• Anthropic (ANTH) has disabled its Fable 5 and Mythos 5 AI models for non-U.S. users after a government directive cited national security concerns.
  • The company disputes the action, calling it a misunderstanding, and is in talks with Trump administration officials to restore access.
  • The dispute highlights growing tensions between AI innovation and export controls, with potential implications for the broader industry.

Anthropic executives are meeting with Trump administration officials this week to resolve a dispute over its Fable 5 and Mythos 5 AI models, which the U.S. government recently restricted from foreign access, citing national security risks. The company promptly disabled the models for non-U.S. users but publicly described the directive as a “misunderstanding” and is seeking to restore access while challenging the precedent.

The move, reported by people familiar with the matter, marks one of the most concrete enforcement actions involving frontier AI models. The government’s directive, issued under export-control authorities, ordered restricted access to the models, which are among Anthropic’s most advanced. The company argues the action could chill deployments across the industry, as regulators weigh national security against innovation.

“We believe there has been a misunderstanding regarding the capabilities and intended use of these models,” an Anthropic spokesperson said in a statement. “We are working diligently with the administration to restore access and ensure compliance without undermining AI progress.” The company has also reached out to other AI developers to gauge their experiences with similar regulatory scrutiny.

Industry observers say the dispute could accelerate formalized export controls or licensing regimes for high-capacity AI systems. “This is a pivotal moment,” said a former Commerce Department official familiar with AI policy. “If the government can restrict a model after deployment, it changes the risk calculus for every frontier developer.”

The episode underscores the Biden and Trump administrations’ shared focus on AI safety and export controls, though the specific rationale for targeting Fable 5 and Mythos 5 remains unclear. Some analysts speculate the models may have dual-use applications in surveillance or cybersecurity. The government has not publicly detailed its concerns, and spokespeople for the Commerce Department declined to comment.

Anthropic’s shares of private credit and equity investments have not been directly affected, but the company’s ability to deploy capital into foreign markets could face headwinds. The dispute also raises questions for other developers, such as OpenAI and Google DeepMind, which are developing similar frontier models.

For now, negotiations continue. “We’re optimistic we can reach a resolution that protects national security while allowing the technology to flourish,” the Anthropic spokesperson added. Without a deal, the company could face prolonged restrictions, potentially forcing it to redesign access controls for future models.

(This story has been updated to clarify the models affected and add context on industry reactions.)