Anthropic Disables Fable 5 and Mythos 5 After US Government Security Directive
AI company Anthropic announced late Friday that it has disabled access to its two most advanced models, Fable 5 and Mythos 5, after receiving a government directive citing national security concerns. In a blog post, the company said it received a letter from the US government at 5:21 p.m. ET on Friday instructing it to suspend access to both models for any foreign national, whether located inside or outside the United States, including the company's own foreign-national employees. Because Anthropic said it cannot reliably separate foreign users from its broader customer base in real time, the practical effect is a worldwide shutoff of both models, while the company's other models, including its widely used Claude chatbot, remain unaffected.
The move marks what several outlets described as the first time a leading AI company has taken a publicly deployed frontier model offline as a direct result of US government intervention. Fable 5 had only recently been released, with Anthropic touting it as state-of-the-art across a range of industry benchmarks and noting it was the first time the company had made such an advanced model broadly available, citing new safeguards designed to block responses in high-risk categories. Mythos 5 builds on the earlier Claude Mythos Preview, which had drawn attention from Wall Street and government officials earlier this year for its cybersecurity capabilities and had been rolled out to a limited group of organizations under a cybersecurity-focused initiative.
According to Anthropic, the government's letter did not specify the exact national security concern behind the order, but the company said its understanding is that authorities had become aware of a technique for bypassing, or "jailbreaking," Fable 5. After reviewing a demonstration of the technique, Anthropic said the vulnerabilities it exposed were relatively simple and had already been identified using publicly available models. The company said it is complying with the directive while disagreeing with the underlying rationale, arguing that a narrow potential jailbreak should not be grounds for pulling a commercial model used by hundreds of millions of people, and warning that applying such a standard industry-wide could effectively halt new model deployments across all frontier AI developers.
Anthropic said it believes the situation may stem from a misunderstanding and is working to restore access to both models as quickly as possible, while apologizing to affected customers for the disruption. The company reiterated its long-standing position that governments should retain the ability to block unsafe AI deployments, but argued that any such process needs to be transparent, fair, and grounded in technical facts — standards it said this particular action did not meet. The episode adds a new dimension to the increasingly tangled relationship between Washington and the AI industry, coming in the same week that reports emerged of major banks testing Anthropic's most advanced systems internally as part of a broader push toward enterprise AI adoption.
