G7 Leaders and Tech CEOs Discuss AI Regulation Amidst US-China Competition

EVIAN-LES-BAINS — Gathered since Monday in Evian-les-Bains for the G7 summit, Western countries are attempting to find common ground on AI regulation amidst intense competition with China.

At the invitation of Emmanuel Macron, about a dozen tech leaders — including Demis Hassabis (Google Deepmind), Arthur Mensch (Mistral AI), Sam Altman (OpenAI), and Dario Amodei (Anthropic) — met on Wednesday for a working lunch on digital issues. On the agenda: cooperation in the artificial intelligence sector and securing this technology. Facing them were the leaders of the G7 member countries, notably including US President Donald Trump.

During this exchange, which lasted over two hours, everyone did their best to avoid directly addressing contentious issues — primarily, the suspension of two powerful models from the American AI company Anthropic, which has shaken the relationship between the US administration and its allies since last week.

The spirit was therefore one of cooperation, according to five guests at the lunch interviewed by POLITICO. The objective was to maintain a link with the United States by rallying around a common goal: confronting China.

Chinese competition was thus discussed several times during the exchange, according to notes taken by a participant at the lunch, who sees it as a sign of growing concern in the United States regarding the technological advances of its Asian adversary.

"Everyone knows it is essential that G7 countries and their allies unite — that we win the AI race and… that we do not let autocratic countries gain an advantage," Victor Riparbelli, CEO of the British company Synthesia, who was among the guests, told POLITICO.

The forced blocking of Anthropic's Mythos 5 and Fable 5 models, which followed Washington's request to reserve them for US users only, was even considered by some as an additional opportunity to improve cooperation among allies on model security.

"We had an excellent exchange on AI," Donald Trump stated after the meeting, adding that ongoing discussions with Anthropic were "going well." This was despite months of tension with the company's CEO, Dario Amodei, over disagreements on the use of its powerful models.

Trump appeared relaxed at the beginning of this meeting, chatting with his table neighbor, OpenAI CEO Sam Altman. The US President also took the opportunity during his first remarks to make a long aside about the recently concluded agreement with Iran — but also to mention his uncle, an engineer at the Massachusetts Institute of Technology (MIT), according to two participants in the discussion.

The suspension of Anthropic's models was not directly addressed, Arthur Mensch, CEO of Mistral AI, indicated during a phone call with POLITICO.

The issue "was rather approached from the angle of supply chain interdependencies. I recalled that maintaining this interdependence is very important, as it will be the vector of stability for the new global economy that is emerging," he specified.

Several leaders took this opportunity to promote the idea of common standards in the development of frontier AI models, meaning the most advanced ones. The G7 had already taken initiatives in this direction in recent years, notably with the preparation of a code of conduct in 2023, under Japan's presidency.

"We must succeed in regulating them better to prevent them from falling into the hands of authoritarian regimes or people who could threaten our cybersecurity or our societies," Emmanuel Macron emphasized after the meeting, warning that "the answer cannot be non-cooperation between democracies."

Now, Emmanuel Macron wants to go further. The French president announced that the G7 would "build in the coming months a platform for discussions and cooperation among a few democracies in the face of the risk of artificial intelligence" in order to "jointly define common standards."

A proposal that echoes the repeated calls from the European Union for increased cooperation in artificial intelligence. The bloc has thus refrained from fanning the flames amidst tensions between Anthropic and the US government, which has de facto excluded European users from these advanced models. The EU's strategy has been to call for closer work with the United States.

Ursula von der Leyen, also present at the table, reiterated this on Wednesday. "I think Europe and the United States must work together on AI," the Commission President stated, praising "common interests in security and a shared responsibility to show the way." This is despite the fact that European AI regulation already imposes model oversight obligations in exchange with regulators.

Heads of state are not the only ones to have seized the opportunity to call for common standards. OpenAI CEO Sam Altman made the same argument, calling for the formation of an international forum on the subject.

"A consensus has emerged among the leaders and executives present in the room to establish such a forum," Chris Lehane, director of international affairs at OpenAI, said during a briefing with journalists after the meeting. For Victor Riparbelli, CEO of the British startup Synthesia, this is the path that would have "the most impact" but would also be "the most realistic way to reach a consensus."

However, some are skeptical, first and foremost Arthur Mensch, who fears that the United States will actually be the only ones driving these common standards. "I think we should be very wary of it because if there is one way to establish oneself and become unshakeable in countries other than the United States, it is to own the standards and endorse them [implement them, editor's note]," believes the CEO of Mistral AI.

"American actors tend to push this narrative: trust us, we will establish standards in our country and we will export them to you," he concludes.