The U.S. Commerce Department placed Anthropic’s newly released Claude Fable 5 and Mythos 5 models under export controls on June 12, forcing the company to disable both systems worldwide and turning a commercial AI launch into a national-security dispute with direct consequences for AI customers and investors.

According to the source material, Commerce Secretary Howard Lutnick sent Anthropic CEO Dario Amodei a letter at 5:21 p.m. ET on June 12 barring access to the models by any foreign national, including foreign-national employees inside Anthropic. Anthropic concluded it had no clean way to comply selectively and disabled both models for every customer by midnight.

Anthropic had released the Mythos-class models on June 9. Fable 5 was described as the public, heavily guarded commercial model, while Mythos 5 was a more powerful underlying system routed to selected organizations for cyber-defense work through Project Glasswing.

Anthropic said the order cited national-security authorities but gave no specific rationale, and the company publicly called the matter a “misunderstanding.” The company said it believed officials were reacting to a method of jailbreaking Fable 5, while arguing that a narrow jailbreak did not justify recalling a model already in broad use.

Model Reliability Takes A Hit

The immediate issue for customers is not only whether Fable 5 or Mythos 5 posed a security risk. It is that a U.S. frontier model was removed from global access on short notice. For companies building products, workflows, and security tools around hosted AI models, the episode turns regulatory shutdown risk from a theoretical concern into a live procurement issue.

The source material says industry concern is already spreading to stack decisions. Buyers may add regulatory exposure to the reasons they keep multiple model providers, while self-hosted and open-weight systems may gain appeal because their access cannot be revoked in the same way by a provider responding to a government order.

The timing also matters for AI companies pursuing public-market confidence. The order landed as major labs were expected to move toward public offerings, placing new attention on whether frontier AI revenue can be treated as dependable if access can change abruptly under export-control authority.

Three Days From Launch To Shutdown

The models went live on June 9 and were shut down three days later. The source material describes the sequence as a 72-hour shift from launch to global disablement: release on June 9, Commerce letter on June 12, and worldwide loss of access by midnight.

The government’s reported concerns center on cybersecurity and possible foreign access. The U.K. AI Safety Institute’s red-team lead said publicly that his team built a jailbreak that produced malicious answers within hours of access and later extended it to multi-step agentic tool calls, according to the source material.

The Wall Street Journal, cited in the source material, reported that Commerce was alarmed by a separate jailbreak report from Amazon, and that Amazon CEO Andy Jassy called Treasury Secretary Scott Bessent to warn that researchers had obtained cyberattack-usable information from Fable 5. Semafor, also cited, reported that suspicion of access by a China-linked group contributed to the directive.

“misunderstanding”

— Anthropic

Security Case Remains Disputed

It is not yet clear which concern most directly triggered the order: the reported jailbreak findings, Amazon’s reported warning, suspicion of China-linked access, broader proliferation risk, or a mix of those factors. The competing accounts do not fully align.

It is also unsettled whether Fable 5 and Mythos 5 had capabilities meaningfully beyond other current models. The open letter from more than 120 cybersecurity executives and engineers argues that comparable security work can be done with GPT-5.5, Anthropic’s other models, OpenAI’s Daybreak, and China’s Kimi 2.7, according to the source material. That claim remains part of the dispute, not a settled finding.

White House Talks Come Next

Anthropic and White House officials are scheduled to meet on June 22, according to the source material. The central question is whether the export controls will be narrowed, lifted, defended publicly with more detail, or used as a precedent for other frontier AI systems.

Customers are likely to watch for operational guidance as much as policy explanation. Any lasting restriction could push more buyers toward multi-model contracts, domestic hosting, or open-weight alternatives.

Key Questions

What happened to Claude Fable 5 and Mythos 5?

The U.S. Commerce Department placed both models under export controls on June 12, and Anthropic disabled them globally because it could not selectively block all foreign-national access in the time provided.

Why did the government act?

The stated legal basis was national security, but the specific rationale has not been fully detailed. Reported concerns include jailbreak findings, cyberattack-usable outputs, possible China-linked access, and reverse-engineering risk.

Is Anthropic saying the models were unsafe?

No. Anthropic called the matter a “misunderstanding” and argued that a narrow, non-universal jailbreak was not enough reason to recall models it said had undergone extensive red-team testing.

Why are AI customers concerned?

The shutdown showed that a hosted frontier model can become unavailable worldwide with little notice. That may affect procurement, product planning, and dependence on a single AI provider.

What happens on June 22?

Anthropic is expected to meet with the White House. The outcome may determine whether the controls remain in place and how similar cases are handled for other frontier AI models.

Source: Thorsten Meyer AI