Anthropic halts Mythos 5 and Fable 5 after US export ban

This is the first time a frontier foundation model has been effectively pulled from the global market by government order rather than a lab’s own risk calculus. The U.S. export directive on Mythos 5 and its public sibling Fable 5 signals that top-tier AI systems are now being treated like dual‑use strategic assets, closer to advanced chips or cryptography than to consumer software. That’s a fundamental shift in how states see the frontier labs.([axios.com](https://www.axios.com/2026/06/12/anthropic-trump-mythos-fable-national-security))

For the race to AGI, the move cuts both ways. In the short term it slows international diffusion of Anthropic’s most capable models, especially for cybersecurity and other sensitive domains where Mythos‑class systems reportedly excel. It also raises the bar for foreign labs: matching Mythos may now require local capability rather than simply buying API access. At the same time, the precedent that Washington can unilaterally choke off a private lab’s flagship model will push companies to build more formal compliance, export‑control, and lobbying muscle into their roadmaps, and may accelerate parallel efforts in Europe and Asia to assert their own sovereignty over frontier compute and models.([boursorama.com](https://www.boursorama.com/actualite-economique/actualites/washington-invoque-la-securite-nationale-et-force-anthropic-a-suspendre-son-ia-la-plus-puissante-d7190e04acf0174aa27ce3f84770af7f?utm_source=openai))

The bigger story is that safety, national security, and industrial policy have fully converged around frontier AI. Developers can no longer assume that once a model is trained and productized it will stay online; state veto power is now real, and likely to be exercised again.