Europe Scrambles for Relevance in the Age of AI

With chatbot and AI development largely coming from the US, some EU entrepreneurs and politicians say local champions are needed to prevent a cultural flattening.
Collage with the European union flag with a brain shaped hole that shows parts of a circuit and a robot hand.
PHOTO-ILLUSTRATION: ANJALI NAIR; GETTY IMAGES

When a Finn talks to an AI helper like ChatGPT, they often get the sense that something is subtly wrong. “You really feel that this conversation is not the way that you would have a discussion in Finland,” says Peter Sarlin. For a start, Finnish people are known for a blunt approach to dialog and chatbots are usually tuned to be overly courteous. But there’s also the fact that most leading chatbots and the large language models behind them are developed in the US and trained on mostly US data. Cutting-edge AI products often come with a tonality that is essentially American.

Sarlin, founder and CEO of Helsinki-based Silo AI, one of Europe’s largest independent artificial intelligence labs, worries that in the age of ChatGPT, regional social nuances across Europe will start to disappear. As chatbots and large language models mostly derived from North American data become ubiquitous, the understanding of what normal conversation looks like “converges toward one,” he says.

That cultural flattening could have huge consequences as large language models start to power not just chatbots, but also many other digital services and products. American models might be able to output text in Finnish, but they can’t think—or appear to think—in Finnish. “These models usually have what you call a reasoning capability, and that reasoning capability should descend from data representing this region,” Sarlin says. “But if you look at most of the models out there, they are dominated by English, and dominated by English that represents North America.”

The concern isn’t just cultural, but economic. If closed source models owned by American companies come to dominate across Europe, the economic value will flow to them, Sarlin says.

The dominance of American models is driving many in Europe to talk about the concept of “AI sovereignty”: making sure that the core “digital infrastructure” behind the AI boom isn’t entirely controlled by private companies outside of the continent. Europe is investing heavily in supercomputers and AI research to try to catch up with the US and create domestic champions. But Europe’s AI challengers are starting from a long way behind. The continent lags a long way behind the US and China in the availability of capital and computing power. And it lacks big homegrown tech companies—the Microsofts, Googles, and Metas—which are vital conduits linking AI products to users.

“What is sovereignty when you don't have any champions?” says Raluca Csernatoni, a research fellow specializing in emerging technologies at Carnegie Europe, a think tank.

European preoccupations with the power of American tech aren’t new. Generation after generation of technology has been dominated by big US companies, whose products have become embedded into Europe’s social and economic infrastructure. Europe’s businesses run on Microsoft Office and Amazon Web Services, and its mobile devices rely on Apple and Google, which also run the app stores. European politics happens on WhatsApp, and its news media happens on Facebook, Instagram, and Twitter. Even the French watch Netflix.

And US tech companies operate on a different scale. Only two of the 10 most valuable public European corporations are in tech: German business software provider SAP and Dutch semiconductor equipment maker ASML. Six of the world’s 10 most valuable public corporations are US tech companies. Nvidia and Microsoft, the two largest, are each worth more than 15 times SAP.

That concentration of power is uncomfortable for European governments. It makes European companies downstream customers of the future, importing the latest services and technology in exchange for money and data sent westward across the Atlantic. And these concerns have taken on a new urgency—partly because some in Brussels perceive a growing gap in values and beliefs between Silicon Valley and the median EU citizen and their elected representatives; and partly because AI looms large in the collective imagination as the engine of the next technological revolution.

European fears of lagging in AI predate ChatGPT. In 2018, the European Commission issued an AI plan calling for “AI made in Europe” that could compete with the US and China. But beyond a desire for some kind of control over the shape of technology, the operational definition of AI sovereignty has become pretty fuzzy. “For some people, it means we need to get our act together to fight back against Big Tech,” Daniel Mügge, professor of political arithmetic at the University of Amsterdam, who studies technology policy in the EU, says. “To others, it means there’s nothing wrong with Big Tech, as long as it’s European, so let’s get cracking and make it happen.”

Those competing priorities have begun to complicate EU regulation. The bloc’s AI Act, which passed the European Parliament in March and is likely to become law this summer, has a heavy focus on regulating potential harms and privacy concerns around the technology. However, some member states, notably France, made clear during negotiations over the law that they fear regulation could shackle their emerging AI companies, which they hope will become European alternatives to OpenAI.

Speaking before last November’s UK summit on AI safety, French finance minister Bruno Le Maire said that Europe needed to “innovate before it regulates” and that the continent needed “European actors mastering AI.” The AI Act’s final text includes a commitment to making the EU “a leader in the uptake of trustworthy AI.”

“The Italians and the Germans and the French at the last minute thought: ‘Well, we need to cut European companies some slack on foundation models,’” Mügge says. “That is wrapped up in this idea that Europe needs European AI. Since then, I feel that people have realized that this is a little bit more difficult than they would like.”

Sarlin, who has been on a tour of European capitals recently, including meeting with policymakers in Brussels, says that Europe does have some of the elements it needs to compete. To be a player in AI, you have to have data, computing power, talent, and capital, he says.

Data is fairly widely available, Sarlin adds, and Europe has AI talent, although it sometimes struggles to retain it.

To marshal more computing power, the EU is investing in high-performance computing resources, building a pan-European network of high-performance computing facilities, and offering startups access to supercomputers via its “AI Factories” initiative.

Accessing the capital needed to build big AI projects and companies is also challenging, with a wide gulf between the US and everyone else. According to Stanford University’s AI Index report, private investment in US AI companies topped $67 billion in 2023, more than 35 times the amount invested in Germany or France. Research from Accel Partners shows that in 2023, the seven largest private investment rounds by US generative AI companies totaled $14 billion. The top seven in Europe totaled less than $1 billion.

Sarlin sees an even bigger gap looming for European companies once they have an AI product ready to launch. “We have a very good basis in Europe in terms of data, compute, and talent,” he says. “But that’s not enough, you need distribution. And that happens through software products, usually software platforms.”

OpenAI’s partnership with Microsoft gives its multiple distribution points, through code hosting platform GitHub, Copilot assistants built into productivity tools, Windows, and Bing. It has a new deal with Apple that provides still more ways for people to interact with its AI services. “Every single interaction gives you a better model, gives you a better product, which affects more users, which again, gives you a better model,” Sarlin says.

It’s here that the limits of Europe’s sovereignty in AI start to show. French government officials, up to and including President Emmanuel Macron, have praised the Paris-based startup Mistral AI in interviews and speeches, and hailed it as a domestic alternative to American companies. Macron, who hosted Mistral’s CEO for dinner at the Élysée Palace, has called the startup an example of “French genius.” The company was founded in April 2023 and is already nominally valued at around €5 billion ($5.4 billion). But in February, Mistral announced a partnership with Microsoft. The US company invested €16 million into the startup and will make its models available to its own customers. Mistral will train its large language models on Microsoft servers. The response in Brussels was one of dismay, as it seemed to undermine the arguments that had been made—including by Mistral itself—that avoiding stronger regulation of foundation models would allow European AI champions to flourish. Competition authorities are already investigating the deal.

Carnegie’s Csernatoni says deals like Mistral’s with Microsoft demonstrate that the kind of “pure” AI sovereignty imagined by many in Europe simply isn’t practical. “In the negotiations on the AI Act, Mistral AI was very much championed as a European technological sovereignty champion, but, of course, the discussions were much more nuanced than that, because you cannot have autarky,” she says. “You need to balance collaborations and partnerships with [companies like] Microsoft to be profitable.”

The EU simply doesn’t have the platform companies with the scale and reach of a Microsoft or a Google. To compete in AI would mean going back a couple of generations to solve an older problem. “Marc Andreessen said in 2011, software's eating the world, software's eating traditional industries … AI is accelerating that,” Sarlin says. “Unless we ensure that we have these Big Tech, sizable software product companies in Europe, we're not going to be able to create value in Europe with AI.”