Anthropic has started turning Claude Mythos 5 back on. The company said on X that the US government cleared it to redeploy the model to a set of American organizations that operate and defend critical infrastructure. Fable 5, the general-use sibling, is still benched.
This is the unwinding of a directive that landed two weeks earlier. On June 12, the government told Anthropic to cut off access to both Mythos 5 and Fable 5 for foreign nationals, and the company responded by pulling both models for everyone rather than trying to sort users by passport.
Who actually gets it back
Not you, probably. The restoration is narrow by design: organizations that run and protect critical infrastructure, the same constituency Mythos 5 was built around in the first place. Anthropic calls it the strongest cybersecurity model in the world, which is the sort of line a company always uses about its newest thing, except in this case a government export control directive treating it as a national security asset lends the claim some weight.
The model first shipped to this crowd through Project Glasswing, a limited program for cyber defenders and software infrastructure providers. Mythos 5 is the same underlying model as Fable 5 with the cybersecurity safeguards lifted, which is exactly why it sets off alarms in Washington. The thing that makes it useful to a bank's security team also makes it useful to whoever is attacking the bank.
The part that's still off
Fable 5 remains unavailable. Anthropic says it is still negotiating with the government to widen Mythos 5 access and bring Fable 5 back for general use, with no date attached. So the company that launched Fable 5 in early June as available everywhere now can't say when ordinary users get it again.
"Mythos 5, our strongest cybersecurity model, can be redeployed," the company wrote, which is a careful way of saying the lab no longer fully controls its own release schedule.
That's the real story here, more than any single model coming or going. The original directive named one specific worry, a method of jailbreaking Fable 5's safeguards for finding software vulnerabilities. Whether that worry has been resolved or simply scoped down to trusted users isn't clear from the announcement, and Anthropic isn't saying.
What changed
For frontier AI, the release decision now runs through the government. A model can be built, benchmarked, priced, and announced as generally available, and still get switched off by directive within days. Then switched partially back on, for a list of approved organizations, by a second directive.
Anthropic says the next steps are expanding the Mythos 5 access list and restoring Fable 5. No timeline was given for either.
