Chinese AI is a risk for Europe. So is shunning it

Especially now that America is becoming a less reliable partner

Illustration: Simon Bailly

Listen to this story

Your browser does not support the <audio> element.

O N JANUARY 20 TH 2025 DeepSeek was an obscure hedge-fund-turned-tech-startup from Hangzhou. Within a week it had become the byword for a new wave of Chinese innovation, after launching an artificial-intelligence model as capable as Silicon Valley’s bleeding edge but much cheaper to build and run. Having slugged it out in China’s cut-throat domestic market over the past year, DeepSeek and its homespun rivals are looking abroad for profits. They will not find the largest ones in America, increasingly out of geopolitical bounds, or the poorer global south. That leaves Europe as the likely recipient of their attention.

To the old continent, new technology from China may seem like a curse. Chinese electric vehicles are already eating German and French carmakers’ Wurst and frites. Several EU countries have tried to restrict access to DeepSeek’s chatbot over fears that it might shunt data from European companies and citizens to China. No one wants to rely on a geopolitical adversary for what is fast becoming critical infrastructure. These worries are legitimate. But in the case of AI, China may, if embraced wisely, be a blessing for Europe.

There are three reasons why European firms should welcome this Chinese onslaught. First, Chinese models are nearly as good as the best that Open AI, Anthropic and Google can offer—which for most users is good enough. Demis Hassabis, Google’s AI supremo, has said that Chinese AI s are only “a matter of months” behind American ones. Like DeepSeek, most cost nothing to access and relatively little to operate.

This cost advantage comes from their openness—the second reason why they ought to appeal to European firms. In contrast to proprietary black boxes peddled by leading American firms, open models can easily be fine-tuned and run on local infrastructure. Using them averts the risk of being locked in to any one provider. If Open AI or Anthropic went belly-up, their customers would be in a bind. If DeepSeek were to fold, users could keep running its models’ “weights”, the parameters learned during training, on their own data and their own servers—which also allays data-theft fears. American firms like Meta also offer open models. But China is leading the way.

There is one last reason why welcoming Chinese AI is in Europeans’ interest: it offers insurance against lock-out, as well as lock-in. Before Donald Trump took the oath of office for the second time, also a year ago, it would have been absurd to worry about European access to American technology. As he recklessly exploits the transatlantic alliance over Greenland, an executive order limiting American AI firms’ business in Europe no longer seems unthinkable. Some European restrictions on American technology, including the computing clouds where AI s reside, are also plausible.

Although, in a fragmenting world, Europe’s best option may be to nurture its own AI industry, it is not about to become a model-building superpower. But it can still be a world leader in putting the technology to work. Already, 37% of EU businesses report using generative AI, on par with America. In manufacturing, European firms are ahead. Using open models, including from China, could further increase their lead.

Open road

Politicians in Europe appear to grasp this. With lots of interest and no voluble AI incumbents begging for protection, the early efforts to ban DeepSeek mostly fizzled out. In January the European Commission launched an effort to identify and remove barriers holding back open models. None of this will guarantee Europe’s techno-independence. Firms will still rely on American hardware, especially chips from Nvidia. Chinese software comes with all the old risks. But for Europe, the bigger one now is to spurn it. ■

---

논증 분석

유형: prescription

핵심 주장

유럽 기업들은 중국 AI의 위험성을 인식하면서도, 이를 현명하게 수용함으로써 경쟁력 강화와 기술 독립성 확보라는 이중적 이익을 얻을 수 있다.

논리구조

1. 전제: DeepSeek을 필두로 한 중국 AI 모델들이 Silicon Valley 수준의 성능을 훨씬 저렴한 비용으로 제공하며, 미국 시장 진출이 막힌 상황에서 유럽을 주요 타깃으로 삼고 있다.
2. 진단: 유럽은 중국 AI에 대해 데이터 주권 침해 및 지정학적 적대국에 대한 핵심 인프라 의존이라는 정당한 우려를 가지고 있으며, 일부 EU 국가들은 DeepSeek 접근을 제한하려 시도했다.
3. 논거: 첫째, 중국 AI 모델은 OpenAI, Anthropic, Google의 최신 모델과 거의 동등한 수준이며(Demis Hassabis는 ‘불과 몇 달’ 차이라고 평가), 대부분 무료로 접근 가능하고 운영 비용도 낮아 경쟁력이 충분하다.
4. 논거: 둘째, 중국 AI 모델의 개방성(오픈소스)은 미국 독점 기업의 블랙박스 모델과 달리 로컬 인프라에서 파인튜닝·운영이 가능하여 특정 공급자에 대한 종속(lock-in) 위험을 방지하고 데이터 탈취 우려도 완화한다.
5. 논거: 셋째, Donald Trump 행정부의 대서양 동맹 훼손 행태를 감안할 때, 미국이 유럽에 대한 AI 기술 접근을 제한할 가능성이 더 이상 비현실적이지 않으므로, 중국 AI는 미국 기술 차단(lock-out)에 대한 보험 역할을 한다.
6. 처방: 유럽은 독자적인 AI 모델 개발 강국이 되기 어렵지만, 오픈 모델(중국 모델 포함)을 적극 활용하여 AI 기술 응용 분야의 세계 선두 자리를 유지·강화해야 한다.
7. 반론: 중국 AI 수용에도 여전히 위험은 존재한다—유럽 기업들은 여전히 Nvidia 칩 등 미국 하드웨어에 의존하며, 중국 소프트웨어의 기존 보안 위험도 사라지지 않는다.
8. 결론: European Commission이 오픈 모델의 장벽 제거에 나선 현 흐름처럼, 유럽에게 중국 AI를 거부하는 것이 수용하는 것보다 더 큰 위험이다.

결론

중국 AI의 위험성은 실재하지만, 세계가 분열되는 현재 지정학적 환경에서 유럽이 중국 AI를 외면하는 것은 수용하는 것보다 더 큰 전략적 실수이다.

Subscribers to The Economist can sign up to our Opinion newsletter, which brings together the best of our leaders, columns, guest essays and reader correspondence.

Explore more

→See the latest from topics you follow