Leaders from the G7 nations have engaged in discussions regarding a proposal that would allow certain "trusted partners" to access advanced artificial intelligence models developed by U.S. companies like Anthropic, according to three diplomatic sources. This move could potentially provide a workaround for restrictions on the use of these technologies by non-American entities.
Recently, Anthropic took the step to discontinue access to its most sophisticated AI models, Fable 5 and Mythos 5, following directives from U.S. President Donald Trump that mandated the company to block foreign nationals due to national security issues.
During the G7 summit in Evian-les-Bains, France, representatives from various countries—including U.S. Commerce Secretary Howard Lutnick—discussed expanding access to these advanced AI models. The concept of "trusted partners" may encompass both countries and private companies, as noted by one source who wished to remain anonymous due to the ongoing discussions.
If an agreement is reached, it could enable G7 nations to utilize these models for enhancing their cybersecurity measures against adversaries, particularly China. A representative from the Trump administration expressed that the president's team is maintaining open lines of communication with allies and is committed to addressing national security issues linked to Anthropic’s technology.
AI leaders from Anthropic, OpenAI, and Google, all currently at the forefront of AI development, are anticipated to converge at a working lunch on Wednesday to discuss regulatory frameworks, AI infrastructure, and networking technologies.
Anthropic has not yet commented on this development. Cybersecurity specialists warn that Anthropic's Mythos model, which identifies vulnerabilities in computer code, could elevate the risk of cyberattacks on banking systems. Additionally, the European Union is reportedly interested in accessing Mythos to evaluate its effects.
Before the implementation of Trump's order, Anthropic had allowed select organizations in over 15 countries to use Mythos for assessing their systems for potential security threats. This included entities from various sectors such as healthcare, communications, and utilities.
The Financial Times was the first to report on the idea of "trusted partners."
