Anthropic announced it will abruptly disable its most advanced AI models for all users after the US government ordered the company to suspend access for foreign nationals, citing national security concerns. The export control directive, received without specific details on the threat, targets the company’s Fable 5 and Mythos 5 models.
According to Anthropic, the government believes there is a method to bypass safeguards in Fable 5, potentially allowing its use in identifying software vulnerabilities. The company called the evidence “verbal” and described the potential jailbreak as narrow and non-universal. “We disagree that finding a narrow potential jailbreak warrants recalling a commercial model deployed to hundreds of millions of people,” Anthropic said in a statement.
The order arrives as tensions between Anthropic and the Trump administration were reportedly easing. Earlier this year, the company refused to allow US military use of its models for domestic surveillance and autonomous weapons, leading to a supply chain blacklist set to take effect later this year. The new directive marks an escalation in US efforts to control AI access, moving beyond chip and tool restrictions to directly limit foreign use of the technology itself.
Anthropic had recently called for greater US oversight of AI, including the ability to block high-risk models. But it criticized Friday’s action as lacking fair, fact-based regulation. The Pentagon’s chief information officer, Kirsten Davies, defended the move on X: “Some things are simply more important than revenue cycles, clickbait and pre-IPO valuation. America First. Always.”
Anthropic, which confidentially filed for a US IPO last month, said it must disable Fable 5 and Mythos 5 for all customers to comply. Access to other models remains unaffected. The company believes there is a “misunderstanding” and is working to restore access quickly.
Earlier this week, Anthropic launched Claude Fable 5, a Mythos-class model with guardrails barring use in sensitive areas like cybersecurity—a restriction some users called overly broad. Experts warn that Mythos models, if misused, could accelerate sophisticated cyber-attacks, particularly in sectors like banking with complex legacy systems. Anthropic noted that rival models show similar abilities to find minor code bugs and that it had collaborated with the US government on safety before launch.
Source: The Guardian