Anthropic has been forced to abruptly disable its most advanced AI models, Fable 5 and Mythos 5, after the US government issued an export‑control order barring foreign nationals from accessing them on national security grounds. Because Anthropic cannot practically separate foreign and domestic users at the model level, the company says it has had to switch the models off for everyone, including large corporate customers and its own foreign staff. Access to earlier Claude‑family models remains unaffected, but enterprises that had begun testing or integrating Fable 5 now face an unplanned rollback.
For CIOs and chief risk officers, the incident is a live case study in concentrated vendor and regulatory risk around frontier models. An upstream policy decision in Washington has instantly invalidated assumptions about availability, data residency and long‑term support that many buyers had written into internal roadmaps and board decks. It also follows months of political friction between Anthropic and the Trump administration, including moves to brand the company a “supply‑chain risk” for federal agencies before a court partially suspended those sanctions.
The message for enterprises is blunt: foundation models are now subject to export controls and security classifications more akin to sensitive dual‑use technologies than ordinary SaaS. Organisations betting heavily on a single frontier vendor will need contingency plans, multi‑model architectures and explicit clauses covering government intervention, data retention and model substitution. The Anthropic order is unlikely to be the last time a regulator reaches deep into an AI vendor’s stack – and next time it may hit production, not just pilots.
