The U.S. government has ordered Anthropic to block all foreign nationals from accessing its advanced Fable 5 and Mythos 5 AI models due to national security risks surrounding a safeguard bypass exploit. Unable to verify user citizenship instantly, Anthropic has completely disabled both models for all global customers under protest.
The Trump administration has issued an emergency export control directive ordering artificial intelligence developer Anthropic to immediately suspend access to its most advanced AI models for all foreign nationals. The unprecedented federal mandate, delivered by U.S. Commerce Secretary Howard Lutnick on June 12, 2026, bars access to foreign individuals, companies, and governments both outside and within the United States.
Because the San Francisco-based developer cannot verify the citizenship status of hundreds of millions of global users in real time, Anthropic announced it had no choice but to abruptly disable its newly launched Fable 5 and Mythos 5 models for all global customers to ensure total regulatory compliance. The sudden intervention marks a major escalation in federal efforts to police software capabilities directly, shifting focus from hardware supply chains to live software access.
Commerce Department Cites National Security and Safeguard Vulnerabilities
According to letters sent directly to Anthropic Chief Executive Officer Dario Amodei, the U.S. Department of Commerce classified the restriction under emergency national security authorities. The government's directive targets the company's highest-performing "Mythos" software tier, which publicly debuted just three days prior to the enforcement action.
While the official administrative letter did not detail the precise underlying intelligence, federal authorities reportedly issued the order after becoming aware of an unmitigated "jailbreak" technique. This exploit allegedly allows external actors to bypass pre-programmed safety parameters, enabling the models to read specialized codebases and identify exploit vectors or software vulnerabilities. A U.S. official later confirmed that the Bureau of Industry and Security (BIS) spearheaded the export control directive to prevent foreign adversaries from leveraging the system for offensive cyber operations.
Anthropic Complies Under Protest as Tech Sector Faces Disruption
In an official public corporate statement, Anthropic confirmed it had fully complied with the legal directive, pulling the plug across its platform and asking partner infrastructure providers like Amazon Web Services (AWS) to revoke model keys globally. However, the company expressed sharp disagreement with the administration's sweeping enforcement strategy.
Anthropic executives stated that the government provided only "verbal evidence" of a narrow, non-universal bypass that allowed the model to discover minor, previously known software flaws. The company noted that it had thoroughly red-teamed Fable 5 for thousands of hours alongside the U.S. government and the United Kingdom’s AI Safety Institute (UK AISI) prior to release. Furthermore, Anthropic claimed that the level of computing capability flagged by the Commerce Department is already widely available on the open market through competing models, such as OpenAI's GPT-5.5.
Widening Friction Between Frontier AI Developers and Regulators
The enforcement action significantly intensifies existing friction between Anthropic and federal agencies. Relationships between the firm and Washington regulators fractured earlier this year when Anthropic refused to permit the U.S. military to utilize its frontier systems for domestic surveillance and fully autonomous weapons systems. That standoff previously resulted in the government placing Anthropic on a conditional supply chain blacklist set to take effect later this year.
The mandate also presents an acute operational crisis for domestic technology firms employing international talent. Because the export control order applies to all "foreign persons" inside U.S. borders, foreign national AI researchers working within the United States on standard visas—including Anthropic's own staff—are legally blocked from accessing or evaluating their own company's state-of-the-art models.
Why It Matters
The federal shutdown of Fable 5 and Mythos 5 establishes a critical legal precedent for the global software sector, demonstrating that the U.S. government will actively recall commercial AI products over safety bypasses. For enterprise clients, software engineers, and global tech investors, this action introduces severe regulatory unpredictability, showing that cloud-hosted operational models can be deactivated overnight without an administrative appeals process.
Key Facts at a Glance
The Mandate: The U.S. Commerce Department issued an export control directive barring all foreign nationals from accessing Anthropic's top-tier models.
Impacted Systems: Anthropic completely deactivated its premier Fable 5 and Mythos 5 models globally for all users to ensure baseline compliance.
The Vulnerability: Federal officials acted on reports of a specialized "jailbreak" exploit that allowed users to bypass security safeguards to analyze software codebases.
Scope of Enforcement: The restriction covers foreign nationals located both internationally and domestically within the U.S., including Anthropic's visa-holding employees.
Frequently Asked Questions
Are older Anthropic models like Claude 3.5 Sonnet affected by this ban?
No. The emergency Commerce Department order applies strictly to the company's newly deployed, high-capability Mythos class tier, specifically the Fable 5 and Mythos 5 models. Other standard commercial models remain operational.
What is the difference between Fable 5 and Mythos 5?
Mythos 5 is the raw, high-performing foundation model that was restricted to vetted cyber defenders under Anthropic’s "Project Glasswing" program. Fable 5 utilized an additional layer of safety classifiers to protect general enterprise users from generating dangerous exploits.
Can American citizens still use Fable 5 or Mythos 5?
No. Because Anthropic cannot instantly verify the nationality and passport data of millions of active API and chat users in real time, the developer had to shut down the models for all global customers to prevent an accidental export control violation.
How does this affect foreign tech workers inside the United States?
Under U.S. export control definitions, a "foreign person" inside the country is subject to the same restrictions as an overseas entity. International researchers, engineers, and developers residing in the U.S. on work visas are legally barred from accessing these models.
Source: Official regulatory compliance disclosures published by Anthropic PBC, administrative statements confirmed by the U.S. Bureau of Industry and Security (BIS), enterprise customer alerts issued by Amazon Web Services (AWS), and initial investigative updates from Axios and Reuters.
