“AI will determine the geopolitical future.” The warning from Monika Schnitzer, the German economist and chair of the country’s Council of Economic Experts, goes far beyond the usual debate over whether Europe is falling behind Silicon Valley. In an interview with DIE ZEIT, Schnitzer argues that Europe’s technological dependence could ultimately become a question of economic power, national security and political independence.
Schnitzer is among twelve international AI experts who have called for Europe to become genuinely competitive in artificial intelligence. Their 200-page strategy plan argues that the continent must confront a widening technological gap. Among the experts are former EU Commissioner Margrethe Vestager of Denmark, Canadian AI researcher Yoshua Bengio and Schnitzer herself, a 65-year-old professor of comparative economic research at Ludwig-Maximilians-Universität Munich.
The starting point is uncomfortable. “What is certain is that Europe is lagging behind in the most powerful AI systems, the frontier models,” Schnitzer tells DIE ZEIT. Closing that gap, she says, will require enormous effort and substantial political determination. The challenge is not simply to produce another European technology product. It is to ensure that Europe does not become dependent on systems whose development, ownership and access are controlled elsewhere.
That distinction lies at the heart of the strategy. Schnitzer argues that Europe needs access to the most powerful AI models because they will be important for both its economy and its security. The question is whether every one of those systems must be developed in Europe. European countries could use models developed in the United States, but such dependence carries a strategic risk. “Our concern is that Europe could then make itself dependent,” she says.
Europe nevertheless possesses assets that could give it leverage. Companies such as ASML and Carl Zeiss occupy crucial positions in chip manufacturing, while advanced chips are essential for developing advanced AI models. The strategy therefore proposes using European technological strengths to create a more balanced relationship with the United States.
The proposal includes a strategic alliance involving countries such as Germany, France and the Netherlands. Schnitzer argues that these countries should use their strength in specialised areas of semiconductor production as part of a wider European strategy. Other countries should be involved, she says, but some governments need to move first. Her proposal is an “alliance of the willing”, designed to avoid a situation in which a single EU member state can prevent collective action by exercising its veto.
The logic becomes particularly stark when technology is treated as a potential instrument of political pressure. Schnitzer points to the recent decision by the US government to withhold Claude Mythos, described as a presumably highly capable AI model. For her, the case demonstrates how quickly access to advanced technology can become uncertain.
“Europe must be better prepared for such scenarios so that it can act in a coordinated way towards the United States,” she says. If a government such as the Netherlands, where ASML is based, attempted to restrict access to its products, it could itself come under substantial pressure. “Alone, there is nothing Europe can do against the United States,” Schnitzer says.
That raises an even more difficult question. If Europe wants genuine technological independence, should it not develop its own frontier models rather than merely attempting to secure access to American or Chinese ones?
Schnitzer’s answer is unusually candid. “In principle, it would be best to take both approaches very seriously.” Europe could exploit its existing technological advantages while investing in frontier models of its own. But she questions whether European governments are actually prepared to provide the necessary resources. “If you invest only half-heartedly, you can waste a great deal of money and time,” she warns.
Her position is therefore not that Europe should abandon the ambition of developing its own frontier model. On the contrary, she argues that an independent European system would have significant advantages. But the costs and risks must be acknowledged. “Even with all the money in the world, there would still be no guarantee of success,” she says. The political choice, therefore, should be made with a clear understanding of what such an undertaking would require.
Europe may nevertheless have opportunities beyond competing directly with the largest AI companies. Schnitzer identifies security as one area where Europe could establish a strong position. Advanced AI systems could be examined for dangerous capabilities and security risks. Britain already has an AI Security Institute, while Germany is expected to establish a similar institution. The strategy also proposes strengthening expertise within European institutions through mechanisms such as a task force for transformative AI.
The experts’ strategy places another resource at the centre of the competition: computing power. By 2030, they want Europe to account for 15 per cent of global AI computing capacity. Schnitzer argues that the figure is significant because it is roughly equivalent to Europe’s current share of global economic output. Computing power, she says, could become a strategic resource comparable to energy.
The crucial requirement is that this capacity should physically exist in Europe. Even if American AI models operate on European infrastructure, having the data centres on European territory would provide greater control over what happens there.
Finding the necessary sites could be relatively straightforward, Schnitzer argues. Germany could use disused industrial land where electricity connections are already available. Spain has substantial renewable electricity and could power data centres with solar energy. Norway, with its hydropower resources, could also offer favourable locations.
But building the infrastructure creates another political challenge: public consent. Data centres can be unpopular with local communities, and Schnitzer argues that Europe must handle this better than the United States. Her proposal is that data centres should be built “in a European way”, meaning that the populations hosting them should receive tangible benefits.
She points to wind turbines in Münsterland as an example. Local people were able to purchase shares in wind projects and benefit financially from the electricity they generated. A similar model could be applied to AI infrastructure. Companies building data centres could contribute to swimming pools or transport infrastructure, for example. “It must be clear: the municipality gets something in return,” Schnitzer says. She therefore welcomes the German government’s plan to give municipalities hosting data centres a greater share of the resulting tax revenues.
Yet the most consequential argument in the interview concerns the possibility that Europe could become not merely technologically dependent but politically powerless in determining how AI develops. Schnitzer rejects the assumption that leading American AI companies will voluntarily slow the technological race simply because they recognise its dangers. Any restraint, she argues, would require an agreement involving all major players and a mechanism for regularly verifying compliance.
Europe could have a role in creating that verification technology. Schnitzer compares the challenge with an arms-control agreement and argues that Europe possesses the talent and expertise to develop reliable mechanisms for monitoring AI safety commitments.
The stakes are already visible in everyday life. “AI will affect all areas of our lives,” Schnitzer says. People already use ChatGPT as a search engine, yet the answers they receive are shaped by decisions made by OpenAI, including decisions about the data incorporated into its models. AI systems could increasingly penetrate intimate areas of human life, particularly through technologies such as deepfakes. At the same time, AI-driven hacking attacks are becoming increasingly common.
This is why Schnitzer’s warning is ultimately about power rather than technology alone. “We therefore need the best AI models in order to defend ourselves,” she says. If Europe fails to approach the issue with sufficient seriousness and urgency, she warns, it may not have a seat at the table when decisions about AI with global consequences are made.
Her conclusion is blunt: “If Europe does not wake up, it will have no effective control over the effects of the most important technology for humanity’s future on its citizens.”
For Europe, the AI race is therefore becoming something more dangerous than a competition between companies or a contest for technological prestige. It is becoming a test of whether the continent can retain meaningful control over its economy, security and political choices in a world increasingly shaped by artificial intelligence. And, according to Schnitzer, the decisive moment is not somewhere in the distant future. It is now.
