AI governance: why alignment doesn't appear on its own
At the G7 in Évian, AI's own developers asked for rules to govern it. The deeper lesson: alignment doesn't appear on its own, and without shared spaces for dialogue no one takes responsibility.
What happens when AI's own developers are the ones asking for rules to govern it.
Last week this became clear at the G7 in Évian, where heads of state sat at the same table as the CEOs of the leading AI companies. That the topic was addressed in a formal session — on the same level as security, international trade or finance — signals that AI is already being treated as public policy.
Which isn't illogical, given the scope and risks AI brings: impact on learning, cybersecurity, financial stability, redesign of the labor market.
But the meeting was far from an orderly consensus. It was the developers themselves who asked for the framework — some calling for a US-led coalition to set the rules, while European countries sought checks on that same concentration of power. It's no small thing: this mirrors what we see on a smaller scale, at the sector and even organizational level. Even bringing together the actors with the most resources and decision-making power, alignment doesn't appear on its own.
And still, the meeting matters. Not for its ability to resolve anything, but because it makes something essential visible: without a shared space for dialogue and exchange, it's impossible to establish where there's agreement and, above all, where there isn't.
That's the starting point. Not the solution — the condition for beginning to see it.
The discomfort is that as long as there are no spaces for putting things in common, exchanging and aligning, no one will take responsibility. Especially in healthcare, where the ecosystem is fragmented but the patient who moves through it is just one.