Anthropic just pulled the plug on its shiny new AI models, Fable 5 and Mythos 5, less than a week after launching them. The reason? The U.S. government told the startup to block foreign nationals from using them, citing national security. Anthropic's response: shut it all down for everyone.
The export-control order arrived at 5:21 p.m. ET on Friday, and it's sweeping. It covers "any foreign national, whether inside or outside the United States," including foreign-national employees at Anthropic itself. The company said it had to suspend access for all users to make sure it was following the rules. Other models are still running fine, it added.
Anthropic posted about the situation on X, saying the government's directive effectively forces it to block access for everyone, not just foreign nationals. The company called the move a "misunderstanding" and said it's working to restore access.
What's the Government Worried About?
The government didn't give a detailed explanation, but Anthropic thinks it's about a potential way to "jailbreak" Fable 5 — a technique to bypass the model's safety guardrails. Anthropic pushed back, saying the reported technique only revealed "a small number of previously known, minor vulnerabilities" and doesn't show a broad failure of safety protections.
"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," the company said.
This isn't the first time Anthropic and Washington have butted heads. Earlier this year, the Pentagon labeled the startup a "supply-chain risk" after it refused to loosen restrictions on military use of its models for autonomous weapons and domestic surveillance. Anthropic sued, claiming unlawful retaliation and free-speech violations. That legal fight is still ongoing.
Why Fable 5 Matters
Fable 5 and Mythos 5 were Anthropic's most advanced AI systems yet, unveiled just last week. Fable 5 was especially notable because it brought capabilities from the company's closely watched Mythos program to a broader audience. Before launch, Anthropic said the models went through thousands of hours of testing with government agencies, third-party researchers, and internal teams.
The company insists no researcher has found a "universal jailbreak" that could broadly bypass the models' safeguards. It also argues that the cybersecurity capabilities the government is worried about are already available in competing frontier AI systems.
For now, Anthropic is complying with the directive while trying to sort things out. It called the government's action not "transparent, fair, clear, and grounded in technical facts."
