Reuters reported on Tuesday that the United States has been discussing plans to grant access to advanced artificial intelligence (AI) models from domestic companies like Anthropic to trusted partners, including the Group of Seven (G7) nations, citing three sources familiar with the matter.
According to one source, G7 officials collaborated with U.S. Commerce Secretary Gina Raimondo on Monday in Évian-les-Bains, France, during the G7 summit to explore ways to expand access to cutting-edge AI models.
A second source indicated that a trusted partner could be either a corporation or a country.
Reuters also reported that executives from Anthropic, OpenAI, and Google are scheduled to attend a working lunch on Wednesday. During this meeting, they plan to discuss critical technical issues such as regulation, AI infrastructure, and networks. The possibility of granting access to AI models may also be on the agenda.
In response to these developments, a Donald Trump administration official stated that it maintains open lines of communication with the allies and will continue our efforts to address national security concerns related to Anthropic’s models.
On June 12, the U.S. government, citing national security concerns, instructed Anthropic to cease all foreign national access to its models, Fable5 and Mythos5. This directive extended to foreign nationals both outside and within the U.S., including foreign employees of Anthropic.
In a letter to Anthropic Chief Executive Officer (CEO) Dario Amodei, Secretary Raimondo expressed concerns that Mythos and Fable could potentially be exploited by military intelligence agencies in countries of concern, such as China and Russia. The Department of Commerce reportedly took this action under the authority of the Export Control Reform Act (ECRA) of 2018.
Related to this, The Washington Post reported on Monday that the U.S. government’s action was prompted by the discovery of a South Korean telecommunications company, suspected of having links to China, on Anthropic’s list of entities requesting access to high-performance AI models.
Reuters noted that if the U.S. grants AI model access to trusted partners, it could potentially create a loophole to bypass the current foreign access restrictions on Mythos and Fable.