Export Controls Hit Anthropic AI Models in Rare Federal Action

The Trump administration has imposed export controls on two of Anthropic's flagship AI models, marking the first time the federal government has used such restrictions against frontier artificial intelligence systems on national security grounds.

The directive, issued last Friday, prevents Anthropic from allowing any foreign national—whether inside or outside the United States—to access its Fable 5 and Mythos 5 models. The ban extends even to the company's own employees who hold foreign citizenship.

Anthropicmet with administration officials this week seeking an agreement to restore access to the models. According to the company, the government claims to have discovered a method to "jailbreak" Fable 5, potentially enabling the model to perform unauthorized actions. Anthropic counters that officials have provided only verbal evidence of "a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws."

Why it matters

This action establishes a new precedent for how the federal government can intervene in AI development and deployment. The restrictions could extend to other major labs including OpenAI and Google, potentially locking enterprise customers out of leading U.S. AI models. For Anthropic specifically, the timing creates significant challenges as the company prepares for an initial public offering later this year—being unable to demonstrate flagship products disrupts enterprise commitments during a critical growth phase.

Second clash with federal authorities

This represents Anthropic's second major conflict with the Trump administration. The company is currently suing the Department of Defense after Secretary Pete Hegseth designated it a supply chain threat, citing Anthropic's refusal to allow the DOD to use its models for autonomous weapons development.

David Sacks, co-chair of the President's Council of Advisers on Science and Technology, disputed any connection between the two actions, stating on X: "It is frankly bewildering that Anthropic hasn't wanted to comply with safety requests that it previously said were its highest priority." Sacks maintains the export control decision stands independent of the prior DOD designation.

Broader industry implications

"Not being able to make available two flagship models does disrupt enterprise commitments and pauses momentum at a critical growth stage ahead of its IPO," said Gartner analyst Arun Chandrasekaran. "At the same time, the ban could eventually extend to other frontier labs like OpenAI and Google, in which case organizations may be locked out from all leading US models equally."

Chandrasekaran characterized the action as "the first major use of export controls to limit access to frontier AI models on national security grounds," suggesting it could establish a framework for future government intervention in the AI sector.

These details were first reported by AI Watch.