US Bars Foreign Nationals From Anthropic's Top AI Models

Anthropic has shut down customer access to its advanced Fable 5 and Mythos 5 AI models after receiving a directive from the US government late on Friday. The order requires the company to suspend access for every foreign national, regardless of location or employment status, and it took effect immediately with no detailed explanation provided.

The directive came at 5:21 p.m. on June 12 and cited national security concerns without elaborating on specific risks. Anthropic stated that the measure applies even to its own employees who are not US citizens and to users both inside and outside the country. All other company models and its Claude chatbot remain unaffected.

Fable 5 was released publicly just days earlier on June 9. The model incorporates capabilities drawn from Mythos, a specialized cybersecurity system previously restricted to select Project Glasswing partners. During internal testing, Fable 5 demonstrated performance exceeding earlier Anthropic releases, including success against tasks where previous versions had fallen short. Mythos itself has been employed by US agencies and approved companies to identify long-hidden software vulnerabilities.

The government provided no public details on why the models suddenly required this level of restriction. Anthropic has indicated it believes the order may relate to reports of a jailbreaking technique that could bypass safeguards in Fable 5. The company described steps it had already taken to limit the model's cybersecurity and biotechnology functions before the directive arrived.

The abrupt cutoff has left researchers, developers, and staff members without notice or recourse. Foreign nationals working at Anthropic were required to stop using the systems at once, disrupting ongoing work that depends on these tools. The broad application to anyone outside US citizenship, including those legally employed in the United States, marks a significant expansion of export-style controls into everyday AI usage.

Observers note that Mythos technology has already proven useful for defensive security work, yet the same capabilities could be turned toward offensive purposes if misused. The decision to block access across the board rather than target specific high-risk users has drawn attention to the lack of transparency in how such orders are issued and enforced. Anthropic confirmed it had no choice but to comply fully to avoid penalties.

The episode underscores the growing tension between rapid AI development and government efforts to maintain control over sensitive technologies. With Fable 5 positioned as a more widely available version of Mythos capabilities, the sudden restriction raises questions about how future models will be released and who will be permitted to use them.