The U.S. government has restricted foreign access to Anthropic's advanced artificial intelligence (AI) models and warned that overseas provision is prohibited without prior approval. This marks a direct intervention by the U.S. in the operations of AI models, indicating a tightening of the country's AI control policies.
According to a letter obtained by Bloomberg News, U.S. Commerce Secretary Howard Rutnik informed Anthropic CEO Dario Amodei on June 12 that individual approval from the Commerce Department is required to provide the 'Preamble 5' and 'Mythos 5' AI models to foreign entities.
Rutnik specified that the export, re-export, and domestic transfer of these models, as well as access for foreign individuals anywhere in the world, are subject to approval. He warned that failure to comply could result in civil and criminal penalties.
While the letter did not specify the reasons for the restrictions, it referenced U.S. laws that allow export controls on civilian technologies that could be used for intelligence activities by hostile nations.
Following the Commerce Department's directive, Anthropic ceased access to the two models late on June 12. Subsequently, company representatives discussed security issues with U.S. government officials, and technology staff reportedly met with Commerce Department representatives on June 15.
The Trump administration has not disclosed the specific background for this action. Anthropic believes the order was issued after the U.S. government identified vulnerabilities in Preamble 5 that could allow for circumvention of its safety measures.
Anthropic has pushed back against what it considers excessive government action. The company stated, "We do not agree with the decision to withdraw a commercial model used by hundreds of millions based on a narrow scope of potential safety circumvention. If this standard is applied across the industry, it could effectively halt the deployment of new models by all frontier AI companies."
This directive is seen as the strongest instance of U.S. government intervention in the distribution and access rights of AI models. By expanding controls to include cutting-edge AI models, alongside semiconductors and AI servers, U.S. export control policies are becoming a key variable in the global operation of AI services.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.