The G7 leaders’ summit hosted by Emmanuel Macron in France closes today after three days of meetings that included, for the first time, a working lunch with the three CEOs running the labs at the frontier of generative AI. Sam Altman of OpenAI, Dario Amodei of Anthropic, and Demis Hassabis of Google DeepMind were all there together, an invitation list that Bloomberg flagged on June 12 and that CNBC and Reuters confirmed on the ground over the weekend. The three rarely share a stage. Altman and Amodei famously avoided being photographed together at the India AI summit in February. Putting them next to each other in front of seven heads of state was a deliberate choice by the French presidency.

The European read on what this summit is for has been consistent since the Paris AI Action Summit in February 2025. Brussels and Paris want regulatory leverage over American frontier labs that the EU AI Act alone cannot deliver, because the Act regulates deployment in Europe but not training, weight access, or the underlying decisions about what gets shipped at all. Putting Altman and Hassabis in a room with Macron, Meloni, Merz, and Starmer, with Carney and Ishiba listening, is the maximum-pressure version of the diplomatic ask. The Europeans want commitments on pre-deployment notice, on sovereign access to weights for safety review, and on advertising-grade transparency about training data provenance. None of those have been agreed to in any binding form.

The American subtext is louder than the European one. Amodei flew to France while Claude Fable 5 and Mythos 5 are still offline at home under last week’s Commerce Department export-control directive. Altman arrived four days after OpenAI announced its Deployment Simulation pre-release method, which is the closest thing any US frontier lab has shipped to the kind of structured safety auditing the EU has been asking for. Hassabis arrived with Gemini 3.5 Pro still in limited Vertex preview, two weeks past the deadline Pichai set on the I/O stage in May. Each of the three has a different reason to want this particular meeting to be useful to them. None of those reasons line up.

The communique drops this afternoon. The actual measure of the summit will be whether the language around AI commits to anything specific, which historically G7 communiques on AI have not. The bigger measure is whether Macron got the photograph he wanted. He did.

g7francemacronopenaianthropicgoogle-deepmindsam-altmandario-amodeidemis-hassabisai-governanceeuropean-sovereignty