After the US government barred non-US citizens from accessing Anthropic's new Fable 5 and Mythos 5 models, the European Commission is weighing the fallout. European experts call it a wake-up call - but disagree on what to do next.

The European Commission says it's currently assessing the practical impact of the US export control order that forced Anthropic to shut down its Fable 5 and Mythos 5 models worldwide.

Thomas Regnier, the Commission's spokesperson for technological sovereignty, told Euronews that emergency measures must "not be discriminatory against partners." He called it a "shared challenge, not one confined to a single jurisdiction or company." The development is "a ​further ⁠illustration of why Europe needs to strengthen its technological sovereignty," he added, according to Reuters.

Days earlier, Anthropic had pulled its most advanced models worldwide following a US government order tied to national security concerns. Talks to restore access are ongoing. Anthropic CEO Dario Amodei will join the heads of other leading AI companies at a working dinner with G7 leaders on Wednesday.

Researchers agree on the alarm, not the response

In a collection of statements published by the Science Media Center, several European researchers also describe the situation as a wake-up call. They disagree sharply, however, on what Europe should do about it.

Thorsten Holz of the Max Planck Institute for Security and Privacy called it striking that a single foreign government order could "shut down a model overnight for all non-US citizens," including European companies. Digital sovereignty, he argued, doesn't mean self-sufficiency. It means being able to use critical technology even during geopolitical conflicts.

Konrad Rieck of TU Berlin put it more bluntly. US models can be "shut off at any time, sometimes for opaque reasons," he said, and Europe needs to develop and operate its own capable models. In the same vein, Gitta Kutyniok of LMU Munich called for an "Airbus moment" for AI, with joint, ambitious investment in foundation models, chip design, and energy-efficient computing. Anyone who waits until the structures are already in place, she warned, will have almost no room left to shape them.

Paul Röttger of the Oxford Internet Institute took a very different stance. He said he doesn't believe more investment in European AI is the answer: "Europe won't be able to develop models like Mythos or Fable 5 in competition with the US." Instead, Röttger argued, access should be secured through contracts, tied to data center investments and backed by credible trade policy threats.

Two more researchers highlighted just how hard building homegrown capacity would actually be. Matthias Hein of the University of Tübingen warned that Europe needs not just one but several providers of its own, because nobody should count on commercial companies to keep releasing open-weight models. Jonas Geiping of the ELLIS Institute Tübingen pointed to the structural barriers behind that goal. French company Mistral has "fallen far behind" over the past two years, he said. Even if new players emerged, the basics are missing: large-scale data centers and enough power generation, which in Germany has dropped back to 1985 levels.

Geiping also cautioned against drawing historical parallels to nuclear weapons tensions, a comparison Anthropic itself likes to make. Unlike nuclear weapons, he argued, AI is deeply woven into the economy. A shutdown or restriction during a diplomatic conflict wouldn't just affect defense capabilities, he said. It could also deal serious damage to the European economy if processes can't function without strong AI.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now