U.S. Orders Anthropic to Block Foreign Access to Advanced AI Models

Anthropic has disabled access to two of its newest artificial intelligence models after the U.S. government ordered the company to suspend use by foreign nationals.

The directive applies to Fable 5 and Mythos 5, two advanced AI systems that had drawn attention because of their potential cybersecurity capabilities. According to Anthropic, the order covers foreign nationals both inside and outside the United States, including foreign-national employees who work at Anthropic.

Because of the scope of the government order, Anthropic said it had to shut off access to the affected models for all customers in order to comply.

The company said access to its other AI models was not affected.

Anthropic Says It Received the Directive Friday Evening

Anthropic said it received the government directive at 5:21 p.m. Eastern Time on Friday. The company said the order cited national security authorities, but did not give specific details about the concern.

In a public statement, Anthropic said:

“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.”

The company also said:

“Access to all other Anthropic models will not be affected.”

The order reportedly came from the U.S. Commerce Department and appears to rely on national security export-control authority. Those rules can restrict not only access by people outside the United States, but also access by foreign nationals inside the country.

That means the restriction may affect some of the same engineers and researchers who helped build or test the models.

Why Fable 5 and Mythos 5 Drew Attention

Mythos 5 had already become a point of concern in Washington and the technology industry because of its advanced cybersecurity abilities.

Anthropic had previously limited access to Mythos, saying the model could be powerful enough to help identify or exploit software weaknesses. The company initially shared it with select organizations involved in maintaining critical computer infrastructure so they could use it defensively.

Fable 5 was later introduced as a more restricted version of the same underlying technology. Anthropic said it had added strong guardrails to reduce the risk that the system could be misused for cybersecurity, biology, or other sensitive topics.

In practice, that meant Fable 5 could be useful for some advanced work while refusing many requests Anthropic considered risky.

Government Concern Focused on a Possible Jailbreak

Anthropic said its understanding is that the government became concerned after learning about a possible method of bypassing, or jailbreaking, Fable 5’s safeguards.

The company disputed the seriousness of the issue. Anthropic said it reviewed a demonstration of the technique and believed it was used only to identify a small number of previously known, minor vulnerabilities.

Anthropic argued that similar results could be produced by other public models without using a bypass.

The company wrote:

“We believe this is a misunderstanding and are working to restore access as soon as possible.”

Anthropic also said it had not received evidence of a universal jailbreak that broadly defeats Fable 5’s safeguards.

The Company Pushes Back While Complying

Anthropic said it is complying with the legal directive and removing access to Fable 5 and Mythos 5.

At the same time, the company strongly disagreed with the government’s action.

Anthropic said:

“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.”

The company added that if the same standard were applied across the AI industry, it could effectively stop new frontier AI model launches.

Anthropic said it believes the government should be able to block unsafe AI deployments, but only through a process that is transparent, fair, clear, and based on technical evidence.

It said:

“This action does not adhere to those principles.”

A New Kind of AI Restriction

Export controls are not new. The U.S. government has long restricted certain technologies, including advanced computer chips, from being sold or transferred to foreign adversaries.

But this directive appears broader and more unusual because it focuses on who may use a specific AI model, not just where a product may be shipped.

The order could theoretically block access by citizens of allied countries such as Canada or Britain if they are considered foreign nationals under the directive. It could also limit the ability of foreign-national employees at Anthropic to work on the affected models.

That is why the move drew immediate attention from AI policy experts, cybersecurity researchers, investors, and technology executives.

Industry Reaction Was Swift

The government order triggered strong reactions across the AI and policy world.

Some critics saw the move as too broad and potentially damaging to U.S. AI development. Others argued that Anthropic’s own messaging around the power and danger of Mythos may have contributed to the government’s response.

Dean Ball, a senior fellow at the Foundation for American Innovation and a former AI adviser in the Trump administration, called the decision “baffling.”

He wrote:

“I have no words.”

Other experts warned that the directive could create uncertainty for AI companies, employees, and investors.

Some compared the situation to earlier fights over encryption export controls in the 1990s, when the U.S. government tried to limit the spread of strong cryptography. Those efforts eventually became difficult to maintain because software and mathematical methods were hard to confine within national borders.

Cybersecurity Debate Remains Unsettled

The dispute highlights a larger question in artificial intelligence: when does a model become dangerous enough to justify government restriction?

Advanced AI systems can help defenders find and patch software vulnerabilities. But the same capabilities could also help attackers discover weaknesses faster.

That dual-use nature makes regulation difficult.

A model that helps protect critical infrastructure may also be useful to hackers. A safeguard that blocks malicious use may also block legitimate researchers. A government restriction that reduces risk may also slow innovation or push talent elsewhere.

Anthropic has argued that its approach to Fable 5 was based on layered safeguards, red-teaming, monitoring, and restrictions on risky outputs. The company said it worked with government and outside organizations to test the model before release.

The government, however, appears to have decided that the risk was significant enough to require an immediate suspension of access by foreign nationals.

Older Claude Models Remain Available

Anthropic said the restriction applies only to Fable 5 and Mythos 5.

Other models, including older Claude systems, were not affected by the order. That means Anthropic customers may still have access to the company’s other AI products, even though the newest and most advanced affected models are unavailable.

The company said it is trying to resolve the situation and restore access.

Why This Matters

The Anthropic order may become an important moment in AI regulation.

Until now, much of the debate around AI safety has focused on voluntary commitments, model evaluations, and proposed oversight. This directive shows the federal government may be willing to use national security powers to directly limit access to specific AI systems.

That raises major questions for AI companies.

Will advanced models need government approval before release? Will foreign nationals be restricted from working on certain systems? Will different countries create competing rules for frontier AI? Will companies become more cautious about describing their models as dangerous?

The answers are still unclear.

But the message is clear: the most advanced AI systems are no longer being treated only as commercial products. They are increasingly being viewed through the lens of national security.

The Bottom Line

Anthropic’s decision to disable Fable 5 and Mythos 5 came after a U.S. government directive barring access by foreign nationals.

The company says it is complying with the order but believes the action is based on a misunderstanding of a narrow potential jailbreak. Government officials appear to be concerned that the models could pose cybersecurity risks if misused.

The move is unusual, disruptive, and likely to intensify the debate over how advanced AI should be regulated.

For now, Fable 5 and Mythos 5 are offline for customers, while Anthropic works to restore access and policymakers confront a difficult question: how powerful can an AI system become before it is treated like a controlled national-security technology?