The Trump administration has partially lifted its ban on the use of Anthropic's next-generation artificial intelligence (AI) model, 'Claude Mitos 5', allowing certain U.S. companies and institutions to access it. However, restrictions on foreign companies, including Samsung Electronics and SK Hynix, remain in place, indicating a tightening of U.S.-centric AI technology controls.
According to reports from Reuters and other outlets on June 28, the U.S. Department of Commerce recently informed Anthropic in a confidential letter that the use of Mitos 5 would be permitted for specific companies and institutions within the United States.
U.S. Secretary of Commerce Gina Raimondo stated in the letter, "Sufficient security measures have been established to allow trusted specific partners to use Mitos 5." This effectively means that only U.S. companies and institutions deemed safe by the government will be granted exceptions for its use.
The letter also noted that the list of approved entities could be changed by the government at any time as needed. While the specific names of the companies were not disclosed, U.S. media speculate that around 100 American firms and research institutions may be included.
As a result of this decision, Mitos 5 will primarily be utilized by verified U.S. companies and institutions, while access for foreign firms is expected to remain restricted.
Consequently, domestic companies such as Samsung Electronics, SK Hynix, and SK Telecom are likely to be unable to use Mitos 5 for the time being. These companies had previously participated in Anthropic's global AI security collaboration program, 'Project Glasswing', which allowed them access to earlier versions of Mitos. However, following the U.S. government's implementation of export controls after the official model's release, their access has effectively been blocked.
Project Glasswing is a program that provides major global companies and institutions with pre-release AI models to verify vulnerabilities and safety in advance. It is reported that domestic companies, including Samsung Electronics, joined the project on June 2 to test the pre-release version of Mitos.
However, on June 12, the Trump administration announced export control guidelines that completely barred access to Mitos 5 and the consumer model 'Fable 5' for foreign nationals, citing national security concerns. This latest action represents a shift from the previous blanket ban, now allowing exceptions only for U.S. companies, reflecting an intention to manage cutting-edge AI technology domestically.
Concerns have been raised within the U.S. AI industry that such extensive export controls could weaken the competitiveness of the American AI ecosystem in the long run. With fewer opportunities for collaboration and technology validation among global companies, the spread of U.S. technology and market dominance may be limited.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.