As the AI world gathers in Paris, the alliance between the U.S. government and Silicon Valley’s tech giants aims to break the European model based on digital regulation.

How to resist? For Columbia jurist Anu Bradford, Europe could win the battle if it stops fearing its own agenda.

With the arrival of Silicon Valley’s technocesarists in Washington, has the U.S. digital model described in Digital Empires changed radically?
In many ways, recent developments reinforce the description of the United States as a technolibertarian, market-driven regulatory model. This framework explains how Americans govern technology: tech companies are doubling their power and are now extending it over democratic institutions and political processes. Consider Elon Musk’s role in President Trump’s election and the donations these companies made for the January 20 presidential inauguration, securing front-row seats in the administration’s opening.
It truly shows an attempt by tech companies to assert their centrality in U.S. decision-making and ensure few policies restrict them.

But something new is also emerging: tech companies are now trying to convince and involve the U.S. government to help them win their regulatory battles in Europe. Mark Zuckerberg refers to European fines in these digital cases as “tariffs.” J. D. Vance previously said the U.S. might have to ask Europeans to drop some of these investigations if they want Washington to continue supporting Ukraine.
The novelty here is the attempt to politicize Europe’s digital agenda and try to exert additional influence through government involvement, which could entail a broader trade war and force Europeans to negotiate. In this sense, we see a greater external threat to the European rights-based model if companies align with the U.S. government, and the latter is willing to take stronger actions to elevate, intensify, and politicize the agenda.
With Trump, Silicon Valley companies are now trying to convince and involve the U.S. government to help them win their regulatory battles in Europe.

At the same time, we are witnessing an authoritarian shift, the type Tim Wu recently called “command capitalism” in The New York Times. In the new alliance between Washington and Silicon Valley, the president is choosing winners and losers to forge a new economy, which may involve turning digital tools into weapons. The U.S. is moving away from Biden administration policies that brought it closer to Europe. As recently as last year, a U.S. court ruled that Google had abused its monopoly position. That same court is now considering breaking up the company: we have truly witnessed a shift in the U.S. toward regulating these companies. That momentum is now in danger.
Instead, we see a mixture of doubling down on the market-driven model and some elements of state control in an increasingly intense trade and tech war.
We may see even more export controls and investment restrictions, along with various policies that are not really consistent with the market-driven model.
With Trump, regulatory tools are becoming weapons against his adversaries.

This Would Bring the United States Closer to the Chinese Model — Less Market-Oriented, More Centralized. Would We Talk About Convergence, as Giuliano da Empoli Suggested?
It is true that we will see increased state control through industrial policy and various restrictions on exports and technological investments. These measures resemble China’s state regulatory model.

There are elements of a kind of authoritarian control. In particular, it is not inconceivable that we might see the temptation to start using antitrust legislation as a weapon. All these regulations we have built could now be used to protect favorites or persecute political opponents. That would be an element where the U.S. would be playing Beijing’s game and following a kind of playbook line from the Chinese Communist Party. However, we must consider another element of paradoxical convergence: as the example of DeepSeek shows, it is market-driven investments that generate the most exciting innovations in China, not those controlled by the state.

On the other hand, Elon Musk is also a fundamentalist of free speech. This is a major deviation from the state-driven Chinese model: Beijing shows no sign of stopping its limits on freedom of expression or government control.

Another difference: President Trump is dismantling the administrative state in many ways. By emptying the state’s capacity, the U.S. may soon lack the public sector’s ability to control these tech giants. The tech companies seem to count on this and are already preparing for a less regulated environment. For example, Meta has announced changes to its content moderation policies, knowing that more harmful content will remain on the platform.

I don’t see that element as convergence with China, although I agree that the strong leader deciding who gets benefits in the digital economy moves the U.S. away from the free market model it traditionally adopted.

Nor do I see the Chinese government tolerating the amount of power that tech companies now have under Trump.

Trump is dismantling the administrative state. By emptying the state’s capacity, the U.S. may soon lack the public sector’s ability to control these tech giants.

What Does Trump’s Spectacular Style of Governance Change in This Regard?
What is very characteristic of Chinese policy-making, including digital policies, is unpredictability and rapid changes. They are tough on big tech companies, then they retract that toughness, etc. That is exactly Trump’s way of governing. He governs through unpredictability, and the only thing we can expect are very sudden moves. This is what happened with the TikTok saga.

First, he wanted to ban TikTok, then he said he did not want to ban TikTok.
Then TikTok is banned.
The Supreme Court says the ban is legal.
Two days later, TikTok is available again because Trump wants to reach a political deal.

That way of governing is exactly how China can govern as an authoritarian country.
It seems capable of orchestrating chaos of unpredictability in the form of pendulum swings. But the TikTok saga shows something else: his own preferences are not stable.

Does This Entail an Internal Risk?
Absolutely: the risk of seeing deep divisions emerge. Right now, he is Elon Musk’s best friend. But I believe many would bet that this romance won’t last four years. So, I think there is an element of unpredictability that now casts a shadow over any policy we might witness.

Internal divisions are already emerging within the Trump administration. An example is immigration. Elon Musk understands it is very important for foreign talent to come to the U.S. and benefit the American innovation economy, despite the MAGA group’s skepticism about any kind of immigration. J. D. Vance believes in enforcing antitrust laws. He has said positive things about former Federal Trade Commission Chair Lina Khan, known for being a firm critic of Big Tech. But many others want to dismantle all antitrust infrastructure. It is important to recognize that some internal forces are not exactly pushing tech policy in one direction.

Regarding DeepSeek, Which You Already Mentioned, Would You Agree with Gary Marcus Saying “The Race for AI Supremacy Is Over for Now”?
DeepSeek changes the narrative about AI.

The way the narrative has been pushed, and I think it has largely been a global narrative, is that the “big” players would inevitably win.

In other words, those with more data, more computing power, and who could afford to buy talent would necessarily have the advantage. The rest basically had to stay on the sidelines while reaping the benefits of the AI revolution.

This suggested that the U.S. would have the advantage because American companies had more data (the Chinese also have a lot, but that data is more national and less diversified). By the way, that’s why TikTok was a problem. TikTok actually gave the Chinese access to global data: in the race for AI, that was a problem for Americans.

DeepSeek changes this dynamic by challenging the “laws of scale” and, therefore, the idea that scale was a fundamental driver for success.

In fact, DeepSeek shows another rule: while Americans may be ahead in fundamental research, the Chinese are very good at commercial applications. They are innovators and manage to apply AI productively. They also demonstrated a more capital- and energy-efficient — and therefore more sustainable — way to develop AI than many had thought.

It could also democratize AI. DeepSeek’s success really calls into question whether projects like Stargate or accepting OpenAI’s pricing are necessary to succeed. This example could also be enriching for Europe.

We can only regret that the breakthrough came from DeepSeek and not Mistral, but we must recognize that Europe can still compete in AI application levels. We probably won’t build those fundamental AI models, but we can find ways to create new applications based on open-source models through innovative engineering. In many ways, it’s an exciting opportunity. It also shows that more competition generates more innovation. It’s quite exciting to see startups succeed when bit technology is gathering so much capital investment support.

The DeepSeek case demonstrates an important rule: while Americans may be at the forefront in fundamental research, the Chinese are very good at commercial applications.

In the Broader Context of the Political Capitalism War Between the United States and China, Is This a “Sputnik Moment”?
Yes, in many ways, U.S. restrictions, the tech war, export controls — including high-end semiconductor export controls — created the conditions for a Sputnik moment for China.

The U.S. incentivized China to innovate around those limitations, which often forces you to be more creative. It also shows how sometimes big companies have so much money that what they do, instead of innovating around those limits, is simply invest more money in the problem and focus on accessing more and greater computing power.

DeepSeek, essentially, is a good wake-up call: the AI race does not inevitably revolve around greatness. For me, this is welcome. But, obviously, many of us have some concerns about China’s dominance in the digital space because it comes with an authoritarian control model.

There Is Another Point to Consider in This Sequence: DeepSeek Was Made Possible Thanks to an Open-Source Model Designed by Meta. Can Open Source Be a Risk?
In many ways, this suggests that Meta deliberately positioned its model as open source, arguing that doing so would help democratize AI, ensuring everyone benefits rather than leaving control solely in the hands of big tech companies. Although Meta may not have explicitly used that exact term, the idea fits their stance.

However, they also acknowledge that allowing others to develop their model benefits them as well. So, their decision is not purely altruistic. Meta further argues that open access improves security and stability since their models incorporate strong security features.

In other words: when others develop these models, Meta could oversee the ecosystem to some extent, allowing for greater control. Conversely, critics argue that making these models widely accessible introduces significant risks, as bad actors can more easily exploit their capabilities. Once a model is open source, its original developers can no longer fully oversee how it is used. I do not have the expertise to resolve this debate or determine the proper balance, but it is important to recognize that, in many ways, this serves as validation for open-source AI advocates.

Some question whether DeepSeek is truly open source. Certain aspects of the model remain unrevealed and, as far as we know, the complete code has not been published. Because of this, staunch advocates of a purely open-source approach would not consider DeepSeek fully open source.

Do You Agree with Former Google CEO Eric Schmidt When He Writes in the Washington Post That, Regarding AI After DeepSeek, “the Balance of Power Now Seems to Be Shifting Along Two Key Axes: One Between the United States and China, and the Other Between Closed-Source and Open-Source Models”?
I agree with this in many ways and would add a third axis: the big versus the small.

Europe’s conclusion from this framework is that one is not out of the game if they remain innovative and can create applications. This opportunity undermines the hypothesis of the inevitability of scale.

This may already be included in Eric Schmidt’s argument, but it should be emphasized as an important conclusion from the DeepSeek saga for Europeans.

What Have Been the Main Changes in China’s Strategy Recently Regarding the Structure and Model of Its Digital Power?
China has been actively preparing for an escalation of the trade and tech war.

DeepSeek is just one proof of how China is accelerating what they consider the existential race toward technological sovereignty. They have been building stronger responses to U.S. export controls but also, and I believe more productively for China, have been finding new ways to innovate. They are clearly doubling down on the idea that they need to be more self-sufficient since they cannot rely on supply chains, as the U.S. is now weaponizing those dependencies. This is the biggest change.

Has China Stopped Regulating?
China continues regulating—but lightly. They are working on a general AI law, but they are also closely watching how they can regulate in a way that promotes Chinese innovation.

Another trend to watch: since the U.S. is increasingly hostile toward China and there is uncertainty about its immigration and visa policies, we might soon see some talent returning to China. For a long time, the U.S. has benefited from European and Chinese talent. But with Trump’s hostility toward immigration, that may change. Beijing will benefit from this, but again, it is also a golden opportunity for Europeans to welcome global talent.

This global talent has truly driven the AI revolution in the U.S., and the crucial question is: where will that talent go now? Before DeepSeek’s breakthrough, major U.S. institutions like Google and OpenAI could attract the best AI scientists because the prevailing view was that innovation must happen where the greatest computing power and data existed to trigger the next breakthrough.

DeepSeek shows that cutting-edge work can be done in places that allow freer, less bureaucratic structures that do not obstruct disruptive innovation. It is an opportunity for global talent to rethink what they need to succeed.

Can the European digital model survive Trump?

It is difficult to know where to start considering this, but one thing the Trump administration is willing to do is implement a “divide and conquer” strategy. This will be a great challenge for Europe.

From Greenland to trade, now everything is transactional, everything is negotiable, and the United States will try to exploit divisions within Europe.

We can no longer believe there will be some fundamental compromise based on values for transatlantic collaboration among traditional allies. At the same time, to the extent that Europeans can identify mutually beneficial agreements, those agreements can be made.

Trump has no principles or convictions: it is about the United States winning. But this also means that if we can identify where potentially both Europe and the United States can win, we can collaborate with Trump.

Europeans must also remember that the United States is a federation: there are many other actors, including states, with whom we can continue collaborating.

Specifically, what should the Union do to extend its digital regulation agenda?

The new European Commission now faces the challenge of implementing and enforcing the huge digital agenda built during the last Commission’s term.

All those legislations (the Digital Services Act, the Digital Markets Act, and the Artificial Intelligence Act) reflect the rights-based model that I believe captures how Europeans want the digital society to evolve.

More concretely, I think we can identify two challenges that will make applying this agenda much harder today.

We have already mentioned the external challenge: with Trump and Musk governing in Washington, tech companies are becoming increasingly assertive in their effort to co-opt the administration to politicize and delay the application of Union legislation on them, as demonstrated by statements from Zuckerberg and J. D. Vance.

The internal challenge is more paradoxical: it is Europe’s own paralysis regarding its regulatory agenda.

Geopolitical pressures have placed technological sovereignty at the center of the Union’s agenda. There is broad consensus that Europeans really need to improve their competitiveness, and the Draghi report was absolutely crucial in fostering that debate. I warmly welcome that conversation but have rejected the opinion that the Union faces a binary choice between digital regulation or innovation. That, to me, would be an irreparable mistake. Europeans can and must build the other pillars of the tech ecosystem to improve European competitiveness. But the way to do that is not by repealing the AI Act or not applying the GDPR. It is not commitment to digital rights that is holding Europeans back: the absence of an integrated digital single market and a true capital markets union are more powerful explanations of why European tech companies are not growing; that and our inability to recruit global talent.

Europe itself is petrified before its own regulatory agenda.

The widespread narrative that digital regulation is holding Europeans back is very dangerous, and the indiscriminate race to reduce bureaucracy everywhere constitutes a real challenge to the regulatory agenda.

How should these threats be addressed?

Regarding the external challenge coming from the fusion of Washington and Silicon Valley across the Atlantic, my view is that Europeans must absolutely refuse to politicize the digital agenda. They must repeat to Americans that “this is not politics, these are binding regulations.” In Trump’s world, law disappears and everything is negotiable. But this is simply not Europe’s world. We must say firmly: “We have no choice to negotiate on them.”

Contentious cases will ultimately be litigated by privacy advocates like Max Schrems, meaning they will go to the Court of Justice of the European Union. And the Court is not willing to negotiate. The United States can try to divide and conquer, pitting some member states against others, but also, if smart, different Directorates-General of the Commission against each other. The U.S. might say: “Why don’t you stop investigating X and we’ll suspend tariffs?” The Directorate-General for Trade might like that, but the Directorate-General for Connectivity, responsible for enforcing digital regulations, cannot afford it: Parliament would tear the Commission apart because it would be considered compromising the rule of law and European values.

To avoid these internal wars, the message must be clear: we will not fall into these traps because we have a court; if you want to fight, fight in court. European leaders must take this very quickly into the legal sphere. If they start giving in to this kind of blackmail and intimidation, there will be no end. For example, today the Union is asked to give up investigating X to avoid tariffs. Suppose they abandon the investigation on X. Tomorrow, it won’t be X, but the U.S. government will demand that the Union accommodate Apple, which does not want to comply with interoperability rules of the Digital Markets Act.

This dynamic would lead Europeans to total and colossal failure.

There are many other aspects that are more political and in which Europeans must engage politically. We cannot disinfect all transatlantic relations from geopolitics and political negotiation. But the digital agenda should not be part of this debate.

And about the internal challenge — the “fear of regulation”?

Europeans must distinguish between the real costs of digital regulations on European innovation versus the actual fundamental problems in the European tech ecosystem that are holding back the European industry.

The AI Act is not what is holding them back, nor is the Digital Services Act.

Of course, we can rethink and recalibrate things to ensure existing regulations are smooth and reviewed when necessary. But we cannot let that strategy displace the need for more fundamental policy reforms.

This brings us fundamentally to the relevance of the Brussels effect: while a Union losing competitiveness must face Trump’s openly imperialist ambitions, is its thesis still valid?

The greatest threat to the Brussels effect is not Trump — but the Europeans themselves.

If they lose faith in enforcing their laws, the concept will soon become empty.

The Union remains one of the largest and wealthiest consumer markets in the world. That has not changed, but if Europe does not get its competitiveness in order, its economy will suffer and its market will shrink every year. That, naturally, will weaken the Brussels effect.

But what continues to matter is not only market size but GDP per capita: the business models of these platforms remain largely tied to advertising. They cannot afford to completely withdraw from the European market and go to India because advertising revenue per user in developing markets like India is only a fraction of what it is in Europe. There is still more purchasing power in Europe, and there is no substitute for the European market anywhere else in the world.

The threat of withdrawing from the European market if EU digital regulations are enforced is also empty, because where would they go? They cannot go to China, they are not allowed to enter, and as I mentioned, expanding presence in India or other developing markets will not generate enough revenue in the near future.

Moreover, we have other allies: global trade does not revolve around the United States. Australians, Koreans, Japanese, and Brazilians are willing to regulate tech companies, and Europeans can and must collaborate with them on trade and digital governance.

The greatest threat to the Brussels effect is not Trump — but the Europeans themselves.

Some have also shown they are capable of resisting.

Exactly. Just look at what happened with the confrontation in Brazil between a Brazilian judge and Elon Musk. Who gave in? Musk. This is very revealing: we, as Europeans, the largest market in the world, can also stand up to American tech giants to enforce our laws. Certainly, in Europe we cannot outsource that and let Silicon Valley start writing the rules for the European Union.

We must trust one thing: Musk was forced to give in because he could not afford to lose the Brazilian market and he cannot afford to lose the European market.

What would be the advantages of turning to these other partners?

In Europe, European laws prevail. And the likelihood that the Brussels effect continues presupposes that tech companies continue respecting European legislation.

If these other markets adhere to the same kind of regulatory principles, then it is more likely these companies will apply these rules worldwide.

This would strengthen the Brussels effect both de facto and de jure.

I think it is really important that Europeans continue engaging with the rest of the world on trade, on digital regulation, and not be carried away by the idea that the world is nothing but the United States.

Washington will be a disruptive factor on most issues over the next four years. We still have to relate to them, but we must also deepen our engagement with the rest of the world. There is no guarantee that the rest of the world will emulate the American techno-libertarian model; other countries, outside Europe, are and will be mistreated both by the Trump administration and by the tech giants. This could lay the foundation for a new alliance.

Are you optimistic about the real capacity of Europeans to face the situation?

The worrying aspect is political weakness: ultimately, the European industry and politics are weak; France is a disaster and Germany is extremely weak today. They face elections with a growing threat of the AfD winning.

Macron can hardly claim a grand vision for Europe because he can barely do so for France. The political weakness of the most powerful European countries is opening Europe to more divisions, which I believe weakens European determination.

In his report, Draghi formulated three imperatives: more innovation, more investment, and more integration. Even those who in normal times would have resisted deeper European integration now have to seriously consider what Draghi said about the unprecedented magnitude of the problems we face today.

And that really shows how small the scale of individual member states is.

I do not advocate a European federation, but we must be pragmatic about the benefits of integration. We do not necessarily have to make the European federation an ideological commitment or an end goal we strive for. But if we consider issues like joint acquisition of military equipment, there are compelling arguments for deeper integration. We cannot afford to build a European defense on a fragmented defense procurement market.

Other countries, outside Europe, are and will be mistreated both by the Trump administration and by the tech giants. This could lay the foundation for a new alliance.

From digital giants to European leaders and entrepreneurs, the world of AI has gathered in Paris since yesterday. Are multilateral meetings like this summit useful?

I would say, to some extent, yes. I am neither naive nor overly optimistic, but it remains important.

Much of the world would benefit from better information and the exchange of views among policymakers. Even some more limited forms of cooperation can be beneficial.

So let me give you an example related to the global governance of AI.

We may never reach a global agreement on issues such as whether countries can deploy AI for mass surveillance. China will never agree to stop AI-powered facial recognition. But there are some common concerns that are not ideological. Like China, the United States does not want AI to fall into the hands of bad actors. They could collaborate on standards to make it harder for these actors to obtain AI models with powerful capabilities.

The same applies to the use of AI as a weapon in military conflicts. Even at the height of the Cold War, there were arms control treaties between the Soviets and the United States. Similarly, countries could now reach some limited agreements, for example, on how we use AI in a military environment. So even amid insoluble conflicts, summits like this can bring some — quoting Harvard scholar Gabby Blum — “islands of agreement.” We cannot afford to give up those achievements in this turbulent and conflict-filled world.