La soberanía digital: Europa construye su propio internet

And the moment it cracked open, spectacularly, was 2013.

Edward Snowden.

Sí, y lo que Snowden reveló no fue solo que la NSA espiaba a ciudadanos corrientes, lo cual ya era bastante grave, sino que espiaba a líderes europeos.

What Snowden revealed wasn't just that the NSA was spying on ordinary citizens, which was bad enough, but that it was spying on European leaders.

Se supo que habían intervenido el teléfono de Angela Merkel durante años.

Merkel's phone had been tapped for years.

Eso fue un escándalo diplomático mayúsculo, y convenció a muchos en Bruselas de que la dependencia tecnológica no era solo un problema económico sino un problema de seguridad nacional.

That was a massive diplomatic scandal, and it convinced many in Brussels that technological dependence wasn't just an economic problem but a national security problem.

Right, so you have this moment where the German chancellor is finding out her personal phone calls have been monitored by an allied government, and the primary instrument of that surveillance is an American company's infrastructure.

That's not a comfortable place to be.

La verdad es que Merkel reaccionó con una frialdad muy alemana en público, pero en privado, según se supo después, estaba furiosa.

Merkel reacted with very German restraint in public, but privately she was furious, according to what came out later.

Y ese momento aceleró dos cosas: primero, el debate sobre el RGPD, que es el Reglamento General de Protección de Datos, que se aprobó en 2016 y entró en vigor en 2018;

That moment accelerated two things: first, the debate around the GDPR, the General Data Protection Regulation, approved in 2016 and in force from 2018;

y segundo, el inicio de conversaciones más serias sobre construir infraestructura digital propia en Europa.

and second, the beginning of more serious conversations about building Europe's own digital infrastructure.

So the GDPR is one of those things that Americans mostly know as an annoying cookie banner.

But it's actually something much more significant than that, right?

It's a legal philosophy about who owns data.

Exacto.

Exactly.

El RGPD parte de la premisa de que tus datos personales son una extensión de tu persona, y por lo tanto están protegidos por derechos fundamentales.

The GDPR starts from the premise that your personal data is an extension of your person, and is therefore protected by fundamental rights.

No son un activo comercial que una empresa puede explotar libremente.

It's not a commercial asset a company can exploit freely.

Esa es una diferencia filosófica enorme con respecto al modelo estadounidense, donde los datos son esencialmente una mercancía que puedes intercambiar por servicios gratuitos.

That's a huge philosophical difference from the American model, where data is essentially a commodity you exchange for free services.

Here's what gets me about that framing.

Europe is essentially saying that the entire surveillance capitalism model, the one that funds Google and Meta and most of the free internet, is philosophically incompatible with European values.

That's a bold claim.

And it has enormous practical consequences.

Bueno, y así lo han entendido los tribunales europeos, que han multado a Google, a Meta y a Apple con cantidades astronómicas en los últimos años.

And that's exactly how European courts have interpreted it, fining Google, Meta, and Apple astronomical amounts in recent years.

Pero aquí llega el problema central: puedes regular las plataformas extranjeras, imponerles multas, obligarlas a cambiar su comportamiento, pero si toda tu infraestructura digital sigue dependiendo de esas mismas empresas, tu soberanía es muy relativa.

But here's the central problem: you can regulate foreign platforms, fine them, force them to change their behavior, but if your entire digital infrastructure still depends on those same companies, your sovereignty is pretty relative.

Es como prohibir un producto pero seguir dependiendo de quien lo fabrica.

It's like banning a product but still depending on the manufacturer.

So Europe decides it needs to build something.

And what it builds, or tries to build, is GAIA-X.

Which is, I have to say, one of the more ambitious and also more chaotic European projects I've come across in a long time.

Mira, GAIA-X es un caso de estudio fascinante precisamente porque ilustra tanto las ambiciones europeas como sus contradicciones.

GAIA-X is a fascinating case study precisely because it illustrates both Europe's ambitions and its contradictions.

La idea original era crear un ecosistema de nube europeo federado, es decir, no un monopolio centralizado como Amazon Web Services, sino una red de proveedores europeos interoperables que compartieran normas comunes de seguridad y soberanía de datos.

The original idea was to create a federated European cloud ecosystem, not a centralized monopoly like AWS, but a network of interoperable European providers sharing common standards around security and data sovereignty.

Sonaba muy bien en el papel.

It sounded very good on paper.

I sense a 'but' coming.

Es que el 'pero' es enorme, Fletcher.

The 'but' is enormous, Fletcher.

En primer lugar, los propios americanos que se suponía que GAIA-X debía contrarrestar, es decir, Amazon, Microsoft y Google, terminaron siendo miembros del consorcio.

First, the very Americans that GAIA-X was supposed to counter, Amazon, Microsoft, and Google, ended up as members of the consortium.

Lo cual es un poco como fundar una asociación de defensa del consumidor e invitar a las empresas que quieres regular a sentarse en el consejo de administración.

Which is a bit like founding a consumer protection association and inviting the companies you want to regulate to sit on the board.

That's, I mean, that's almost too on the nose as a metaphor for the problem.

You set up an alternative to dependence on American tech, and the first thing you do is invite American tech to run it with you.

La verdad es que hay una explicación pragmática para eso: muchas empresas europeas ya tenían sus sistemas integrados con Amazon Web Services o Microsoft Azure, y migrar todo de golpe era económicamente inviable.

To be fair, there's a pragmatic explanation: many European companies already had their systems integrated with AWS or Azure, and migrating everything at once was economically unfeasible.

Así que la idea fue incluirlos en el proceso y establecer normas que también ellos tuvieran que cumplir.

So the idea was to include them in the process and set rules they'd also have to follow.

Pero el resultado fue que GAIA-X se diluyó bastante y perdió parte de su fuerza original.

But the result was that GAIA-X got watered down and lost much of its original force.

So GAIA-X as a cloud play has had mixed results, to be generous.

But there's another front where Europe is actually making more concrete progress, and that's in artificial intelligence.

And specifically there's one French company that's become a genuine symbol of this whole effort.

Sí, Mistral es un caso realmente interesante porque rompió varios tópicos de golpe.

Mistral is a really interesting case because it shattered several myths at once.

Primero, demostró que se podía crear un modelo de lenguaje de primer nivel fuera de Estados Unidos.

First, it proved you could build a frontier language model outside the US.

Segundo, lo hizo con un equipo relativamente pequeño y con mucho menos financiación que OpenAI o Google DeepMind, lo cual era toda una declaración de eficiencia técnica.

Second, it did it with a relatively small team and far less funding than OpenAI or Google DeepMind, which was a statement of technical efficiency.

Y tercero, eligió una estrategia de código abierto, que es políticamente muy astuta porque hace que sea mucho más difícil acusarlos de intentar crear un monopolio europeo.

And third, it chose an open-source strategy, which is politically very shrewd because it makes it much harder to accuse them of trying to create a European monopoly.

The open-source angle is interesting because it cuts against the sovereignty argument in a way.

If your AI model is open source, anyone in the world can use it.

Including, in theory, your adversaries.

So you're not really controlling who benefits from it.

A ver, esa es una objeción legítima, pero creo que mezcla dos cosas distintas.

That's a legitimate objection, but I think it mixes two different things.

La soberanía digital no significa necesariamente que solo tú tengas acceso a la tecnología.

Digital sovereignty doesn't necessarily mean only you have access to the technology.

Significa que no dependes de otro para tenerla tú mismo.

It means you don't depend on someone else to have it yourself.

Si Europa tiene sus propios modelos de inteligencia artificial que puede usar, auditar y modificar sin depender del permiso de una empresa americana, eso ya es un avance enorme en términos de autonomía estratégica, aunque el modelo también lo use todo el mundo.

If Europe has its own AI models it can use, audit, and modify without needing an American company's permission, that's already a huge step toward strategic autonomy, even if everyone else uses the model too.

No, you're absolutely right about that distinction.

The dependency is the problem, not the exclusivity.

And I think that's actually where the conversation gets more interesting, because the dependency argument isn't just about espionage or data privacy.

It's about what happens in a crisis.

Exactamente.

Exactly.

Fíjate en lo que pasó con los semiconductores.

Look at what happened with semiconductors.

Cuando Estados Unidos decidió restringir las exportaciones de chips avanzados a China, de repente quedó claro para todo el mundo que controlar la cadena de suministro tecnológico es un instrumento de poder geopolítico.

When the US decided to restrict exports of advanced chips to China, it suddenly became clear to everyone that controlling the technological supply chain is an instrument of geopolitical power.

Y Europa se dio cuenta de que ella misma era vulnerable en ese sentido, porque no fabricaba prácticamente ningún chip avanzado en su territorio.

And Europe realized it was itself vulnerable in that sense, because it manufactured virtually no advanced chips on its own territory.

Which is why the European Chips Act happened.

Forty-three billion euros to try to build semiconductor manufacturing capacity in Europe.

TSMC is building a fab in Dresden.

Intel was supposed to build one in Magdeburg, though that's hit some serious turbulence.

Y ahí está el problema crónico de Europa, que es la distancia entre la ambición política y la ejecución industrial.

And there's Europe's chronic problem, which is the distance between political ambition and industrial execution.

Los europeos somos muy buenos escribiendo documentos estratégicos brillantes.

Europeans are very good at writing brilliant strategic documents.

La Comisión Europea produce libros blancos de una calidad intelectual impresionante.

The European Commission produces white papers of impressive intellectual quality.

Pero llevar eso a la práctica, construir fábricas, atraer inversión, crear ecosistemas industriales, es donde las cosas se complican.

But turning that into practice, building factories, attracting investment, creating industrial ecosystems, that's where things get complicated.

Look, I want to push back on something here, because I think there's a version of this critique that slides into being unfair.

The regulatory framework that Europe built around data protection, the GDPR, was initially mocked by American tech companies as bureaucratic overreach.

It is now effectively the global standard.

California modeled its own privacy law on it.

Brazil did too.

India is doing it.

No, tienes razón, y es un punto importante.

You're right, and it's an important point.

Hay un fenómeno que los académicos llaman el 'efecto Bruselas', que básicamente significa que cuando la UE establece una norma técnica o regulatoria, las multinacionales prefieren aplicarla a nivel global antes que mantener dos sistemas distintos para el mercado europeo y el resto del mundo.

There's a phenomenon academics call the 'Brussels Effect,' which basically means that when the EU sets a technical or regulatory standard, multinationals prefer to apply it globally rather than maintain two different systems for the European market and the rest of the world.

Así que las normas europeas acaban siendo normas globales por defecto, sin que Europa haya tenido que imponérselas a nadie.

So European standards end up becoming global standards by default, without Europe having to impose them on anyone.

The Brussels Effect is real and it's genuinely underappreciated outside Europe.

But here's the tension I keep circling back to.

Regulatory power and industrial power are not the same thing.

Europe has significant regulatory power.

It does not yet have equivalent industrial power in tech.

And regulations without industrial alternatives can feel a bit like telling someone to swim but taking away the pool.

Bueno, esa metáfora me gusta.

I like that metaphor.

Y creo que captura el debate interno que está teniendo lugar ahora mismo en Europa entre dos facciones.

And I think it captures the internal debate happening right now in Europe between two factions.

Por un lado están los regulacionistas, que creen que el papel de Europa es fijar las normas éticas y legales del uso de la tecnología, aunque no la fabrique.

On one side are the regulationists, who believe Europe's role is to set the ethical and legal standards for technology use, even if it doesn't manufacture it.

Y por otro lado están los industrialistas, que dicen que sin capacidad productiva propia, la soberanía digital es una quimera, y que Europa tiene que ser más agresiva apoyando a sus propias empresas tecnológicas.

On the other are the industrialists, who say that without productive capacity of its own, digital sovereignty is a pipe dream, and Europe needs to be more aggressive in backing its own tech companies.

And the Draghi report landed right in the middle of that fight.

The former ECB president and Italian prime minister produced this 400-page document last year that was basically a very polite way of saying Europe is falling behind and needs to get serious.

El informe Draghi fue importante porque venía de alguien que no puede ser descartado como un euroescéptico ni como un demagogo populista.

The Draghi report mattered because it came from someone who cannot be dismissed as a Eurosceptic or a populist demagogue.

Es un hombre del establishment europeo por excelencia, el hombre que salvó al euro con aquella frase de 'whatever it takes', y sin embargo dijo abiertamente que Europa está perdiendo la carrera tecnológica y que si no cambia su forma de invertir y de regular, las consecuencias serán muy graves.

He's the ultimate European establishment figure, the man who saved the euro with that 'whatever it takes' line, and yet he said openly that Europe is losing the tech race and if it doesn't change how it invests and regulates, the consequences will be severe.

So there's the regulatory piece, there's the industrial piece, and then there's a third dimension that I think often gets left out of this conversation, which is the cultural one.

Because digital sovereignty isn't just about where your servers are.

It's about whose values are embedded in the systems that shape your information environment.

Eso es fundamental, y es quizás el argumento más profundo a favor de la soberanía digital europeo.

That's fundamental, and it's perhaps the deepest argument for European digital sovereignty.

Los sistemas de inteligencia artificial no son neutrales: reflejan los valores, los sesgos y las prioridades de quienes los diseñaron.

AI systems are not neutral: they reflect the values, biases, and priorities of the people who designed them.

Un modelo de lenguaje entrenado principalmente con datos en inglés y diseñado en Silicon Valley va a tener una visión del mundo que no necesariamente coincide con la de un ciudadano francés, o español, o polaco.

A language model trained primarily on English data and designed in Silicon Valley is going to have a worldview that doesn't necessarily align with that of a French, Spanish, or Polish citizen.

I spent a lot of years in parts of the world where the information environment was controlled by an outside power, and the effect on how people thought about themselves and their own history was profound and usually damaging.

I'm not saying Silicon Valley is doing that deliberately.

But the effect can be real even without the intent.

Es que la cuestión de la lengua es especialmente sensible para Europa.

The language question is especially sensitive for Europe.

Cuando los modelos de inteligencia artificial se entrenan principalmente en inglés, las lenguas minoritarias, y aquí hablo no solo del euskera o el galés sino del catalán, del polaco, del húngaro, quedan sistemáticamente infrarrepresentadas.

When AI models are trained primarily in English, minority languages, and here I'm talking not just about Basque or Welsh but Catalan, Polish, Hungarian, are systematically underrepresented.

Y eso tiene consecuencias reales sobre qué idiomas sobreviven en el entorno digital y cuáles se van marginando progresivamente.

That has real consequences for which languages survive in the digital environment and which get progressively marginalized.

Which is why Spain, and the broader Spanish-speaking world, is actually doing something quite interesting on this front.

There are now specific AI projects focused on Spanish and its regional variants that are explicitly designed as a counterweight to English-dominant systems.

The language sovereignty argument and the political sovereignty argument are connected.

Sí, proyectos como MarIA del CSIC en España, o la iniciativa de la Real Academia Española para crear recursos lingüísticos de alta calidad para el español.

Projects like MarIA from Spain's CSIC, or the Real Academia Española's initiative to create high-quality linguistic resources for Spanish.

Pero también hay iniciativas paneuropeas para lenguas como el finlandés o el checo que tienen mercados demasiado pequeños para ser comercialmente atractivos para las empresas americanas.

But there are also pan-European initiatives for languages like Finnish or Czech that have markets too small to be commercially attractive for American companies.

Nadie en San Francisco va a invertir en entrenar un modelo de primer nivel en esloveno porque no hay dinero que justifique ese esfuerzo.

Nobody in San Francisco is going to invest in training a frontier model in Slovenian because there's no money to justify that effort.

So the market failure argument is real.

Commercial AI will optimize for scale, and scale means English and Mandarin, and maybe Spanish.

Everything else is a rounding error from a profit motive perspective.

Which means if you want those languages to thrive in the AI era, you need public investment.

Exactamente.

Exactly.

Y fíjate que este es uno de los pocos argumentos a favor de la soberanía digital que tiene consenso político relativamente amplio en Europa, porque trasciende la división entre derecha e izquierda.

And notice that this is one of the few arguments for digital sovereignty with relatively broad political consensus in Europe, because it cuts across the left-right divide.

Los conservadores lo enmarcan como protección de la identidad cultural nacional, los socialdemócratas lo enmarcan como corrección de un fallo de mercado, y los liberales lo enmarcan como inversión en competitividad.

Conservatives frame it as protecting national cultural identity, social democrats frame it as correcting a market failure, and liberals frame it as investment in competitiveness.

Todos llegan al mismo sitio por caminos distintos.

They all arrive at the same place by different routes.

I want to spend a minute on the counter-argument, because there is one and it's not stupid.

The counter-argument is that digital sovereignty, when you push on it hard enough, starts to look a lot like digital protectionism.

And protectionism has a pretty poor track record as an economic strategy.

La verdad es que esa crítica tiene algo de fundamento.

That criticism has some merit.

Si la soberanía digital significa simplemente que los europeos tienen que usar tecnología europea aunque sea inferior, eso es proteccionismo puro y duro y va a generar ineficiencias reales.

If digital sovereignty simply means Europeans have to use European technology even if it's inferior, that's pure protectionism and will generate real inefficiencies.

Pero si significa que Europa crea las condiciones para que puedan desarrollarse empresas tecnológicas competitivas en su territorio, con normas claras y acceso a capital, eso es política industrial, que es algo muy distinto.

But if it means Europe creates conditions for competitive tech companies to develop on its territory, with clear rules and access to capital, that's industrial policy, which is a very different thing.

The distinction you're drawing is between defensive sovereignty, just block out the Americans, and constructive sovereignty, build something that can actually compete.

And I think one of the more honest things you can say about where Europe is right now is that it's attempting the second while sometimes sliding into the first.

Sí, y esa tensión se ve muy claramente en el debate sobre la Ley de Inteligencia Artificial europea, que fue la primera regulación de IA del mundo.

That tension is very visible in the debate around the European AI Act, the world's first AI regulation.

Por un lado es un logro legislativo genuino, un marco serio para gestionar los riesgos de la IA.

On one hand it's a genuine legislative achievement, a serious framework for managing AI risks.

Pero sus críticos, incluyendo a los fundadores de Mistral, argumentaron en su momento que la regulación era tan exigente que pondría a las empresas europeas de IA en desventaja competitiva frente a las americanas y chinas, que no tenían que cumplir esas normas.

But its critics, including Mistral's founders, argued at the time that the regulation was so demanding it would put European AI companies at a competitive disadvantage against American and Chinese rivals who didn't have to comply.

So Mistral, the flagship of European AI sovereignty, lobbied against the European AI Act.

That's a genuinely delicious irony.

Bueno, mira, no lo llamaría irría exactamente.

I wouldn't call it irony exactly.

Lo llamaría realismo empresarial.

I'd call it business realism.

Mistral tiene que competir globalmente, no solo dentro de Europa.

Mistral has to compete globally, not just within Europe.

Y si tiene que cumplir requisitos regulatorios que sus competidores estadounidenses no tienen, eso le crea una desventaja estructural real.

And if it has to meet regulatory requirements that its American competitors don't, that creates a real structural disadvantage.

El problema de fondo es que Europa está intentando hacer dos cosas simultáneamente que generan tensiones entre sí: ser el árbitro ético de la tecnología global y ser un jugador industrial competitivo en ese mismo sector.

The underlying problem is that Europe is trying to do two things simultaneously that create tension with each other: be the ethical arbiter of global technology, and be a competitive industrial player in that same sector.

Right, and you can't fully be both at the same time.

The referee and the player are different roles with different incentives.

So the big question for the next decade is which one Europe actually chooses, or whether it manages to thread that needle in a way that nobody has quite managed yet.

La verdad es que yo soy moderadamente optimista, aunque sé que eso va contra el espíritu de los tiempos.

Honestly, I'm moderately optimistic, even though I know that goes against the spirit of the times.

Europa tiene ventajas reales: tiene talento técnico, tiene mercado, tiene capital, aunque distribuido de forma ineficiente, y tiene algo que no se puede comprar, que es décadas de experiencia institucional construyendo consensos transnacionales complejos.

Europe has real advantages: technical talent, a large market, capital albeit inefficiently distributed, and something you can't buy, which is decades of experience building complex transnational consensus.

Hacer que veintisiete países se pongan de acuerdo en algo tan técnico como una regulación de IA es, objetivamente, un logro político notable.

Getting twenty-seven countries to agree on something as technical as AI regulation is, objectively, a notable political achievement.

The extraordinary thing is that when you zoom out, this is one of the most significant geopolitical questions of the next thirty years.

Who controls the infrastructure of the digital world controls, in large part, the information environment and therefore the political possibilities of the people living in it.

And Europe is essentially arguing that that question should not be settled entirely by the market logic of a handful of American companies.

Y hay algo más que añadir, que es que el éxito o el fracaso del proyecto europeo de soberanía digital importa más allá de Europa.

And there's something more to add, which is that the success or failure of Europe's digital sovereignty project matters beyond Europe itself.

Si Europa demuestra que es posible construir una economía digital competitiva con normas estrictas de protección de datos y de uso ético de la inteligencia artificial, eso abre un camino para el resto del mundo.

If Europe proves it's possible to build a competitive digital economy with strict data protection and ethical AI standards, that opens a path for the rest of the world.

Si fracasa, el argumento por defecto será que la única forma de ser competitivo en tecnología es abandonar esas normas.

If it fails, the default argument will be that the only way to be competitive in tech is to abandon those standards.

Y eso sería una derrota para todos.

And that would be a loss for everyone.

I think that's exactly the right note to end on.

The stakes are not just European.

They're global.

And the experiment Europe is running right now, messy and contradictory and bureaucratic as it sometimes is, matters to everyone who uses the internet.

Which is everyone.

Bueno, Fletcher, por una vez no tengo nada que objetar a lo que has dicho.

Well, Fletcher, for once I have nothing to object to in what you just said.

Aunque seguro que en el próximo episodio encuentro algo.

Though I'm sure I'll find something for the next episode.