US Bars Foreign Nationals From Anthropic's Top AI Models

Anthropic moved quickly Friday to cut off access to its newest AI models after receiving a direct order from US government agencies. The directive required the company to block every foreign national from using Fable 5 and Mythos 5, no matter where those users were located or whether they worked for Anthropic inside the United States.

The order arrived at 5:21 p.m. and left the company with little choice but to disable the tools for all customers to stay in compliance. Other Anthropic products, including its Claude chatbot, were left untouched. Officials gave no detailed explanation for the action beyond broad national security language. Anthropic said it believes the move stems from reports that someone had found a way to bypass safety limits on Fable 5 shortly after its public launch on June 9.

Mythos 5 has long been restricted to a small group of approved partners because of its strength at spotting software weaknesses that had gone unnoticed for years. Fable 5 was meant to bring many of those same abilities into wider use, though some of its more sensitive functions were already limited. Both models now sit behind the new government barrier. Foreign employees at Anthropic itself lost access along with everyone else, a detail that shows how wide the order reached.

The sudden cutoff raises questions about how federal agencies decide which technologies count as too risky for anyone outside a narrow circle. No public evidence has been offered that the models were already being misused abroad, yet the response was total and immediate. Companies that had started testing Fable 5 for legitimate work now face an abrupt halt with no clear timeline for restoration.

Anthropic had taken steps it considered sufficient to keep the models secure before release. Those precautions did not satisfy the officials who issued the directive. The episode follows a pattern in which Washington cites national security to restrict advanced tools without spelling out the exact threat or allowing affected parties much room to respond.

Foreign nationals inside the country who hold valid work authorization at American firms are now treated the same as users located overseas. That approach sweeps in large numbers of engineers and researchers employed by domestic companies. Anthropic described the order as requiring suspension for all such individuals, whether inside or outside the United States.

The models in question sit at the frontier of what current AI can do in cybersecurity. Their ability to surface long-hidden vulnerabilities gives whoever controls them a significant edge. By forcing a complete block rather than a narrower fix, the government has concentrated that edge further. Anthropic complied at once, but the lack of specifics in the directive leaves open whether the real concern was a technical loophole or simply the spread of powerful capability beyond official channels.

No other Anthropic offerings were included in the order. The company continues to operate its main chatbot and earlier models without interruption. The targeted action against just these two releases, however, signals that officials view the newest cybersecurity-focused systems as carrying special risks that warrant extraordinary measures. How long the restriction will last and what additional requirements might follow remain unknown.