The standoff over Anthropic's two most powerful AI models is starting to thaw. On Friday, the Commerce Department let the company turn Mythos 5 back on for a limited set of US users. Now, according to Axios, Fable 5 could follow as soon as this week, pending a final two sign-offs.
Mythos first
Commerce Secretary Howard Lutnick cleared Mythos 5, which Anthropic calls its strongest cybersecurity model, to be redeployed to US organizations that operate and defend critical infrastructure, a group reported at roughly 100 institutions. The model has built-in guardrails meant to deter use in cyberattacks or bioterrorism, and Anthropic notes it has never been freely available. The letter, first reported by Semafor, said Anthropic had "worked with the U.S. government to address risks" tied to both models.
Fable could be next
Fable 5 is the bigger prize for ordinary users, and it is close. Axios reports that most agencies have agreed it can safely return; only the Pentagon and the National Security Agency still have to sign off. Insiders expect the limits could lift as soon as this week, though nothing is final until those two approvals land.
The terms are still fuzzy. Anthropic had offered Fable 5 at no extra cost on several paid Claude plans through June 22, and it isn't clear whether returning subscribers get that free access back or whether Fable comes back behind new fees or identity checks.
What the blackout cost
The shutdown was not abstract. When access vanished on June 12, developers found automated jobs frozen mid-task, and companies scrambled to swap in rivals, including cheaper Chinese models. Anthropic had billed Fable 5 as the most capable model ever released to the public; in early testing it highlighted, the payments company Stripe used it to overhaul a 50-million-line codebase in a single day, a job that would have taken its engineers more than two months by hand.
The fight underneath
The episode has unsettled a lot of founders. "This is how these companies go out of business," wrote one, arguing you cannot fund billion-dollar training runs while begging the government to approve customers one at a time. Garry Tan, who runs Y Combinator, called the model-by-model approach "a solid way to salt the ground and kill all innovation by small startups." Both Anthropic and OpenAI are now pushing the administration to replace the case-by-case reviews with a clear, written process. OpenAI said as much when GPT-5.6 got its own gated release the same week: "We don't believe this kind of government access process should become the long-term default."
This whole saga is finally starting to unwind. The government pulled both of Anthropic's most powerful models in June, and they are coming back one at a time. Mythos 5, the cybersecurity model meant for big institutions, is already back, though only for about 100 organizations. Fable 5, the model regular people had on their Claude plans, looks close behind.
There is a good chance Fable 5 returns soon, but no one knows how good a version it will be. It could come back watered down, with enough new guardrails to make it a shadow of the model people fell for, and depending on where you live, you may not get access at all.
I am rooting for it anyway. Fable 5 was the most exciting model Anthropic had shipped, and I am still eager to get my hands on it and put it through its paces. I will be testing it and sharing the results right here. ## Sources