OpenAI、Anthropic、Google DeepMind 等 AI 巨頭將出席 G7 峰會

OpenAI, Anthropic, Google DeepMind, and other leading global AI companies’ executives will attend this week’s G7 summit held in France. OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Google DeepMind CEO Demis Hassabis, along with about a dozen technology industry executives, will attend a lunch meeting during the Evian Summit on Wednesday to discuss cutting-edge AI risks, infrastructure, and sovereignty issues.
(Background: The strongest AI from Anthropic is restricted from export by the United States! Fable 5, Mythos 5 go offline globally)
(Additional background: Can Claude Fable 5 be unlocked? Presidential adviser: Fixing one issue can lift restrictions)

Key Takeaways

• The three major CEOs of OpenAI, Anthropic, and Google DeepMind, plus about a dozen other tech executives, will attend a G7 Evian Summit lunch in France on Wednesday
• The lunch will cover cutting-edge AI risks, infrastructure, sovereignty, and children’s online protection, while the G7 will also discuss a “Trustworthy Partner” AI framework in parallel
• The backdrop is that in June, the United States imposed export controls on Anthropic’s Fable 5 and Mythos 5 on national security grounds, and negotiations between the two sides are still ongoing

The leaders of the three major frontier AI companies worldwide are set to sit down together at a G7 lunch table this week. According to CNBC, OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Google DeepMind CEO Demis Hassabis, along with about a dozen other tech executives, will attend a lunch meeting at the G7 summit in Evian, France, on Wednesday.

The meeting is expected to discuss cutting-edge AI risks, infrastructure, and sovereignty issues. The French Élysée Palace previously said that children’s online protection will also be one of the key topics. In addition to the three major labs, the attendee list includes Arthur Mensch of Mistral AI, Aidan Gomez of Cohere, Meta’s Alex Wang, and Salesforce CEO Marc Benioff, among others.

AI giants become negotiation power players

Jessica Brandt, a senior research fellow for technology and national security at the Council on Foreign Relations (CFR), told CNBC that if heads of state today want to make credible commitments on AI issues, they need cooperation from a small number of private-sector executives who are actually building this technology—possibly even securing their endorsement. She said, “We are seeing a shift in who can sit at the negotiation table, which also sends a signal about where power lies.”

The ability for AI companies to participate in G7 negotiations is, in itself, a symbol of geopolitical influence. At the same time, Europe is also seeking to add some checks and balances to the United States’ AI leadership.

The U.S. “export controls” are not lifted yet

As the G7 summit gets underway, Anthropic is still negotiating with the U.S. government over export controls on its two models, Fable 5 and Mythos 5. Previously, the U.S. demanded a pause on all foreign access to these two models on the grounds of national security, citing concerns that the models could be jailbroken and used to uncover software vulnerabilities and facilitate cyberattacks. This is also the first time the United States has used export control tools originally meant to regulate chips and weapons and applied them to a commercial AI model that has been used by hundreds of millions of people.

Emerson Brooking, a senior fellow at the Atlantic Council, said the U.S. export controls on Anthropic’s models “changed everything,” because the U.S. has already shown that it is willing to cut off channels through which G7 countries—and even treaty allies—can obtain certain AI capabilities.

According to reports, G7 leaders are discussing a “Trustworthy Partner” framework, under which allies would be selectively granted permissions to access advanced models from U.S. companies. Canadian Prime Minister Mark Carney said that these restrictions highlight the dangers of excessive reliance on a small number of U.S. suppliers.

Frequently Asked Questions

Which AI executives will attend this G7 summit?

OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Google DeepMind CEO Demis Hassabis, along with about a dozen tech executives including Mistral, Cohere, Meta, Salesforce, and others, will attend a G7 lunch in Evian, France, on Wednesday to discuss frontier AI risks, infrastructure, and sovereignty.

Why does the U.S. impose export controls on Anthropic’s Fable 5 and Mythos 5?

In June, the U.S. government cited national security concerns, fearing the models could be jailbroken, used to uncover software vulnerabilities, and facilitate cyberattacks, and demanded a suspension of all foreign access—marking the first time the United States has used export control tools against a widely used commercial AI model that has been used by hundreds of millions of people.