U.S. Orders Anthropic to Suspend Advanced AI Models for All Foreign Nationals

The U.S. government has ordered artificial intelligence company Anthropic to abruptly suspend access to its newly released Fable 5 and Mythos 5 models for all foreign nationals, citing unspecified national security concerns. The emergency export-control directive, issued late Friday afternoon, forced the company to pull the plug on its most capable systems just three days after their public debut.[1][2][4][5]

Because Anthropic cannot instantly verify the citizenship or nationality of every user in real time, the company was forced to disable the models globally for all customers to ensure compliance. Access to the company's older models, such as the Claude 3 series, remains unaffected by the order.[2][3][5][6]

The scope of the directive is unusually broad and presents a logistical nightmare for the AI developer. Issued by Commerce Secretary Howard Lutnick, the order applies not just to users located outside the United States, but to any foreign national regardless of their physical location. Crucially, this includes Anthropic's own foreign-national employees working inside the U.S., effectively locking them out of the very systems they helped build.[2][3][4][6]

While the government's letter did not detail the underlying national security threat in writing, Anthropic stated that officials provided verbal evidence of a potential "jailbreak" vulnerability in Fable 5. The government demonstrated a method where the model could be bypassed by asking it to read a specific codebase and fix software flaws, raising fears that the system's advanced coding capabilities could be weaponized by foreign adversaries to develop cyber exploits.[2][4][5][6]

Anthropic complied with the directive under protest, arguing that the alleged vulnerabilities are minor, relatively simple, and already known to the industry. The company noted that it had successfully replicated the same bypass technique on other publicly available models, including OpenAI's GPT-5.5, which is not currently subject to similar export controls.[2][4][6]

"We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic said in a statement, warning that applying this standard evenly would essentially halt all new frontier model deployments across the industry.[2][4]

The directive marks a profound shift in how the U.S. regulates artificial intelligence. Historically, tech companies have relied on geographic geofencing to comply with export controls, blocking IP addresses from sanctioned countries. By tying access restrictions to individual nationality rather than network location, the government is forcing cloud providers to navigate complex identity verification, API key scoping, and enterprise attestations.[3]

The sudden shutdown follows months of escalating tension between Anthropic and the U.S. defense establishment. Earlier this year, the Department of Defense labeled Anthropic a "supply chain risk" after the company established strict red lines against the military use of its technology, a designation Anthropic is currently fighting in court.[6]

Adding to the political friction, the Trump administration reportedly attempted to pressure Anthropic into pausing the release of Fable 5 before the Commerce Department ultimately issued the binding directive. Cybersecurity experts note that Anthropic's own marketing may have played a role; by heavily advertising the models' ability to rapidly convert disclosed software vulnerabilities into working exploits, the company effectively invited strict government scrutiny.[2][6]

Anthropic has promised to publish further technical details regarding the vulnerability and is working to restore access to the models as soon as possible. In the meantime, the broader technology sector is left grappling with a new reality where frontier AI models are treated with the same stringent, nationality-based controls as advanced military munitions.[2][3][5][6]