France’s effort to set up a G7 group for testing advanced AI models has stalled for now. Anne Le Hénanff, France’s digital minister, told a French Senate hearing on November 4 that the United States was not ready to join the other six members in a joint statement on the idea, according to reports of Politico Europe’s account. The outcome matters to companies that build or deploy advanced models. Shared testing or incident reporting among G7 governments would shape what they are asked to disclose, and to whom.
What France proposed
Le Hénanff said she held ministerial meetings in September and October to seek commitments to a high-level group. The group would test frontier models and share information about risks and incidents. She said Washington had asked the others to wait for the findings of its own assessment.
France issued a chair’s summary after a virtual meeting of G7 technology ministers on October 20. As reported, the summary said ministers had considered closer coordination on identifying risks, sharing incident information and creating a dedicated high-level group. It was not a joint statement endorsed by all members.
The U.S. position
A White House official, speaking anonymously to Politico, described the proposal as a poorly defined body that could create unnecessary bureaucracy. The official pointed instead to President Trump’s Super Intelligence Force, which was described as bringing government and industry together. The U.S. G20 presidency uses similar language. Its priorities refer to “pioneering innovations in SI and emerging technologies.”
Where the agreed G7 text stands
The last consensus G7 text on the subject is the digital ministers’ declaration of May 29. It recognizes that AI systems can be misused for cyberattacks or to develop chemical, biological or radiological capabilities. It calls for an innovation-friendly approach to those risks. It also welcomes the revised Hiroshima Process reporting framework, and ministers said they would start discussions with members of the International Network for Advanced AI Measurement, Evaluation and Science and with the OECD on comparing risk assessment frameworks.
The declaration does not commit members to joint testing or to sharing incident data. The French summary of the meeting says 56 companies and organizations stated their willingness to complete the Hiroshima reporting framework. For now, that voluntary reporting remains the shared G7 channel for companies.
Le Hénanff has proposed resuming talks in December and said France would continue the work with European partners, drawing on its national AI evaluation institute, INESIA, and the research institute Inria. The presidency’s agenda lists a November 9 conference in Paris on AI resilience through openness and cooperation, organized by the Paris Peace Forum, the Global Partnership on AI and Inria. France’s presidency ends with the year, and the May declaration notes that the United States holds the G7 presidency in 2027.
Sources
- G7 digital and technology ministerial declaration, May 29, 2026 (French G7 presidency, PDF)
- Direction générale des Entreprises: G7 digital ministers’ outcomes
- G7 Évian presidency agenda
- U.S. G20 presidency: priorities
- Mezha: France and US clash over G7 AI safety effort (citing Politico Europe)
- UA.News: US did not support France’s initiative on advanced AI safety (citing Politico Europe)
