The U.S. order, motivated by a risk to national security and export controls, imposes a cutoff for any foreign national.
Anthropic announced on Friday June 12 that it had deactivated Mythos 5 and Fable 5 following a request from the Trump administration citing a risk to national security and export controls. The order focused on cutting off access for any foreign national, inside or outside the United States, including foreign employees of the American start-up.
Fable 5 is the consumer version of Mythos 5, restricted in terms of cybersecurity, while Mythos 5 is only accessible to a restricted network of digital infrastructure companies and countries. Washington justifies this measure by the discovery of a circumvention of certain Fable 5 safeguards and bases its action on export controls. According to La Tribune, the injunction comes from the US Department of Commerce and orders the withdrawal of the two models; the directive initially only affected foreign users, but Anthropic, unable to distinguish between nationalities, suspended access entirely.
Reactions in France and Europe
Gabriel Attal, secretary general of Renaissance, judged that “the AI war has already begun”: “We cannot count on others, because that makes us vulnerable. The United States’ decision shows this. Anthropic is their Strait of Hormuz.” On
Jean-Luc Mélenchon (La France insoumise) believes that the measure “proves the urgency of being independent and sovereign” and describes it as a “political settling of scores” between the White House and Anthropic. LFI calls for “the mobilization of national savings” for strategic digital infrastructures and “the opening of negotiations at the UN” to regulate AI.
Edouard Philippe (Horizons) warns: “We have mastered neither the models nor the calculation capabilities of AI, a technology as essential as electricity or the Internet.” Europe must “wake up”, favor European solutions and simplify “standards which benefit American Big Tech”.
Bruno Retailleau (Les Républicains) wishes to “reorient public procurement towards sovereign solutions” and pleads for “doing with AI what we did with nuclear power: thinking of this technology as part of our sovereignty”. He told Dario Amodei: “You are welcome in Paris”. Olivier Faure (Socialist Party) stressed that “we have entered a new world, where we can only count on ourselves first” and called for “building a real European power”. Thomas Regnier, spokesperson for the European Commission, believes that the blockage reinforces “Europe’s need for technological sovereignty”.
In Europe, the decision is seen as the materialization of the risk that the United States will cut off access to major digital services by using tech and AI giants as levers of geopolitical pressure. Donald Trump has repeatedly threatened the European Union with retaliation in the event of sanctions against American companies: at the end of August 2025, he spoke on X of “export restrictions on highly protected American technologies and electronic chips”. Washington had already restricted the sale of Nvidia’s most advanced AI chips to China.
European politicians emphasize their dependence on models and calculation capacities, the risk of extraterritoriality via export controls and call for public procurement in favor of sovereign solutions and real European power.
Regulatory U-turn and Anthropic’s position
After his election, Donald Trump first canceled the regulatory framework established by the Biden administration, which provided for control before the release of the models. But in early June, two months after Mythos 5 went online in April, he signed a decree inviting companies to submit, on a voluntary basis, their most advanced models for government review up to thirty days before their release.
Since its creation in 2021, Anthropic has been warning of the potentially catastrophic dangers of AI (loss of control, creation of biological weapons, cybersecurity). Dario Amodei has called for reinstating export controls on the most advanced chips and AI and granting the government the right to block a model if security measures are breached.
Anthropic considers the measure “disproportionate”, disputes that the discovery of a potential circumvention justifies “the recall of a commercial model installed among hundreds of millions of people” and denounces a “misunderstanding”. The company regrets that the decree does not detail its reasons and calls for a procedure that is “transparent, fair, clear and based on technical facts”.
Criticisms in the ecosystem
Anthropic’s restricted launch policy and its calls for regulation were described as “fear marketing” by Arthur Mensch (Mistral AI) and seen as a possible constraint for small players and open source models by Yann Le Cun, who quipped: “We reap what we sow”.
