Tech CEOs Join G7 Leaders for Closed-Door AI Session in France

Today’s Focus

The G7 summit in France ended Tuesday with an unusual guest list. President Donald Trump and the leaders of Britain, France, Germany, Italy, Canada and Japan were joined by the chief executives of OpenAI, Anthropic, Google and other AI firms for a closed-door session on artificial intelligence, according to CNBC.

The meeting, hosted by French President Emmanuel Macron at a chateau outside Évian, focused on what the leaders’ communiqué called “trusted partners” access to advanced AI models, chips and compute, CNBC reported. C-SPAN carried Trump’s opening remarks at the session, in which he described the assembled CEOs as “the people building the future.”

Macron used the summit to push a “sovereign AI” agenda, arguing that European countries need their own compute capacity and model providers rather than relying entirely on U.S. firms, according to France 24. He invited Trump to a follow-on dinner at the Palace of Versailles after the summit closed.

Cohere CEO Aidan Gomez, in an interview with Fortune on the sidelines, framed the choice facing G7 governments as “sovereign AI or digital serfdom,” arguing that countries without domestic AI capacity will end up dependent on a handful of foreign vendors.

The leaders’ statement did not commit to specific export-control changes, but officials told NPR that the U.S. is weighing looser chip-export rules for allied governments that sign on to shared security standards. Ukraine, Iran and trade tensions dominated the rest of the summit agenda, The Guardian reported, with AI emerging as the surprise headline on the final day.

The Debate

Supporters argue

Backers of the joint session say it formalizes a partnership that already exists in practice. Macron, in remarks reported by France 24, said democracies “cannot regulate what we do not understand” and that bringing AI leaders into the room with heads of state is how serious policy gets made.

OpenAI CEO Sam Altman, quoted by CNBC, said the meeting was “a signal of where power sits” in 2026 and argued that aligning Western governments and frontier labs is the only way to prevent authoritarian regimes from setting global AI norms. Anthropic’s Dario Amodei told the same outlet that coordinated export rules would help safety-focused U.S. labs compete with Chinese rivals.

Trump, per C-SPAN’s recording, said American companies “are going to dominate AI” and that loosening chip-export rules for allies would strengthen, not weaken, U.S. leverage. Cohere’s Gomez argued in Fortune that without sovereign capacity, mid-sized democracies will lose the ability to set their own rules on speech, surveillance and defense uses of AI.

Critics argue

Critics say the optics of tech CEOs sitting beside heads of state at a G7 summit confirm fears about corporate capture of public policy. Sen. Elizabeth Warren (D-MA), in a statement reported by NPR, said the session showed “the people who profit from AI writing the rules for AI.”

European digital-rights group EDRi told The Guardian that no civil society organizations, labor unions or independent researchers were invited to the closed-door discussion, calling the format “a lobbying meeting dressed up as diplomacy.” The group warned that “trusted partner” frameworks tend to lock in incumbents and freeze out smaller competitors.

German Green MEP Alexandra Geese, quoted by France 24, said Macron’s “sovereign AI” pitch risks becoming a subsidy program for a few European champions while doing little for citizens’ rights. She pointed to the EU AI Act’s enforcement gaps as the more pressing issue.

Some Republican lawmakers also pushed back. Sen. Josh Hawley (R-MO) told Fox News that he opposes any loosening of chip-export controls, arguing that advanced semiconductors are “the oil of the 21st century” and should not be shared even with treaty allies without strict conditions.

What the experts say

Independent researchers say the summit reflects a real shift in how AI policy is being made, but caution that the evidence on “sovereign AI” is thin. A 2025 Stanford AI Index report found that 73 percent of notable foundation models released that year came from U.S.-based organizations, with China second at 15 percent and the entire EU at under 4 percent, underscoring the concentration Macron cited.

Helen Toner of Georgetown’s Center for Security and Emerging Technology told NPR that export-control regimes work only when allies coordinate enforcement, and that past U.S. chip rules have leaked through third countries. She said any new “trusted partner” tier would need verification mechanisms that do not yet exist.

Brookings senior fellow Tom Wheeler, a former FCC chair, wrote in a recent analysis that direct CEO-to-leader meetings can speed coordination but tend to underweight downstream harms documented by academic researchers, including labor displacement and election-related disinformation. The OECD’s 2025 AI Policy Observatory tracked 1,100 national AI policies across 70 countries, finding that fewer than 10 percent included binding measures on competition or worker protections.

By the Numbers

7: G7 member countries whose leaders attended the AI session, plus the European Union, according to CNBC.
73%: share of notable foundation models released in 2025 by U.S.-based organizations, per the Stanford AI Index.
Under 4%: EU share of notable foundation models in 2025, per the same Stanford report.
1,100: national AI policies tracked across 70 countries by the OECD AI Policy Observatory in 2025.
Under 10%: share of those policies with binding competition or worker-protection measures, per the OECD.
$500 billion: total announced 2026 capital expenditure on AI infrastructure by the largest U.S. hyperscalers, as reported by Fortune.
15%: China’s share of notable foundation models in 2025, per the Stanford AI Index.