Anthropic Restricts Access to Advanced AI Models After U.S. Export Control Order

Anthropic said it is disabling access to two of its most advanced AI models after receiving a U.S. government order restricting foreign nationals from using them, marking a significant step in Washington’s efforts to control access to powerful artificial intelligence systems.

The company said the directive applies to its Fable 5 and Mythos 5 models. Although the order is aimed at limiting access for foreign nationals, Anthropic said it must remove access for all customers in order to comply.

The move comes at a time when U.S. officials are paying closer attention to how advanced AI models could be used in cybersecurity, national security and other sensitive areas.

Government Cites National Security Concerns

According to Anthropic, the U.S. government cited national security concerns in its order but did not provide detailed evidence to the company. Anthropic said it understands the concern relates to a possible way of bypassing safeguards in Fable 5.

Those safeguards are designed to prevent misuse in areas such as identifying software vulnerabilities. The company said the issue appears to involve a narrow potential jailbreak rather than a broad failure of the model’s safety systems.

Anthropic disagreed with the government’s action, saying that such a concern should not be enough reason to recall a commercial model already being used at scale. The company said it is working with officials to resolve the issue and restore access.

Access Removed for All Customers

Anthropic said the order directly targets foreign access, but the company’s implementation means access to the affected models is being removed more broadly.

AWS, Amazon’s cloud division, said Anthropic asked it to revoke access to the affected models for users across all regions. The company’s other AI models are not expected to be affected by the order.

The restriction applies to Fable 5 and Mythos 5, which Anthropic described as among its most capable systems. Fable 5 was recently introduced as part of the company’s newer generation of advanced AI models.

Dispute Over Model Safety

Anthropic said it had worked with the U.S. government and others on safety measures before releasing Fable 5. The company has also said the model includes guardrails for potentially risky use cases, including cybersecurity-related activity.

The company argued that rival AI systems show similar abilities to identify minor software bugs, suggesting the concern may not be unique to Anthropic’s model. Anthropic warned that using this standard across the industry could affect other frontier AI developers as well.

The disagreement shows the growing tension between AI companies and regulators over how to measure safety risks in advanced models. Companies want to release more capable systems for commercial and research use, while governments are increasingly concerned about misuse.

A Shift in U.S. AI Controls

The order points to a possible shift in how the United States approaches AI regulation and export controls. Until now, much of the focus has been on restricting access to advanced chips and infrastructure used to train powerful AI systems.

This case suggests that government controls may also extend more directly to AI models themselves, especially when officials believe those models could create national security risks.

If similar actions are applied more broadly, AI companies may face new limits on how they release, distribute or host their most advanced systems for international customers.

Broader Pressure on Anthropic

The development adds to existing pressure on Anthropic’s relationship with the U.S. government. Reuters reported that the company had earlier been placed on a Pentagon supply chain blacklist after a disagreement over the use of its AI models for domestic surveillance and fully autonomous weapons systems.

Anthropic has positioned itself as an AI company focused on safety and responsible deployment. The latest dispute shows that even companies emphasizing safeguards may face regulatory challenges as their systems become more capable.

What Happens Next

A U.S. official confirmed that the Commerce Department issued an export control directive requiring access to Fable 5 and Mythos 5 to be suspended for foreign nationals.

Anthropic said it believes there has been a misunderstanding and is working to restore access as soon as possible. For now, customers using the affected models may need to rely on other Anthropic systems or alternative AI tools.

The case is likely to be watched closely by AI companies, cloud providers, policymakers and enterprise users. It raises a central question for the industry: how should governments balance national security concerns with commercial access to advanced AI technology?