US Lifts Export Ban on Anthropic Claude Fable 5 After Jailbreak Fix

For weeks, Anthropic’s most advanced public-facing model sat effectively locked away — not because of a technical failure, but because of a US government export control order that barred foreign nationals from touching it. Now, after weeks of negotiations with the Trump administration, Anthropic Claude Fable 5 is cleared to come back online. The US Department of Commerce has lifted export controls on both Claude Fable 5 and Mythos 5, ending one of the more dramatic regulatory standoffs in recent AI history.

Key takeaways

• The US Department of Commerce lifted export controls on Claude Fable 5 and Mythos 5, allowing Anthropic to restore global access.
• Access restoration begins Wednesday on Claude platforms, including Claude.AI and Claude Code, with AWS, Google Cloud, and Microsoft Foundry to follow without a set timeline.
• The original export control directive, issued in early June, barred all foreign nationals — including Anthropic’s own non-US employees — from using either model.
• Amazon researchers flagged a jailbreak vulnerability that triggered the government directive; Anthropic has since developed a new safety classifier to block such attempts.
• Blocked Fable 5 requests will be rerouted to the Opus 4.8 model, with users notified when that happens.

Anthropic to Restore Global Access to Claude Fable 5

The Department of Commerce’s decision to lift controls on both Fable 5 and Mythos 5 clears the path for a broad redeployment. Anthropic confirmed it will begin restoring access on Wednesday to users worldwide across Claude platforms — including Claude.AI and Claude Code. Fable 5 will also count toward up to 50% of weekly usage limits through July 7 for Pro, Max, Team, and selected enterprise plan users, giving returning subscribers an immediate incentive.

Cloud platform availability is a different story. Access on AWS, Google Cloud, and Microsoft Foundry will be re-enabled as soon as possible, Anthropic said, but no concrete timeline has been set. For enterprise clients relying on those infrastructure layers, that ambiguity is the most operationally significant detail in the announcement.

Commerce Secretary Howard Lutnick posted on X that his department had worked closely with Anthropic over the past two weeks “to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America’s leadership in AI.” That framing matters: it signals that the administration now views the model — and the company — as an asset rather than a liability, at least for the moment.

Background: Government Export Controls and Jailbreak Concerns

To understand why this restoration is significant, it helps to recall exactly how severe the original restriction was. In early June, Anthropic received what was described as a Friday evening ultimatum from the Trump administration. The government issued an export control directive invoking national security authorities and ordering Anthropic to suspend access for “any foreign national, whether inside or outside the United States” — including foreign national Anthropic employees.

That meant both Fable 5 and Mythos 5, models the company had spent the preceding week publicly hyping as state-of-the-art, were effectively yanked from a large portion of their intended user base almost immediately after launch. Enterprise clients with international teams found their staff locked out. Anthropic’s own workforce was partly affected.

The trigger for all of this: Amazon researchers had flagged a jailbreak vulnerability in the models. That discovery set the export control directive into motion. The concern, broadly, was that a model capable of being jailbroken could be exploited by foreign actors in ways that implicated national security — a rationale that, however contested in industry circles, gave the government legal standing to act.

The timing carried a competitive sting. The government crackdown on Anthropic coincided with a notable rise in capable Chinese open-source models that were proving nearly as powerful and significantly cheaper than leading US alternatives. Several tech investors and executives argued publicly that restricting Anthropic’s rollout was effectively handing Chinese developers valuable catch-up time. That pressure almost certainly accelerated the resolution.

Anthropic’s Safety Enhancements and Operational Adjustments

Restoring access wasn’t simply a matter of flipping a switch. Anthropic had to demonstrate it had addressed the underlying vulnerability before the government would lift controls. The company published a detailed blog post outlining its revised safeguards, new information-sharing processes with the government, and prerelease testing protocols for future models.

The core technical fix: Anthropic says it has trained an improved safety classifier that specifically targets and blocks the jailbreak behavior that originally prompted the directive. When a user request to Fable 5 trips that classifier, the request is blocked — and automatically rerouted to Opus 4.8 instead. Users receive a notification when this happens, maintaining service continuity while keeping the higher-risk outputs from reaching them.

This kind of tiered fallback architecture is worth noting. It reflects a broader design philosophy where safety interventions don’t simply result in error messages but in degraded-but-functional alternatives. Whether the classifier is robust enough to hold under adversarial pressure is a question only extended real-world deployment will answer.

Anthropic also said it is collaborating with the government on new AI safety and information-sharing processes — a shift in the company-regulator relationship that goes beyond this specific incident. The blog post framed these as ongoing commitments rather than one-time fixes, suggesting the company is positioning itself for a more institutionalized relationship with federal oversight bodies.

Cloud Platform Access and What Comes Next

Wednesday’s global rollout on Claude-native platforms is the first phase. The second — restoring Fable 5 on AWS, Google Cloud, and Microsoft Foundry — remains open-ended. For many enterprise developers who build on those cloud layers, the model still isn’t available in the environment where they actually work.

Separately, Anthropic said it has already restored Mythos 5 access for some US organizations following government approval granted on June 26. The company also said it will continue working with the government to expand Mythos 5 access to more domestic and international partners through its Glasswing program — its cybersecurity initiative that provides selected organizations access to advanced AI for defensive security testing.

What makes this resolution interesting from a regulatory standpoint is what it reveals about the emerging government-AI company dynamic. The Trump administration moved quickly and bluntly — a Friday-night ultimatum, a broad national security framing — but ultimately negotiated a resolution within weeks rather than months. The speed suggests both sides had strong incentives to reach an agreement. For Anthropic, every week offline was a commercial and reputational cost; for the administration, a prolonged shutdown of a leading US AI model while Chinese competitors advanced was its own kind of strategic liability.

The real test of the new equilibrium will come with the next model launch. Anthropic has now committed to prerelease government coordination and expanded information-sharing. Whether that framework holds — and whether it slows or shapes future releases — may define how the next chapter of US AI regulation actually plays out.

FAQ

Why was access to Claude Fable 5 initially restricted?

Access was restricted due to a US government export control directive, issued in early June, prompted by concerns over potential AI jailbreak vulnerabilities flagged by Amazon researchers. The directive barred all foreign nationals, including non-US Anthropic employees and enterprise clients’ international staff, from using either Fable 5 or Mythos 5.

When will access to Claude Fable 5 be restored globally?

Anthropic plans to begin restoring access globally on Claude platforms — including Claude.AI and Claude Code — starting Wednesday, following the Department of Commerce lifting export controls on the model.

What safety measures has Anthropic implemented to prevent AI jailbreaks?

Anthropic developed an improved safety classifier that specifically targets and blocks the jailbreak behavior at issue. If a Fable 5 request is blocked by the classifier, users are notified and the request is automatically rerouted to the Opus 4.8 model. Anthropic also published a blog detailing broader safeguards and new information-sharing processes with the government.

Will Claude Fable 5 be available on major cloud platforms immediately?

Not immediately. Access on AWS, Google Cloud, and Microsoft Foundry will be re-enabled as soon as possible, Anthropic said, but no specific timeline has been announced. Global restoration on Claude-native platforms takes priority in the initial rollout.

Article produced with the assistance of artificial intelligence and reviewed by the editorial team.