Anthropic Blocks Fable 5 and Mythos 5 Access After US Government Directive

Anthropic has disabled access to Claude Fable 5 and Claude Mythos 5 for all customers after receiving a US government export control directive. In an official statement, the company said the order required it to suspend access to both models for any foreign national, whether inside or outside the United States.

The company said the directive also covered foreign-national Anthropic employees. Because Anthropic said it could not apply that nationality-based restriction reliably in real time, it removed access to both models for all customers. Other Claude models remain available.

The suspension came only days after Anthropic introduced Claude Fable 5 and Claude Mythos 5 on June 9. Fable 5 was positioned as the general-use model, while Mythos 5 carried the more sensitive cybersecurity and biology capabilities.

Why Anthropic shut down both models

Anthropic said it received the government directive at 5:21 p.m. ET on June 12. The company said the letter cited national security authorities but did not provide detailed public evidence of the underlying concern.

According to Anthropic, the government’s concern centered on a possible way to bypass, or jailbreak, Fable 5. The company said the technique it reviewed involved asking the model to read a specific codebase and identify software flaws.

Anthropic disputed the severity of the finding. The company said the demonstrated technique produced only a small number of previously known, minor vulnerabilities and did not show a broad or universal jailbreak.

Model	Original availability	Status after directive	Main issue
Claude Fable 5	General-use Mythos-class model with safeguards	Disabled for all users	Government concern over possible safeguard bypass
Claude Mythos 5	Limited to vetted partners	Disabled for all users	High dual-use capability in cybersecurity and biology
Other Claude models	Available to eligible users	Still available	Not covered by the directive

Fable 5 and Mythos 5 were designed for different access levels

Anthropic’s Claude Mythos page says Mythos 5 was available only to a small group of vetted partners, with a goal of broader access later. The company described Mythos 5 as highly capable in cybersecurity, biology research, and healthcare.

Fable 5 used the same underlying model as Mythos 5, but with stronger safeguards for cybersecurity and biology. Anthropic said queries in those riskier domains were routed away from Fable 5 to Opus 4.8.

The distinction matters because the shutdown affected both the public-facing model and the more restricted partner model. Customers who had just started testing Fable 5 lost access, while vetted partners also lost access to Mythos 5.

Data retention was already a concern for some customers

Before the shutdown, Anthropic had already changed the data handling rules for Mythos-class models. Its data retention guidance says prompts and completions for covered models are retained for at least 30 days.

That requirement applies to Claude Fable 5 and Claude Mythos 5. Anthropic says the retention is used for safety monitoring, with automatic deletion after 30 days except in limited cases such as safety investigations or legal requirements.

The retention rule raised enterprise questions even before the government directive. Some organizations that rely on zero data retention policies had to review whether they could use Mythos-class models for sensitive work.

The directive raises broader AI export control questions

The case is now bigger than one company. Reuters reported that G7 leaders discussed a possible trusted-partner framework for access to advanced US AI models after the Anthropic restriction.

That discussion matters because frontier AI models can support both defensive and harmful work. A model that helps researchers find software flaws can also help attackers if it is misused.

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of…

— Anthropic (@AnthropicAI) June 13, 2026

The current dispute shows how difficult model access controls can become when governments treat AI capabilities like export-controlled technology. Nationality-based restrictions are especially hard to enforce across global cloud products, enterprise accounts, contractors, and employees.

What customers should do now

Customers who used Fable 5 or Mythos 5 should move workloads to another available Claude model or another approved provider while the restriction remains in place. Anthropic said access to all other Claude models is not affected.

Developers should also check applications, agents, API settings, and internal tools that referenced Fable 5 or Mythos 5 directly. Any hardcoded model name could now fail until access is restored or the app routes to a fallback model.

• Check whether apps or workflows call Fable 5 or Mythos 5 directly.
• Route production workloads to another approved model.
• Review data retention requirements before moving sensitive workloads.
• Tell internal users which Claude models remain available.
• Monitor Anthropic status and policy updates for restoration details.

Anthropic says it disagrees with the decision

Anthropic said it is complying with the legal directive, but it disagrees that the reported jailbreak concern should require recalling a commercial model used by customers. The company warned that applying the same standard across the industry could halt new frontier model deployments.

The company also said it supports a government role in blocking unsafe deployments, but only through a transparent, fair, clear, and technically grounded process. Anthropic said it is working to restore access as soon as possible.

The company’s statement frames the shutdown as a compliance move, not a voluntary product retirement. That leaves open the possibility of access returning if Anthropic and the government reach a new arrangement.

Why the models drew government attention

The June 9 launch post described Fable 5 as strong at vision, long-context tasks, agentic coding, research, and complex knowledge work. Anthropic also highlighted Mythos 5 for drug design, molecular biology, genomics, cybersecurity, and healthcare.

Those capabilities explain why governments may see these systems as sensitive. The same model class that can help defend critical software or accelerate science may also create risk if advanced misuse controls fail.

Anthropic’s Mythos access page says the company planned to open broader trusted access over time. The government directive now complicates that plan and may push the industry toward stricter approvals for frontier AI releases.

Customers are waiting for a path back

The immediate business impact is clear. Customers who adopted Fable 5 days after launch had to switch models quickly, while trusted partners lost access to Mythos 5 capabilities used for specialized research and security work.

Anthropic’s retention policy will also remain part of the discussion if access returns. Enterprises will need to evaluate both government access limits and data handling requirements before moving sensitive workloads back to these models.

Reuters reported that allied governments are discussing trusted access, which suggests the shutdown may lead to a more formal system for who can use the most advanced US AI models. Until then, Fable 5 and Mythos 5 remain unavailable.

FAQ