The White House has imposed export restrictions on Anthropic's AI models, Fable 5 and Mythos 5, citing security concerns. These restrictions were enacted partly due to suspicions that a group linked to China had accessed the Mythos 5 model, according to a source familiar with the situation. This development raises significant questions regarding national security risks associated with Chinese access to American AI technologies.
The concerns included an alleged method to bypass the models' security features, prompting discussions between the Trump administration and Anthropic's CEO, Dario Amodei. The administration's request for Anthropic to address the security issue was reportedly declined, leading to the restrictions being implemented.
David Sacks, the White House AI czar, stated that the decision to restrict access was made reluctantly after Anthropic prioritized the continued availability of the consumer model over safety. The restrictions prevent access to the models for customers outside the U.S. and for foreign nationals within the country.
Anthropic responded by blocking access to the models for all users to ensure compliance with the directive from Commerce Secretary Howard Lutnick. The Mythos model had previously been released on a limited basis through Project Glasswing, aimed at securing critical software from potential misuse.
Concerns were validated when researchers at Amazon reportedly managed to bypass the security features of the Mythos 5 model. This information was communicated to U.S. officials, prompting a swift response from the White House. Anthropic expressed disagreement with the administration's decision, arguing that the finding of a narrow potential jailbreak should not warrant the recall of a model used by millions.
Anthropic recently filed to go public, and the two models were released to the public shortly before the restrictions were imposed.