The recent decision by the US administration to restrict access to Anthropic's advanced AI model Mythos for non-US nationals marks a turning point. For the first time, the world has witnessed what Dr. Sergey Lagodinsky, Vice Chair of the Greens/EFA Group in the European Parliament, calls the 'AI kill switch' in action. If the White House chose to, an estimated 88% of the global population could lose access to American frontier models overnight. Meanwhile, China is expected to achieve similar capabilities within a year. Europe, caught between these two emerging AI empires, risks becoming a digital colony.
Mythos, capable of hacking sensitive software worldwide, has transformed AI from a business model into a weapon of mass destruction. The problem is stark: Europe's capabilities are nowhere close. The initiative founded to research and contain the risks associated with Mythos, Project Glasswing, includes no European actors. This is a wake-up call for Brussels.
Scale: The Only Path Forward
First and foremost, Europe needs scale. This is not the race we chose, but it is the race we are in. Compute power and innovation are driven by enormous investments—trillions, not billions. Such equity will not come from public pockets alone; it must be private. The European Commission must convene an emergency summit with leading European businesses to secure substantial and reliable pledges. The big bosses must realize that their own survival and the future of Europe as a business space that can defend itself are at stake.
At the same time, the EU capital market union and the single digital market must be completed in fast-track mode. Scale is not only about money; it is also about size. National solutions will not suffice. We need French-German gigafactories and European space constellations. The common market and common synergies are assets we must leverage. As a recent survey shows, Europeans back digital sovereignty but balk at higher costs—making efficiency and scale even more critical.
Multilateralism Against Digital Imperialism
Second, the Commission must marshal international partners. Pooling AI compute capabilities beyond European borders towards a consortium of middle powers is essential. The EU is the only superpower capable and willing to use multilateralism to confront digital imperialism. The EU's international digital strategy already sets out the framework for digital foreign policy, proclaiming 'global partnerships' as its goal. Now it is time for Europe to lead on these efforts.
In parallel, the Commission should build a new relationship with the US. It is a hard pill to swallow: in the medium term, Europe will depend on American compute and chip infrastructure to grow its own providers at the application layer. Europe has leverage, such as the Dutch company ASML's unique chip printing technology, but this leverage is limited compared to US capabilities. What Europe needs is a path of strategic parallelism: cooperating with Washington where necessary while building capacities and infrastructure that will one day allow it to stand alone.
Regulatory standards, often criticized by deregulation advocates, can help. They anchor American technologies in European trust and transparency. These are the only foundations that can secure Europe in a world of threats and coercion, while also securing American tech access to European markets. The European Parliament's approval of the digital euro is a step in this direction, aiming to curb US payment dominance.
Public Trust: The Missing Ingredient
The United States is experiencing a 'techlash'—growing mistrust towards AI infrastructure among citizens. Europe will only avoid similar reactions if innovations in energy efficiency and water use become integral to transatlantic cooperation. Social redistribution of resources and profits is also necessary so that citizens are not squeezed by AI giants. Only with public acceptance can high-speed investments be politically and economically fruitful. As the Capgemini AI Chief notes, the trust gap between humans and machines remains wide.
These are uncomfortable truths. Rallying capital, herding allies, and bargaining with an unpredictable partner cut against European instincts for caution and consensus. But the American kill switch has already shown that we have entered uncharted AI territory. The only open question is whether Europeans will manage to take their destiny into their own hands and become a geopolitical technology player. The alternative is to wait for someone else to switch us off. That is something we should never accept.
