponents of free speech and digital freedom in this administration. Free speech and anti-censorship are arguably—I would say—the most important American battles of our time. A lot of people don’t know how free speech and these battles are linked to public diplomacy in the State Department. To start, what is public diplomacy?

One of my favorite parts of getting this nomination was watching all my friends and family wait a polite beat to ask that question after congratulating me.

When we think of diplomacy, we’re ordinarily thinking about the relationship between the American government and foreign governments. Two ambassadors shake hands and strike a minerals deal—that’s diplomacy. Public diplomacy is different. It is my privilege and charge to lead the relationship between the American government and the foreign public.

That includes things like educational and cultural exchanges, such as the Fulbright Program. It includes our fast-twitch media-response assets in global public affairs. More relevantly than ever, it includes our engagement with the information environment—the backdrop, the operating system on which these conversations run.

Under the prior administration, that included things like the Global Engagement Center’s censorship efforts, which were the subject of the Murthy v. Missouri Supreme Court litigation. The State Department, along with other organs of the executive branch, would contact Twitter or Meta and say, “We think that these Charlie Kirk tweets are disinformation. You should take them down.”

That apparatus resided inside my part of the State Department. In the State Department’s reorganization, I acquired the Digital Freedom Office, which is basically, under my tenure, the opposite of that. I’m pursuing transparency, truth, and reconciliation on prior censorship, and I’m making freedom of expression a primary prong of our public diplomacy.

We live in a world where the internet mostly works in America. You can text, you can post, you can send memes, and you can criticize our government. But in a lot of countries, that’s just not the case, including some surprising countries.

When the U.S. says it supports internet freedom abroad—and you’ve been very vocal about this—even in places like Europe, what does it mean in practice?

This has gone through different permutations over time. If you think back to the era of the Arab Spring and Occupy Wall Street, I think the foreign-policy establishment was really excited about internet freedom because they saw, correctly, that it made conversation more disintermediated. It enabled a populist periphery to challenge legacy authorities, and that was generally good for openness and freedom and generally bad for entrenched, closed, controlling institutions.

After the Arab Spring, I think people got nervous that there might be an American Spring. In 2016, Trump and Brexit brought a lot of those anxieties to the fore, and you saw efforts to transform a lot of government internet-freedom and digital-diplomacy initiatives into disinformation curation, the suppression of what they called adverse narratives, and so on.

Getting inside this apparatus and seeing it firsthand, what you really see is that a lot of the people involved in these efforts had prosocial intentions. They wanted an ecosystem where people had more access to true information, where people were less likely to be misled by adversary information operations, which are a real thing.

Especially in the age of AI, we are seeing foreign adversaries intentionally introduce malicious false propaganda into our information arteries. To combat that, people just went overboard. We see this every time technology drastically changes and the structure of human communication changes.

When the telegraph was invented, people were worried that it was going to diminish everyone’s attention span. When the printing press was invented, people said, “All these heretics are going to be able to print their own Bibles. What are we going to do about it?”

We’re living through a moment that is on par with, or exceeds, any of those communications revolutions. Naturally, there is an urge to put the innovation back in the bottle, or at least harness it and control it so that legacy institutions can decide who it’s for and what it does.

I don’t think it’s consistent with American values or American interests to use or fetter technology that way. I think we want to use it to make people free.

I’d love to dive into that more because digital freedom is a national-security issue, especially with our adversaries. You touched on how it’s never been easier for adversarial nations to create disinformation and flood our channels.

What was the old mandate for digital freedom under the previous administration, and how have you changed it? What are the priorities for you right now?

The State Department has undergone a reorganization, and I’ve actually acquired new authorities in the National Defense Authorization Act to promote internet freedom that my office didn’t even have before.

We had a digital-freedom team that did some good work, and they’re still doing it: combating malware and spyware, foreign cyberattacks, and so on. I think our Digital Freedom Office was involved in so-called information-integrity initiatives with some international organizations that did a mix of good things and bad things.

I’m all for promoting content provenance—helping people determine, when they’re looking at a piece of content online, where it came from, whether it’s AI-generated, and whether it’s true. We just wanted to empower users to do that rather than institute opaque, tyrannical choke points far upstream of where the user sees the information.

Government-funded NGOs were making decisions about what arguments regarding pediatric transgender medicine people should be allowed to see. That was the kind of thing that was countenanced before; it’s not countenanced now. We’re much friendlier to initiatives like censorship-circumvention VPNs and to initiatives like X’s Community Notes, which put that power in the hands of the users, or the crowd, in a way that’s transparent.

We’ve talked on the stage today about the Project Maven moment as the turning point for tech and American dynamism. When I think about whether there’s a Project Maven moment that you would point to with the EU and other governments, I’m thinking about their current attacks on free speech, which I know you’ve been spearheading for the administration.

I love the touchstone to Project Maven because I feel like most people in tech and in this room think of it in a double-edged way. Project Maven catalyzed awareness that tech innovation and national security were one interest. Project Maven also sparked some very ideologically driven employee revolts at Google.

Both of those things are on everyone’s mind now with the proliferation of AI. All of the smart money knows, and everyone in this room knows, that AI is the next thing, and all of the policymakers know it, too.

The economist Tyler Cowen has this great phrase. He talks about AI with a Western soul, and I completely agree with him that that is the greatest soft-power tool we can possess. AI that reasons in an individualistic, rules-based way and prioritizes user consent, for example—those are all Western principles.

That is going to be the underlying reasoning model on which so much of the world’s communication and commerce runs. The proliferation of a Western AI stack is really a top priority for our entire administration. It should be a top priority for anyone who cares about freedom.

You asked about a similar catalyzing moment with the EU and foreign allied tech regulation. One question I get when I engage on these freedom-of-expression issues in places like the EU and the U.K. is, “Why are you being so hard on us? Russia and China censor the internet.”

The answer is that we are much harder on Russia and China. We treat these countries as adversaries in several key arenas. But when Russia and China censor the internet, they just firewall it off. They don’t purport to levy fines on American companies for allowing Americans to engage in First Amendment-protected speech on American political issues.

The EU did that. I think one key moment in August 2024 was when a now-former European Commission official, Thierry Breton, sent a letter to Elon Musk threatening him with regulatory penalties if he aired an upcoming interview on X with then-presidential candidate, former president, and now-current President Trump.

The interview hadn’t even happened yet, so it wasn’t like President Trump had said something the EU wanted to censor. It was more like, if you allow Donald Trump to speak on your platform, you will face regulatory liability in the EU. Then, in the same letter, Breton made reference to another ongoing investigation against X that they ostensibly said had nothing to do with speech.

This was the investigation that recently culminated in a €120 million fine. It was about whether your blue checks are assigned in an authentic, truthful way and whether your algorithm is transparent enough. In that letter, he basically said, “If you let President Trump speak, that is going to increase your exposure to adverse findings of liability in this other regulatory crackdown.”

This is an issue I litigated in my private practice before the Supreme Court, before I came into the administration. The viewpoint-skewed enforcement of ostensibly content-neutral regulations is both insidious and inevitable when you have something that is this politically pitched.

I think the idea that these European censorship laws—laws that make it illegal to insult a politician, as it is in Germany, for example, or illegal to blaspheme Islam, as some prosecutors recently argued in the UK—when you transpose those to a transnational internet and then tell American companies that they can face fines of up to 6% of global revenue if they transgress those laws, you force us to have this international conversation.

The conversation that I’ve had in some contexts, like on X, is confrontational and feisty, but there are other conversations that happen in diplomatic contexts that I think have been constructive. We have to have the conversation now because digital space is going to be more important, not less important, for international relations and commerce. AI is going to be more important, not less important. We need rules of the road that preserve that spirit of liberty and creation that enabled all the founders in this room to build what they did and made America the engine of innovation that it is.

Yes.

Absolutely. I want to get more into that because, yes, you’ve been very public. You’ve led some of the legal sanctions on some of these bureaucrats who did try to harm American companies. But as you said, the State Department is also—Secretary Rubio went over and gave, I would say, a triumphant speech about the relationship between the EU and the US, and how that needs to be forged so that we can jointly make sure that we are the leaders of the free world and that our adversaries, particularly on things like AI, don’t take control.

I would love to understand how that relationship is going, in your opinion. What are some of the things that have changed in the last several months in terms of these tactics working to make sure we have internet freedom?

I think we really value these alliances. As Secretary Rubio underscored in Munich, we engage with Europe on these issues because we want our European allies to be safe, strong, and prosperous like us—not just so that they can defend themselves vis-à-vis NATO, but because we comprise 1 civilization and have a lot of shared interests.

If you engage with Europeans bilaterally or multilaterally, they will all affirm—and I think a lot of them believe—that free expression is 1 of our shared interests. As a lot of these regulatory actions reach their final stages and we get the opportunity to negotiate and go back and forth on specifics, it’s really my hope that constructive progress can be made—not to the point where we have identical speech regimes in all of these countries, but to the point where there’s not an insidious and sweeping censorship contagion, as House Judiciary Republicans’ recent investigation suggested.

Now, I love that you brought up Tyler Cowen’s brilliant quote on AI with a Western soul. I think that’s a very important and pithy way to put it. A lot of people in this room are building with AI right now, and the rules around AI are changing fast.

There’s a lot of regulation abroad around digital safety and misinformation that’s in some ways becoming the petri dish for lobbying groups and organizations in America to bring that here. Maybe tell some of us what the most troubling legislation you’re seeing is in different pockets of the world and what tech should do to combat these sorts of free speech laws.

I love that question. I’ll talk about the AI regulatory landscape, and then what tech should do is a separate, juicy question that I’d love to get into.

I think copyright is a big issue. There are bones of the internet that we take for granted, built into its structure because the internet grew up in America. One of those is Section 230 of the Communications Decency Act. That’s why all these platforms exist: You can publish third-party content without being exposed to the same range of legal liabilities that a newspaper necessarily would for publishing the same content.

Section 230 is 1 of those structural features of the internet that has made it what it is. It’s actually an artifact of American law, which a lot of people don’t appreciate. The other is the fair-use doctrine.

We’ve seen a lot of good rulings from courts that training AI is fair use. If you showed a class of kindergarteners a library of books and they learned from it, teaching that class of kindergarteners would be fair use. If you pay for the same books and then use them to train an LLM, a lot of courts are saying that’s also fair use, which is great.

I think we are seeing regulatory temptations in other countries, including perhaps the EU, to treat copyright in a different way, which would be very devastating. Similarly, on the IP front, there’s a lot of interest in transparency, which is understandable. But if you force companies to disclose aspects of the AI that let foreign adversaries, for example, reverse-engineer the model weights, you’re really compromising the American and Western competitive edge.

I also think it would be very dangerous for AI to be subject to some of the content regulations that Europe currently imposes on what they call very large online platforms and very large online search engines. One frustrating aspect, as an American lawyer dealing with foreign laws, is that you expect statutes—especially statutes that would impose potential criminal liability or serious civil liability—to be very specific.

These statutes just say that you have to do risk assessments for all of these things: risk assessments for hate speech, risk assessments for speech that could adversely affect civil discourse, or hurt someone’s well-being. What does that mean? Does it mean the AI is too good, and people use it too much and it hurts their well-being?

Especially with an LLM, anyone who develops these things knows that you can be very careful and impose a lot of safety architecture, but it will emit unpredictable responses sometimes. I’ve seen draft legislation that imposes strict criminal liability if the LLM is even capable of generating certain kinds of content. I don’t mean child pornography; I mean content that you might not even be able to see under the First Amendment.

I think that kind of strict-liability regime, which degrades the Section 230 protection layer and creates adverse incentives against creative training of models, is something I don’t like to see.

Yeah. Now, I know you’ve been watching the Department of War’s negotiations with various LLM companies, and you’re a lawyer, as are several of the undersecretaries at State who’ve been very vocal on it. Maybe talk us through your views on contracting with the Department of War and, more broadly, how to think about AI, free speech, and alignment in a national-security context.

Look, I think we have several great AI companies in the United States. I defer to lawyers at the Department of War and elsewhere on which of them meet certain statutory thresholds, but I think what is essential from a national-security and national-defense perspective is that AI keeps its Western soul.

These really important debates about what kill shot an autonomous weapon should take or what the scope of data synthesis should be—these important debates should happen in the way that they’ve always happened under our Constitution: in courts, on statehouse steps, and in these crucibles of democratic deliberation. They shouldn’t be subject to the fiat of Silicon Valley executives or tech workers.

As you mentioned, we’ve seen woke tech workers in our prior era make decisions—for example, that it should not be permissible on Twitter to call a convicted sex offender a male, which she is. They’re entitled to that opinion, but the reason we have these democratic deliberative bodies and processes that have served us so well for 250 years is so that we can have these courts, which we’ve crafted to be deliberative and careful, think seriously about questions like unlawful search and seizure, what is too invasive, and what kind of surveillance should be allowed.

Then they write down, in a principled, consistent way, what positions we should abide by. Maybe it’s because I’m a lawyer, but I think the rule of law needs to be a touchstone, and I think you’ve seen that reflected in some of the administration’s positions.

Absolutely. Going back to something you said about Twitter, now X, I would argue that Elon buying Twitter, displacing the trust and safety team there, and changing the nature of free speech for a lot of the country was the most consequential moment in this movement. But it's not something that can be done now, and it's not a playbook that can be easily replicated. Elon can do it, but not many companies can.

So, if you have some advice for the people or companies in this room, how can the U.S. government encourage the private sector to encourage free speech at their companies, even if they don't have the ability to buy a company like Elon does and put his fingerprint on it?

I can think of several ways, and I'm pursuing all of them as best I can, for the government to encourage private companies to favor free speech. One is just not to create regulatory cudgels that can be wielded in a capricious, arbitrary way, like we've seen in Europe with this blue-check investigation, for example, or like we've seen in some of these debanking cases. I represented the NRA in a prominent case where the regulations being enforced were ostensibly viewpoint-neutral, but there was strong evidence that they were being enforced disproportionately against banks that allowed pro-gun groups to contract for financial services, and we've seen that before.

We should have a regulatory environment that's crisp and principled, where it's always clear what you have to do and what you can do to comply with the law. That's one thing we can do. Another thing we can do is, to the extent that we regulate tech companies—and there is going to be some regulation, as there already is with any emerging industry—we should have regulations that favor viewpoint neutrality.

I think we've seen, since 2020 and especially since 2024, the tech industry really come to the side of free speech in ways that it hadn't before, catalyzed significantly by Elon's purchase of X. That's great, and I think a lot of founders are not only patriotic, but they have that kind of gray-tribe freedom impulse in them. I think it's natural, but to the extent that there are incentives in our law favoring one kind of content moderation or another, we need to favor viewpoint neutrality.

I used the phrase “viewpoint neutrality” deliberately because that's a concept out of First Amendment law, but that doesn't mean founders shouldn't try or shouldn't offer users tools to curate and navigate the information environment in other ways. Spam content and pornographic content will behave differently than other content; people have more negative engagements with it. If you offer people the ability to see less of that in their feeds, or to see less content with foreign provenance, for example, that's not viewpoint-based suppression. That's not a viewpoint-based distinction, and I think our regulations should be kinder to that kind of content moderation.

Obviously, as government officials, we need to stand up for our companies and our industry when their interests and American political freedoms are threatened. If the U.S. government threatened Le Monde or Vivendi—these are French platforms—for hosting an interview with Emmanuel Macron, the French government wouldn't stand for it. We shouldn't stand for it, either. That's what these sanctions signaled.

Obviously, we have a lot of foreign policy priorities in the administration, and these really are critical allies with whom we share so much. But when we send signals in this policy domain, I think they should be signals that are consistent with free speech.

Absolutely. Well, as we always say, we invest in and support the Second Amendment so that we can enjoy the First. Under Secretary Rogers, thank you so much for being here and for the work you do.

Thank you so much for having me.