World leaders gathered at the G7 summit in Evian-les-Bains, France are weighing a plan that would allow a select group of so-called “trusted partners” to gain access to top-tier American artificial intelligence models, three diplomatic sources revealed Tuesday.
The discussions follow a decision made last week by Washington to cut off foreign nationals from Anthropic’s most advanced AI models, a move the U.S. government justified on national security grounds.
According to one of the sources, a number of delegates raised the idea during informal conversations with U.S. representatives — primarily U.S. Commerce Secretary Howard Lutnick — on the sidelines of the G7’s opening dinner at the French lakeside resort.
A second source, who asked not to be identified because the discussions are still ongoing, said the “trusted partners” in question could include both countries and private companies.
A third source confirmed that no official statement regarding the proposed framework was anticipated Wednesday, even as technology-related issues were scheduled to be part of the day’s G7 agenda.
Cybersecurity experts have raised concerns about one specific Anthropic model called Mythos, which is designed to identify weaknesses in computer code. They warn it could significantly amplify attacks on banks’ technology systems. The model has not yet been made available to any European banks, and the European Union has been seeking access to Mythos to better understand its potential implications.
The Financial Times was first to report on the “trusted partners” proposal.
