AI Governance & Access

On the evening of Friday, June 12, 2026, Anthropic received an export-control directive from the U.S. government ordering it to suspend all access to two of its newest models — Claude Fable 5 and Claude Mythos 5 — for any foreign national, whether inside or outside the United States, and including the company's own foreign-national employees.

Because Anthropic could not cleanly separate foreign-national access from everyone else's on short notice, it complied by disabling Fable 5 and Mythos 5 for all customers, globally. The two models had only been released days earlier — Fable 5 to the public and Mythos 5 to partners — on June 9. Every other Anthropic model stayed online.

The directive itself did not spell out the national-security concern. Anthropic's stated understanding is that the government had become aware of a "jailbreak" path on Fable 5; the company says it reviewed the evidence and characterizes it as a narrow, non-universal technique — essentially asking the model to read a codebase and fix software flaws — whose capability is already widely available from other models. Reporting (Semafor, June 14) indicated the order was driven in part by suspicion that a China-linked group had accessed the new model.

This is, as far as we can tell, the first time the U.S. government has reached in and switched off a frontier model in active deployment. It is a different lever than the familiar chip and weights export controls: it targets *who is allowed to query a running model*, not what hardware ships abroad. For a benchmark site, that distinction matters — model availability is now a variable a government can change overnight, independent of price, quality, or the vendor's own roadmap.