Anthropic announced it would “abruptly disable” its cutting-edge AI model for all users after the U.S. government ordered it to suspend foreigners’ access to the model, citing national security concerns.
Anthropic said in a statement that the company has received an export control order suspending access to Fable 5 and Mythos 5 to all foreign nationals, without specifying any specific national security concerns.
Anthropic said it understands that the government believes there are ways to circumvent, or “jailbreak”, safeguards that prevent Fable 5 from being used to identify software vulnerabilities.
The order comes just as earlier disputes between Trump administration officials and Anthropic, which is pursuing an IPO, were showing signs of easing across branches of the U.S. government.
Anthropic’s relationship with the government broke down this year after the U.S. military wouldn’t allow its AI models to be used for domestic surveillance or fully autonomous weapons systems. The government has responded by placing Anthropic on a supply chain blacklist, which is expected to come into effect later this year.
The action also marks a significant escalation in U.S. efforts to thwart the AI capabilities of foreign adversaries. For years, U.S. export controls have focused on the chips and tools that power AI, rather than restricting foreign access to the AI itself.
Anthropic said the government has only given “oral evidence of a limited and non-universal escape possibility.”
“We do not agree that the discovery of a narrow jailbreak possibility should be cause for a recall of a commercial model that has been deployed to hundreds of millions of people,” the company said in a statement.
The government directive and Anthropic’s response highlight growing tensions between AI developers and regulators over how to assess risks from so-called “jailbreaking,” or methods used to circumvent safety measures in models.
As recently as Wednesday, Anthropic was calling for increased U.S. oversight of AI, including the ability to block models with unacceptable risks. But it said the government’s actions on Friday did not follow the principles of fair and fact-based regulation.
Kirsten Davis, the Pentagon’s chief information officer, said in a post on X that the Pentagon supports national security priorities.
“Sometimes it’s simply more important than revenue cycles, clickbait and pre-IPO valuations. America first. Always,” Davis said.
Anthropic secretly filed for an IPO in the US last month, putting it ahead of rival OpenAI in the race to enter the public market.
Earlier this week, Anthropic unveiled an AI model named Claude Fable 5 that represents a new layer of functionality that the company calls the “Mythos Class.” The model comes with guardrails that prohibit its use in dangerous areas such as cybersecurity, and some users have complained that it’s “too broad,” Anthropic said.
In the wrong hands, experts say the Mythos model could dramatically accelerate sophisticated cyberattacks, especially in sectors such as banking that rely on complex, interconnected, and often decades-old technology systems.
Anthropic said it worked with the U.S. government and others on safety ahead of Fable’s launch, and that models from competing AI providers have shown similar ability to find minor bugs in code.
“The ultimate impact of this order is that we must suddenly disable Fable 5 and Mythos 5 for all customers to ensure compliance. Access to all other Anthropic models is unaffected,” Anthropic said.
Anthropic said it believes there was a “misunderstanding” and is working to restore access to the model as soon as possible.
U.S. officials confirmed that the Department of Commerce has issued an export control directive that halts all foreign access to Fable 5 and Mythos 5.
