Europe's AI Safety Paradox: Sidelined from debate while facing inevitable impact

Christine Lagarde, president of the European Central Bank, laid out Europe's predicament in blunt terms this week: the continent can reject artificial intelligence and watch its economy stall, or embrace it and surrender to tools built by American and Chinese companies. There is no third door. Yet even as this choice hardens into something unavoidable, Europe has found itself almost entirely absent from the conversation that might shape what happens next.

The absence is structural. Europe produces regulators, not titans. It has no OpenAI, no Anthropic, no Nvidia—no company large enough to sit at the table where the world's most powerful AI labs decide what risks matter and how fast to move. The continent's strength in regulation, which has already irritated the White House, counts for almost nothing in a debate dominated by American and Chinese voices. When Donald Trump and Beijing dismiss the idea of slowing down AI development, Europe's input becomes almost ceremonial. Ursula von der Leyen, president of the European Commission, said this week she would invite the major AI labs to discuss pacing. The statement landed with barely a ripple.

Lagarde's answer to this powerlessness is self-sufficiency. If Europe builds its own AI technology and invests in the datacentres that form the backbone of these systems, it can break the threat of being cut off by Washington or Beijing. The logic is straightforward: dependency dissolves when you control your own tools. Margrethe Vestager, the former EU competition commissioner who has spent years clashing with American tech companies, co-authored a report this week that echoed the same warning. Europe's citizens are already bearing the costs of decisions made elsewhere, she wrote. Building domestic AI capacity is not a luxury—it is a matter of sovereignty.

Vestager points to the EU AI Act as proof that Europe has something to contribute to the safety conversation. The law requires developers of high-risk AI systems to report safety incidents and address the possibility of a powerful model slipping out of control. It is rare legislation, genuinely novel in its scope. But she acknowledges a hard truth: if an existential threat exists, it demands a global response, not a European one. That response cannot be left to the United States and China alone. She calls for the UN General Assembly to move beyond the rivalry between Xi and Trump and create space for the rest of humanity to have a voice.

Yet the framing of the crisis itself may be working against Europe. Frederike Kaltheuner, an adviser at the AI Now Institute, argues that American tech executives have shaped the entire debate around a narrow idea: that there is a frontier of AI capability that matters above all else, and that access to it determines a nation's future. This narrative serves the companies that control the frontier. The reality is messier. Many businesses are not adopting the most advanced AI models—the ones that generate the most alarm—because they are expensive and because companies fear what might happen to their data. If that trend holds, Europe may not be as badly positioned as the doomsday talk suggests.

The EU AI Act, despite its flaws, addresses something concrete: how ordinary people encounter AI in their daily lives. It sets rules for labelling AI-generated content, for using AI in medicine and hiring decisions. These are not frontier concerns. They are the concerns of a functioning society. But the act is unlikely to become a global standard. Georgina Kon, a partner at the law firm Linklaters, notes that other countries have deliberately chosen to differentiate their approach from the EU's, citing the heavy compliance burden and the act's narrow territorial focus. Tech companies will still want to sell to Europe, but they will not feel constrained by European rules.

Kaltheuner worries that the doomsday narratives will push Europe in exactly the wrong direction. Fear of being left behind could drive the continent to adopt American AI models across its economy, across its public sector, across its schools and hospitals. That would be a different kind of crisis—not a technological one, but a sovereignty one. Europe would wake up to find its future running on infrastructure controlled by a handful of American companies. The choice Lagarde described—embrace dependency or build your own—may be false. There is a third option: Europe could reject the framing altogether and ask what it actually needs AI to do, rather than what Silicon Valley says it must have.