U.S. Export Control Directive Forces Anthropic to Disable Claude Fable 5 and Mythos 5

Anthropic has disabled access to Claude Fable 5 and Claude Mythos 5 after receiving a U.S. government export control directive that bars foreign nationals from using the models.

The company said in an Anthropic statement that the directive applies to any foreign national, whether inside or outside the United States. That includes non-U.S. Anthropic employees.

Because Anthropic said it could not reliably separate access by nationality across all products and customers, it disabled Fable 5 and Mythos 5 for everyone. Other Claude models remain available.

What Happened to Claude Fable 5 and Mythos 5

Anthropic received the directive on June 12, 2026, at 5:21 p.m. ET. The company said the letter did not provide specific details about the national security concern behind the order.

The move came only days after Anthropic launched Claude Fable 5 and Claude Mythos 5 on June 9. Fable 5 was positioned as a general advanced model, while Mythos 5 was aimed at a smaller group of cyberdefenders and infrastructure providers.

Anthropic said it believes the government’s concern relates to a method of bypassing, or jailbreaking, Fable 5. The company said it reviewed a demonstration of the technique and described the results as a small number of previously known, minor software vulnerabilities.

Model	Original role	Current status
Claude Fable 5	Public Mythos-class model with stronger safeguards	Disabled for all customers
Claude Mythos 5	Restricted model for selected cyberdefenders and infrastructure providers	Disabled for all customers
Other Claude models	General Claude product lineup	Not affected by the directive, according to Anthropic

Why the Export Control Order Matters

The directive is unusual because it affects access to cloud-based AI models rather than only physical exports such as chips, servers, or equipment.

The Bureau of Industry and Security says in its end-use and end-user controls guidance that BIS may inform a person by specific notice that a license is required for an export, reexport, or in-country transfer when there is an unacceptable risk of diversion to specified end uses.

That kind of notice is often described as an “is informed” letter. Legal analysts say this case raises new questions because the directive appears to apply to access by foreign nationals globally, including foreign nationals working inside the United States.

Anthropic Disputes the Safety Rationale

Anthropic said Fable 5 had gone through extensive red-team testing before launch. The company said it worked with the U.S. government, the U.K. AI Security Institute, private third parties, and internal teams for thousands of hours of testing.

The company also said Fable 5 included strong safeguards against cybersecurity misuse. It argued that the reported bypass did not justify a universal shutdown because the observed flaws were already known and minor.

In the same public statement, Anthropic said the order forced an abrupt service change and that it was working with the government to restore access as soon as possible.

Mythos 5 Was Designed for Sensitive Cybersecurity Work

Anthropic described Mythos 5 as the same underlying model as Fable 5, but with some safeguards lifted in certain areas for trusted cybersecurity use.

The company said Mythos 5 would initially be deployed through Project Glasswing in collaboration with the U.S. government. Anthropic described the model as having the strongest cybersecurity capabilities of any model in the world.

In its launch post for Fable 5 and Mythos 5, Anthropic said the models could help cyberdefenders secure important software and assist life sciences research, coding, knowledge work, and vision tasks.

Legal Questions Are Still Open

The Export Administration Regulations include controls for certain end uses and end users. The EAR Part 744 rules cover several end-use and end-user restrictions, including military-intelligence controls.

However, the Anthropic directive raises a newer question: how export control law applies to remote access to advanced AI models when the service is delivered through cloud infrastructure.

A Just Security analysis noted that the order appears to push export control tools into a frontier AI access dispute, creating important questions about statutory authority, process, and scope.

Issue	Why it matters
Foreign national access	The directive affects users outside the U.S. and foreign nationals inside the U.S.
Cloud model access	The case tests how export rules apply to hosted AI services
Employee access	Non-U.S. employees may be blocked from working with restricted models
Enterprise customers	Companies using global teams may face access and compliance uncertainty
Future AI regulation	The directive could become a precedent for other frontier AI models

EU Officials Are Reviewing the Impact

The European Commission is assessing the practical consequences of the U.S. directive, according to Reuters.

A Commission spokesperson said AI progress can bring benefits, including in cybersecurity, but also creates risks. The spokesperson also said any measures should not discriminate against international partners.

The reaction shows the diplomatic risk around AI export controls. A U.S. national security action that cuts off allied users can create friction with partners that also rely on U.S. AI services.

What This Means for Enterprise AI Customers

For companies, the immediate impact is access disruption. Customers that had started testing Fable 5 or Mythos 5 must move workloads to other Claude models or competing AI systems.

Businesses with multinational teams face a broader issue. If future AI models receive similar restrictions, companies may need nationality-aware access controls, model-specific compliance reviews, and clearer policies for employees working across borders.

The case also shows why enterprises should avoid building critical workflows around a single frontier model without fallback options. Export controls, safety interventions, and regulatory changes can affect model availability with little warning.

• Review whether any workloads depended on Fable 5 or Mythos 5.
• Move affected workflows to supported Claude models or approved alternatives.
• Check internal access policies for foreign national and cross-border AI use.
• Update procurement and legal teams on possible model-level export restrictions.
• Prepare fallback plans for high-risk AI dependencies.

BIS Controls Could Shape Future AI Access

The BIS guidance on end-use and end-user restrictions explains that license requirements can apply even when items do not otherwise require a license, depending on the end user or end use.

If advanced AI model access gets treated more like controlled technology, AI labs may need new compliance systems for customers, employees, contractors, APIs, and cloud deployments.

The EAR Part 744 framework already gives BIS tools to impose end-use and end-user controls. The Anthropic case may now test how far those tools can extend in the AI service market.

Foreign Access Is the Central Compliance Challenge

Foreign national restrictions can create practical problems for AI companies. Model development, testing, safety research, customer support, and enterprise deployment often involve international employees and contractors.

That makes a blanket foreign-national restriction difficult to implement without shutting down access more broadly. Anthropic chose that broader shutdown to avoid violating the directive.

The European Commission’s review, reported by Reuters, suggests other governments may press for more clarity if U.S. actions affect allied researchers, businesses, and public institutions.

A Precedent for Frontier AI Regulation

The Fable 5 and Mythos 5 shutdown marks one of the clearest examples yet of U.S. export control pressure reaching a commercial AI model after launch.

Legal debate will likely continue. The Just Security legal analysis says the case raises unresolved questions about how export control authorities should apply to frontier AI and whether emergency-style actions provide enough transparency.

For now, the outcome is clear for users. Claude Fable 5 and Claude Mythos 5 are offline, Anthropic is seeking a path to restore access, and the U.S. government has shown it may use export controls directly against advanced AI model availability.

FAQ