Future TechnologyThe bosses of Anthropic and Google DeepMind pitched G7 leaders on a shared AI rulebook
Key takeaways
- Anthropic CEO Dario Amodei and Google DeepMind CEO Demis Hassabis held a closed-door session at the G7 summit in France
- They pitched a US-led coalition of allied nations to coordinate AI development
- The proposal focused on shared safety standards, joint testing infrastructure and coordinated export controls
- It signals the AI race is now as much about government policy as it is about products
Two of the most powerful people in AI spent part of the G7 summit in France making a pitch to world leaders. Anthropic's Dario Amodei and Google DeepMind's Demis Hassabis held a closed-door session arguing for a US-led coalition of allied countries to coordinate how advanced AI gets built and tested. The focus, according to reports, was on shared safety standards, joint testing infrastructure and coordinated export controls.
Strip away the summit setting and the idea is straightforward. Right now every major country is writing its own AI rules, and the labs have to navigate a patchwork. A coordinated approach among allies would mean common safety tests, shared facilities to run them, and a united line on which chips and models can be exported where.
Why the labs want this
It is worth being clear-eyed about the motive. When the companies building the most powerful AI ask governments to write coordinated rules, they are also shaping rules they will have to live under. Get a seat at the table early and you help define what "safe" means, which tests count, and where the export lines fall. That is good positioning, and it can be genuinely useful at the same time.
Why it matters
This is the clearest sign yet that the AI race has moved beyond products. It now touches chips, electricity, cloud capacity and national security, and governments are writing the rules that decide who benefits. A coordinated allied bloc would be a big deal: it would set the global default for AI governance the way earlier coalitions shaped trade and telecoms.
The hard part is that coordination sounds simple and almost never is. Allies disagree on export controls, on how tightly to regulate, and on who pays for shared testing. A closed-door pitch is a long way from a signed agreement. But the direction of travel is set. The next phase of AI will be argued over in summit rooms as much as in research labs, and the people building the technology intend to be in both.