The US government has ordered Anthropic to cut off foreign nationals from its two most advanced AI models, Fable 5 and Mythos 5. Rather than build country-by-country restrictions, Anthropic went further — disabling both models for all customers worldwide.
According to India Today, Anthropic said the directive cited national security concerns but provided no detailed explanation. The Decoder reports the order specifically invoked alleged "jailbreak" risks — vulnerabilities that could let users sidestep the models' safety guardrails. According to MSN, the U.S. Department of Defense had flagged Fable 5 and Mythos 5 as security risks prior to the shutdown.
Anthropics is complying but openly disputing the rationale. According to The Decoder, the company says the cited vulnerabilities are minor and are also present in rival systems, including GPT-5.5 — a pointed argument from a company that has built its brand around AI safety.
The tech world is divided. According to Business Insider, reactions range from calling the move "baffling" government overreach to praising it as a sound national-security call.
The original order targeted foreign nationals under export-control rules, but Anthropic's decision to extend the blackout to domestic users as well — described by theweek.in as a case where the US order "backfired" — left even American customers locked out of models they were already paying to use.
If this precedent holds, it means the US government can effectively kill a cloud AI product overnight for the entire planet, not just adversarial states — a power with sweeping implications for how AI companies build, price, and deploy their most capable systems.