The United States has lifted export controls on Anthropic Fable 5 and Mythos 5, clearing the company’s most powerful AI models to return worldwide after a 19-day shutdown that pulled them offline for users everywhere.
The US Department of Commerce removed the restrictions on Tuesday, 30 June 2026, and Anthropic began switching the models back on the following day, as reported by Al Jazeera.
For anyone who lost access mid-project, the reboot ends nearly three weeks of genuinely frustrating downtime.
Why Anthropic pulled Fable 5 and Mythos 5 offline
This whole saga kicked off on Friday, 12 June 2026, when Anthropic abruptly cut off both models. The move was not a bug or a server outage. The Trump administration had ordered the company to block all foreign nationals, including its own employees, from touching Fable 5 and Mythos 5, and Anthropic complied by shutting the door completely.
The trigger was a security scare rather than a policy whim. Amazon researchers found a way to bypass Fable 5’s safety guardrails, a technique the industry calls a jailbreak, and coaxed the model into identifying software vulnerabilities.
In one case it went further and produced code showing how a particular flaw could actually be exploited.
What the reversal means for users
The episode is the clearest sign yet that the most capable AI models are now treated as strategic assets, governed less like ordinary software and more like advanced technology subject to national security review.
Anthropic, the San Francisco lab behind the Claude family of models, has long sold safety as its edge, which made a government-ordered blackout an awkward test of that promise.
With the controls gone, Fable 5 started rolling out globally on Wednesday, 1 July 2026 across Claude Platform, Claude.ai, Claude Code and Claude Cowork. Cloud access through Amazon Web Services, Google Cloud and Microsoft Foundry is being switched back on as fast as the plumbing allows, so expect a staggered return rather than one clean flip of a switch.
Mythos 5 is a different story altogether. It is coming back only for approved organisations inside the US, and only after a government review signs off on each one individually.
That split tells you how nervous regulators still are about the most capable frontier models ending up in the wrong hands.
What happens next
Anthropic now has to prove the jailbreak that started all this has been properly patched, while regulators watch how the re-released models behave out in the wild.
Expect tighter safety reporting and slower, more conditional rollouts to become the template every frontier lab follows the next time Washington gets spooked by a security finding.
