The G7 summit in Évian-les-Bains wrapped its three-day program today with a working lunch dedicated to artificial intelligence. OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Google DeepMind CEO Demis Hassabis, and Mistral CEO Arthur Mensch sat with the assembled heads of state — the first time the three rival US frontier-lab chiefs have appeared at a G7 together, joined by roughly a dozen other tech executives. The session theme, set by host Emmanuel Macron, was "ensuring a safe, rapid and effective deployment of artificial intelligence." Macron closed the summit with a state dinner for Donald Trump at the Palace of Versailles.
The substance behind the photo op is a widening transatlantic split. European leaders used the summit to press for binding international rules on frontier models and to signal discomfort with how concentrated the AI industry has become in a handful of American companies. The US delegation, by contrast, arrived having recently published a National Policy Framework that proposes preempting state AI laws for three years and pulling federal funding from states whose laws it deems onerous. The working lunch produced no joint communiqué on AI, which itself is the news: the G7 cannot yet agree on what "safe deployment" obliges governments to do.
This is the first G7 since the 2024 Apulia summit at which AI has been the headline topic, and it lands at a moment when the run-rate revenues of the assembled labs — Anthropic alone is reportedly past $30 billion — are bigger than the GDPs of some of the smaller G7 economies. The pattern over the last twelve months has been clear: the labs get larger and faster, governments stage convenings, and the actual rulemaking happens in the EU AI Act timeline, in US state legislatures, and in a handful of executive orders. International coordination keeps slipping to the next summit.
For learners, the useful frame is to watch what governments do, not what they say at summits. The EU AI Act's August general-purpose-model deadline, the Colorado AI Act's June 30 effective date, the discussion draft of the Great American AI Act, and the federal preemption push are the policy events that will actually shape how AI products ship in the next year. Summit photos are signaling; statutes and regulations are the wiring.