Part of the reason it hurt was that it happened over the weekend. No months of warning, no press conference. Customers were informed by Anthropic that it had been told to restrict foreign nationals’ access to its Fable 5 and Mythos 5 models due to national security concerns. As a result, European businesses that had discreetly integrated these tools into their everyday operations were suddenly locked out.
According to Anthropic, the official explanation was limited to a technical flaw. However, the timing seemed off. The company has been referred to by President Trump as a “out-of-control Radical Left AI company” in the past, and it’s difficult to separate that history from a decision that directly affected foreign users, including a sizable portion of the Fortune 500 Europe. The message that European executives took away, regardless of the true motivation, was more straightforward and unsettling: a foreign government has the ability to abruptly and without warning turn off the AI systems that run their company.
The current scramble is motivated by that realization. In the words of Marc Warner, chief technology officer of Accenture and head of the U.K. AI company Faculty, “nearly nobody had taken seriously the idea that the U.S. could simply cut off access.” They’re taking it seriously now. Speaking with people in this setting gives the impression that the topic of theoretical risk suddenly changed to operational reality.
The numbers show why this is so important. The European Commission and Parliament claim that over 80% of Europe’s digital infrastructure and products are supplied by foreign companies. With over 300,000 business clients worldwide, Anthropic’s revenue from Europe, the Middle East, and Africa has increased by almost nine times in just one year. It’s not a side relationship. Infrastructure is that.
The lack of alternatives is startling. In this case, Warner’s observation that technology markets tend to produce a limited number of winners seems uncomfortably true: Europe lacks a frontier AI model. UpCloud, a cloud provider based in Helsinki, is positioning itself as a hedge. According to its CEO, Arno Schäfer, the discourse has truly shifted, with digital sovereignty now being viewed as a business continuity issue rather than an abstract political one. It’s possible that he is correct in saying that this incident ultimately influenced decision-makers more than any policy paper.
Thus far, adaptation appears to be more akin to hedging than a clean pivot. European companies are reevaluating contracts that previously assumed permanent access to American AI, diversifying their suppliers, and distributing workloads among providers. According to Reuters, big European businesses are stepping up their efforts to distribute risk among providers; this is a defensive rather than a victorious move. Here, no one is commemorating a breakthrough. They are controlling their exposure.
In terms of politics, the incident has given Europe’s push for sovereignty new life. When you consider how slowly continental tech infrastructure projects typically proceed, Austria’s proposal to host Anthropic inside EU borders seems ambitious. New funding for domestic AI, cloud, and semiconductor development has been announced separately by the European Commission. It’s really unclear if that will result in anything approaching parity with American frontier labs in the coming years. In press releases, these initiatives tend to sound more decisive than they actually are.
Additionally, a more subtle and difficult to measure cultural change is taking place. Similar to how businesses began reconsidering semiconductor supply chains following the chip shortage years ago, executives who previously viewed AI vendor selection as a procurement detail are now treating it as a strategic risk category. Nobody was completely weaned off Taiwan after that earlier episode. It’s also unlikely that this will completely wean Europe off of American AI.
More multi-vendor contracts, more investment in European alternatives, and more political rhetoric about sovereignty that surpasses actual capability appear to be the most likely outcomes of this slow and uneven rebalancing. According to the CUDO Compute study that Fortune cited, almost half of UK businesses were already cautious about relying too much on the cloud. That anxiety was not brought on by Anthropic’s limitations. All they did was give it a name and a date.
