Anthropic, the San Francisco AI lab behind Claude, abruptly suspended access late Friday to its most capable models, Claude Fable 5 and Claude Mythos 5, after the U.S. government ordered the company to block use by foreign nationals. To comply, Anthropic temporarily disabled those models for everyone, even U.S. users, while its other Claude models remain online. The sudden move yanks a frontier model from public use just days after its broad release and drops national security rules squarely into the middle of the AI rollout.
In a statement, Anthropic said the government had issued an export control directive suspending all access to Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States, and that it must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. The company said it disagrees with how the government is handling the matter and is pressing for more details as it works to restore access. According to AP News, Anthropic received the directive Friday afternoon and said the Commerce Department had not yet provided the evidence behind its national security concern.
How Washington Framed The Move
Officials are describing the action as an export control measure meant to keep foreign nationals away from cutting-edge AI capabilities. As reported by The New York Times, the Commerce Department directed the restriction after receiving, according to people familiar with the matter, concerns about a potential jailbreak that could let a model identify software vulnerabilities. The decision is one of the most aggressive U.S. steps so far to limit the spread of frontier AI technology.
Why Anthropic Shut The Models For Everyone
Anthropic said the Commerce directive covered foreign nationals, whether they were inside the United States or abroad, a requirement that is notoriously hard to enforce in a shared cloud environment. Rather than try to sort users by nationality in real time, the company concluded that switching off the models entirely was the only practical way to comply. TechCrunch reports that Anthropic received the order at 5:21 p.m. ET and moved quickly to disable both models. Fable 5 had been released publicly on June 9 as a guarded, consumer-facing iteration of Mythos, according to The Guardian, which makes the window of public access especially short.
Local And Industry Impact
The cutoff hits Anthropic’s San Francisco headquarters directly and ripples through local teams, enterprise customers, and startups that had been testing the new models. Anthropic has been sharing Mythos class access through Project Glasswing, a vetted program that included roughly 40 organizations that maintain critical infrastructure, a fact the company described previously and that national reporting later confirmed. Hoodline offered an April look at Project Glasswing, and The New York Times noted the roughly 40 partners involved.
Legal Implications
The move leans on long-standing Commerce Department rules that treat the release of controlled technical information toU.S. persons as a “deemed export,” a legal category that can require licensing or outright prohibition. The Bureau of Industry and Security explains that disclosing technology or source code to a foreign person in the United States is treated as an export to that person’s most recent country of citizenship or residency. That framework gives regulators a way to clamp down on cutting-edge tools, but it also raises tough engineering and civil liberties questions about how companies should identify nationality and segment services. Bureau of Industry and Security guidance lays out the deemed export rules in detail.
What Comes Next
Anthropic said it disagrees with the directive and is asking for clarity while it works with regulators to restore access. The company has warned that applying this standard broadly could slow frontier model deployments across the industry. TechCrunch reports that Anthropic plans to publish more details and pursue technical fixes, but for now, customers will be routed to lower-capability Claude models. For San Francisco, the episode is a clear reminder that national security policy can rewrite local product roadmaps overnight.
