Anthropic’s Claude Fable 5 is back. The company is rolling out its previously suspended AI model to users worldwide starting July 1, just one day after US export restrictions on the model were officially lifted.

The turnaround is remarkably fast. Fable 5 was globally suspended on June 12, less than a week after it launched, because export controls made it impossible for Anthropic to reliably verify user nationalities. Now, 18 days later, the model is being switched back on across Claude Platform, Claude.ai, Claude Code, and Claude Cowork.

What happened and what’s changing

Here’s the timeline. Claude Fable 5 and its sibling model, Claude Mythos 5, both launched on June 9 as part of Anthropic’s Mythos-class generation. Three days later, US export controls kicked in, and Anthropic pulled the plug on global access to both models because it couldn’t comply with nationality verification requirements while keeping the service running.

The export controls on Fable 5 were lifted on June 30. Anthropic wasted no time announcing the redeployment for the following day.

But the model isn’t coming back exactly the way it left. Anthropic has used the downtime to bolster the model’s cybersecurity safeguards. New classifiers are being implemented to restrict access to sensitive tasks, and the company is working with Amazon, Microsoft, and Google on developing a jailbreak severity framework.

To sweeten the return, Anthropic is giving paid users a promotional window. Subscribers on Pro, Max, Team, and select Enterprise plans will receive up to 50% of their weekly usage limits free through July 7. After that date, access reverts to the standard usage credit system.

Mythos 5 stays locked down

While Fable 5 gets a global green light, its more powerful counterpart isn’t so lucky. Claude Mythos 5, designed specifically for cybersecurity applications, remains restricted exclusively to approved US organizations dealing with critical infrastructure.

There are no current plans for broader public access to Mythos 5.

The bigger picture for AI and its investors

Any developer who integrated Fable 5 into their workflow between June 9 and June 12 had their capabilities yanked away with no warning. The promotional free access through July 7 is nice, but it doesn’t compensate for the scramble those teams went through.

The collaboration between Anthropic, Amazon, Microsoft, and Google on a jailbreak severity framework is arguably the more consequential development buried in this announcement. When competitors start building shared safety infrastructure, it signals that the industry sees regulation as inevitable and is trying to shape it rather than react to it.

Export controls can appear with minimal notice and reshape the competitive landscape overnight. Anthropic handled this episode relatively quickly, but for any organization betting its infrastructure on a specific AI provider, the lesson is that geographic access to frontier models is now a variable, not a constant.