The U.S. government has ordered AI company Anthropic to suspend access to its most advanced models, Fable 5 and Mythos 5, for all foreign nationals, citing national security concerns. Anthropic said it received an export-control directive requiring immediate action and complied by disabling access to the models while it seeks clarification from government officials.
According to Anthropic, the government's concern appears to involve a potential “jailbreak” technique that could bypass some of the models’ safeguards and potentially assist in identifying software vulnerabilities. The company disputes the severity of the risk, arguing that the alleged weakness is limited and not unique to its systems, but it nevertheless shut down access to ensure compliance with the order.
The directive marks a significant expansion of U.S. AI export controls. Previous restrictions largely focused on advanced semiconductors and computing hardware, whereas this action directly targets access to frontier AI models. The restrictions reportedly apply to all foreign nationals, including individuals located within the United States and even some of Anthropic’s own international employees.
The move comes amid broader tensions between Anthropic and the U.S. government over AI policy and national security. Earlier disputes involved the company’s refusal to relax restrictions on military uses of its technology, leading to legal and regulatory clashes. Anthropic has described the latest action as a misunderstanding and says it is working to restore access as quickly as possible while continuing discussions with officials.
