Europe Debates AI Sovereignty After Anthropic Cutoff
A US order that switched off Anthropic's top models worldwide now has Europe debating whether to build its own AI models or just secure access.
Evgenii Arsentev · PhDEurope is now openly debating how to cut its dependence on US artificial intelligence after a single Washington order forced Anthropic to switch off its Fable 5 and Mythos 5 models for every non-US citizen worldwide. The European Commission has started assessing the practical fallout, and its spokesperson for technological sovereignty, Thomas Regnier, said such measures must 'not be discriminatory against partners,' calling the situation a 'shared challenge.'
The shutdown has hardened a long-running argument among European researchers into two camps. One says the continent must build its own frontier models. Thorsten Holz of the Max Planck Institute warned that a single foreign government able to switch models off 'overnight for all non-US citizens' is a direct threat to digital sovereignty. Konrad Rieck of TU Berlin argued Europe needs its own capable models precisely because US options can be 'shut off at any time.' Gitta Kutyniok of LMU Munich called for an 'Airbus moment' — pooled European investment in both foundation models and chip design.
The other camp: don't try to win
Not everyone thinks building is realistic. Paul Röttger of the Oxford Internet Institute doubts Europe can catch up and would instead lock in access through contracts and trade-policy leverage. The skeptics have the numbers on their side: Mistral, Europe's best-known lab, has fallen far behind over the past two years, and the continent lacks the large-scale data centers and power generation needed to train frontier models at scale. Researcher Jonas Geiping noted that German power generation 'has dropped back to 1985 levels' — a blunt reminder that AI ambition ultimately runs on electricity, not slogans.
Why it matters beyond Europe
This is the first time a major AI provider has been switched off for an entire class of users by government order, and it quietly sets a template. If you live or run a business outside the US, the practical lesson is that access to a frontier model is a political variable, not a permanent utility. The tool you lean on today can vanish because of a decision made in another country's capital, and you get no vote in it. My own read is that the 'build it ourselves' debate will outlast this particular crisis: even if access is restored, the fear of the off switch won't go away. Talks to restore access are reportedly ongoing, with Anthropic CEO Dario Amodei due to join G7 leaders at a working dinner.
Don't wire your work to a single model or provider. Keep your prompts and important outputs saved somewhere portable, and have a real fallback ready — a second vendor, or an open model like Gemma you can run yourself. Test it once now, while nothing is on fire, so a geopolitical decision can never take your work fully offline.

Author
Evgenii Arsentev
PhD · Chief Product Officer at a healthtech company
Want to actually build this?
Guides explain. The free course transforms — personalized, gamified, and built to get you shipping fast.
◉ Start the free courseSource: the-decoder.com