The Anthropic Fable ban shows why Europe cannot keep relying on American tech
The American ban on foreign access to Anthropic’s Fable 5 and Mythos 5 models should be treated as a warning.
This is not just a story about one AI company. It is a story about power. It shows, in plain terms, what happens when European businesses, governments, universities and public services build their digital lives on technology that can be controlled elsewhere. Anthropic said the United States government ordered it to suspend access to Fable 5 and Mythos 5 for foreign nationals. That included people outside America and even foreign national employees inside Anthropic. The company then disabled the models for all customers to comply. That is the point. Access to critical technology was not decided in London, Brussels, Berlin, Paris or Dublin. It was decided in Washington. For years, Europe has treated American technology as the default. Cloud storage, office software, email, analytics, AI models, chips, app stores, advertising systems and developer tools are often bought from the same handful of American companies. The products are good. In many cases they are the best. The problem is not quality. The problem is dependency. When a foreign government can affect access to tools used by European organisations, Europe has a sovereignty problem. When sensitive data sits under the control of companies subject to foreign law, Europe has a sovereignty problem. When public money flows out of Europe to rent the same platforms every month, instead of growing local capability, Europe has an industrial problem. The Anthropic case makes the risk visible. A model can be removed. A service can be restricted. Terms can change. Prices can rise. Access can become political. A company can promise stability, but if it is governed by another state, that promise has limits. This matters even more in AI because AI is becoming part of ordinary work. It is being added to legal review, public services, education, security, healthcare, software development and compliance. If Europe relies almost entirely on American models and American cloud providers, then Europe is not shaping the next technological era. It is renting it. The EU has started to recognise this. Its recent technological sovereignty package is aimed at semiconductors, AI, cloud and open source. The Commission has also pushed the idea that critical public contracts should favour technology controlled in Europe, especially in sensitive sectors such as defence, healthcare, banking and energy. The phrase used by supporters of the movement says it clearly: build European, buy European, protect European. That should not be dismissed as protectionism. It is basic resilience. No serious person is saying that Europe should cut itself off from American technology. That would be foolish. American companies have built extraordinary products and Europe should work with them where it makes sense. But there is a difference between cooperation and dependency. At the moment, Europe is far too close to dependency. The UK has the same issue. Britain has produced excellent research, strong universities and world class talent, but too often the economic value ends up elsewhere. DeepMind began in the UK and became Google DeepMind. Arm began in Cambridge and is no longer British owned. Too many promising firms sell early, scale abroad or become dependent on foreign infrastructure. The result is familiar. Britain invents, America scales, and Britain buys the finished product back. That pattern cannot continue forever. There are already European and British alternatives worth backing. In AI, Mistral is the clearest European contender. Its argument is simple: customers should be able to run models with more control over data, deployment and infrastructure. OVHcloud is now planning frontier AI models of its own, building on Europe’s cloud base and supercomputing capacity. In cloud, Europe has OVHcloud, Scaleway, StackIT, IONOS, Hetzner, Clever Cloud, Exoscale and others. In privacy and collaboration, there are firms such as Proton, Nextcloud and Element. In the UK, there are serious companies in AI, cyber security, chips, robotics and autonomous systems. Are all of these products better than the American options today? No. That is not the test. The test is whether they are good enough to adopt in more places, improve with real customers and become stronger through demand. Technology ecosystems are not built by applause. They are built by procurement. Governments, councils, universities, hospitals, charities and large companies decide the future every time they sign a contract. If every contract goes to the American default, European competitors will remain small. If European competitors remain small, buyers will say they are not mature enough. It becomes a loop. The way out is not blind loyalty. It is deliberate buying. Public bodies should be required to ask whether a European or UK provider can meet the need before defaulting to a foreign giant. Sensitive workloads should have a higher bar for foreign control. Cloud and AI contracts should measure sovereignty, not just price and features. Buyers should ask who controls the service, where the data sits, which law applies, who can switch it off and whether there is a practical exit plan. For private companies, the same logic applies. A small business does not need to move everything overnight. It can start with lower risk categories. Analytics, email, file storage, backup, monitoring, documentation and internal knowledge tools are all areas where European options already exist. More sensitive organisations should go further, especially those handling health data, legal data, financial data, children’s data, national infrastructure or public sector work. The UK also needs to take its own sovereign AI plans seriously. A £500 million Sovereign AI Unit and new investment in compute are positive steps, but they will only matter if they lead to real companies, real customers and real domestic capability. Funding alone is not enough. British firms need procurement routes, compute access, patient capital and customers willing to buy from them before they are globally dominant. Europe and the UK do not need to copy Silicon Valley in every detail. They need to build the parts of the stack that matter for freedom of action. That means cloud, AI models, chips, identity, security, payments, productivity software, communication tools and open source foundations. Some of it can be shared across Europe. Some of it can be built by the UK. Some of it can be done through partnerships. But the direction has to be clear. The Anthropic Fable ban is a useful shock because it strips away the illusion that technology is neutral. Technology is owned. It is governed. It sits inside legal systems. It can be used as leverage. If Europe wants control over its hospitals, schools, courts, councils, businesses and public services, it cannot leave the digital foundations of those institutions entirely in foreign hands. The answer is not panic. It is procurement, investment and discipline. Build European. Buy European. Protect European.