Anthropic’s Fable 5 is back after US shutdown it called 'a misunderstanding'

Anthropic’s Fable 5 is back after US shutdown it called 'a misunderstanding'

Anthropic’s most advanced public Claude model is returning after a US government directive forced it offline - but the episode raises bigger questions about who controls access to frontier AI.

Fable 5 is back after the US government lifted export controls that had forced the company to suspend access to Fable 5 and Mythos 5 earlier in June. The company said it believed the shutdown was based on “a misunderstanding,” after officials raised concerns about a possible jailbreak and national security risk.

In a statement on its website, Anthropic said it is not against the US government having the power to block unsafe AI releases, but argued that this should not have been the case with Fable 5.

“As we have stated publicly, we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles.”

Anthropic is now working to restore access to Fable 5 for its users.

The more interesting question now is not simply whether Fable 5 is available again, but what this episode says about the future of AI model launches. The most powerful AI models may no longer be treated like ordinary software updates. Going forward they may increasingly be treated like strategic technologies that governments can pause, restrict, and negotiate over.

What happened?

On June 9, Anthropic released Fable 5, a restricted version of its Mythos 5. Anthropic said Fable 5 had been released with safeguards designed to prevent misuse in cybersecurity attacks, while the full Mythos 5 was kept under tighter controls because of its more advanced capabilities.

On June 12, Anthropic received a US government export control directive. The directive suspended access to Fable 5 and Mythos 5 for foreign nationals, including Anthropic employees. Anthropic said the practical effect was that it had to disable the models for all customers to ensure compliance.

The issue appears to have centered on whether it was possible to jailbreak Fable 5 and bypass its guardrails. (A jailbreak is essentially a method of persuading an AI model to bypass its safety restrictions). Anthropic pushed back, saying it had not been shown evidence of a broad or universal jailbreak.

Access is now being restored after the US Commerce Department lifted the restrictions. Reuters reports the restrictions were lifted after enhanced safeguards were put in place.

In its response to the US government’s restrictions, Anthropic argued that “perfect jailbreak resistance is not currently possible.” Its case was that if every narrow jailbreak is enough to force a model offline, then no frontier AI model may ever be safe enough to launch. The company warned that applying this standard across the industry could “halt all new model deployments.”

Why this matters beyond Anthropic

Until recently, a new AI model launch mostly meant faster answers, more coding features, or smarter replies. Fable 5’s shutdown shows that frontier models are now powerful enough that governments may step in before, during, or after their launch. That changes the relationship between AI companies, users, developers, and regulators in a way we're going to have to get used to.

OpenAI’s GPT-5.6 rollout has already been limited for similar reasons. Reuters reported on June 26 that OpenAI had delayed a full public launch of GPT-5.6 at the US government’s request, with access limited first to a small group of vetted partners whose details were shared with the authorities.

Fable 5 may be back, but the recent developments have changed the mood around frontier AI model launches. A model can be announced, celebrated, pulled offline, negotiated over, and restored all in the space of a few weeks. Anthropic may call this one “a misunderstanding,” but it also looks like a preview of how the most powerful AI systems may be governed from now on.