The timeline of Claude Fable 5's return
Claude Fable 5: from shutdown to return (2026)
From the export controls to their removal and the global relaunch. Source: Anthropic official news.
Claude Fable 5 is Anthropic's highest-performing AI model. Shortly after its launch, it was taken offline worldwide by the June 12 export controls and stayed dark for about two and a half weeks. The news here is that those controls have finally been removed and the model is back. For the full story of the shutdown, see our guide to Claude Fable 5 going offline.
From the June 12 shutdown to the June 30 removal
The shutdown began when the US government applied export controls to Fable 5 and Mythos 5 on June 12. On June 30, those controls were lifted, and the models were cleared to return.
"On Friday, June 12, the US government applied export controls to our newest models, Claude Fable 5 and Claude Mythos 5." / "As of today, June 30, the export controls on Fable 5 and Mythos 5 have been lifted." — from Redeploying Fable 5
The directive called for suspending access by foreign nationals, but because Anthropic could not verify nationality in real time, it disabled the models for all customers, including US nationals, to ensure compliance. In other words, it was not a case of "some users can't use it" — nobody in the world could use it during the shutdown.
The global return on July 1 (four products)
With the controls lifted, Anthropic announced it would bring Fable 5 back to users worldwide the next day. The return covers four products — the Claude Platform, Claude.ai, Claude Code, and Claude Cowork — all available to users globally.
"Fable 5 will be available starting tomorrow, Wednesday, July 1, to users globally on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork." — from Redeploying Fable 5
The cause of the shutdown and the strengthened safeguards
What triggered the shutdown, and what was strengthened for the return
The immediate trigger for the shutdown was an external vulnerability report. Let's look at why Fable 5 was seen as risky, and how Anthropic responded.
The trigger: a bypass found by Amazon researchers
The government acted on a report that Amazon researchers had found a way to bypass Fable 5's safeguards. The technique, as described, prompted the model to identify a number of software vulnerabilities.
"The export control directive on June 12 came after the government became aware of a report in which Amazon researchers had found a method of bypassing Fable 5's safeguards." / "prompting it so that it identified a number of software vulnerabilities" — from Redeploying Fable 5
The ability to "find and fix" vulnerabilities helps defenders but can also serve attackers. Because of this dual-use nature, a model that can pinpoint weaknesses at high accuracy tends to draw scrutiny on national-security grounds.
A new classifier blocks the technique in over 99% of cases
As a condition of the return, Anthropic added to its safeguards. A newly deployed classifier now blocks the specific reported technique in over 99% of cases.
"The new classifier means that the specific technique described in the Amazon report is blocked in over 99% of cases." — from Redeploying Fable 5
During this period, Anthropic says it worked closely with the government and partners such as Amazon over about two weeks to review the report and the evidence. Beyond closing the technical gap, the return was built on aligning with the authorities to reach agreement on reopening.
What the return means, and how to prepare
Claude Fable 5 return status (as of July 1, 2026)
This return means more than just "back to normal." It touches the relationship between government and AI companies, and how users should think about risk management.
Aligned with the Executive Order
Anthropic frames the whole response as aligned with the June 2 Executive Order, "Promoting Advanced Artificial Intelligence Innovation and Security." It also states that for models that materially advance the capability frontier in areas relevant to national security, it will give designated government partners expanded early access.
"Anthropic has worked closely with the US government as it developed the approach reflected in the June 2 Executive Order on Promoting Advanced Artificial Intelligence Innovation and Security." / "For models that materially advance the capability frontier in areas relevant to national security, we will provide designated government partners with expanded early access." — from Redeploying Fable 5
It signals an era in which the release of frontier models advances together with national-security review, not on performance alone.
Prepare for shutdowns with a multi-model setup
The Fable 5 episode is a reminder that even a top-performing model can suddenly become unavailable due to regulation or national-security decisions. It is back now, but the same thing could happen to another model. Keeping key tasks runnable on two or more models means your work keeps moving even if one is halted.
For coding help, for example, it's worth keeping prompts that work on both the models covered in our guide to Claude and those in our guide to ChatGPT's GPT-5. Options that automatically switch between models internally, such as the system in our guide to Sakana Fugu, are another way to reduce dependence.
When feeding long reference material to an AI, converting web pages to Markdown first preserves the heading and table structure and tends to improve accuracy.



