Leaders of the Group of Seven (G7) and chief executive officers (CEOs) of global artificial intelligence (AI) corporations on the 17th (local time) agreed to jointly regulate top-tier "frontier models" to block China's pursuit in AI. Frontier models are the most advanced AI humanity has built, powerful enough to be used for cyberattacks and even weapons development. Anthropic's "Fable" and "Mythos," which the United States recently cut off access to for security reasons, are representative examples. A strategy of containment against China that had been limited to advanced weapons and strategic materials is spreading into the AI domain.
On the 18th (local time), U.S. President Donald Trump, French President Emmanuel Macron, and other G7 leaders met for more than two hours with 11 AI corporations' CEOs and agreed to set standards to regulate the most advanced AI models. Victor Riparbelli, CEO of Synthesia, the world's largest AI video production platform, told Politico after the meeting, "Everyone agreed that it is important for the G7 and allied countries to join forces to win the AI race and prevent authoritarian states from gaining the upper hand." From the AI industry, attendees included OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, DeepMind CEO Demis Hassabis, and Arthur Mensch, CEO of French AI corporation Mistral.
According to participants, China came up repeatedly throughout the meeting. Politico, citing participant memos, reported, "While attendees agreed that U.S. corporations are taking the early lead in AI, they identified China as a real competitor," adding that "they noted China has already secured the massive electricity needed to run the most powerful AI models."
President Macron told local reporters after the meeting, "We need better regulation to keep frontier models out of the hands of an authoritarian regime." He added, "A small number of democracies will cooperate to set standards and create a platform for discussion." European Commission President Ursula von der Leyen also said, "I believe Europe and the United States must cooperate in AI," adding, "Both sides share security interests and have a responsibility to lead together."
CEO Altman proposed creating an international forum to practically evaluate and test AI models. Chris Lehane, OpenAI's global lead, said after the meeting that "there was a consensus among leaders and CEOs on establishing a forum." The G7 plans to flesh out frontier model regulatory standards at a Minister-level meeting in September.
Because most top-tier AI labs are concentrated in the United States, there was considerable wariness that the United States would monopolize the standards. Experts worried that U.S.-led regulation could lead to AI technology dependence across allied countries. The European Union is expanding AI-related technology cooperation with countries including Korea and Brazil to reduce reliance on the United States. Calls to accelerate development of homegrown advanced AI models are also growing, centered in Europe. Arthur Mensch of France's Mistral said, "It is true that the United States' stock line is 'trust us, we will set the standards here and then export them to you.'"
This agreement also came five days after the United States cut off access to frontier-grade AI models for allied countries. On the 13th, the U.S. government ordered Anthropic to restrict access to the latest models Fable 5 and Mythos 5 on security grounds. As a result, citizens of the G7 outside the United States were prevented from using the latest AI models by a decision of the U.S. administration rather than their own governments, and allied countries pushed back strongly.
However, multiple officials who attended the meeting told Politico that "leaders and CEOs, facing China as a common adversary, were in a cooperative mood regardless of the Anthropic situation." After a luncheon, President Trump said, "We had an excellent meeting on AI," adding, "Talks with Anthropic are also going well."