G7 digital ministers strike limited AI child-safety deal

The G7 digital ministers’ statement is another sign that AI governance is shifting from abstract principles to concrete levers like child-safety rules and infrastructure constraints. For the first time in this forum, countries explicitly recognized that frontier AI build‑out is an energy problem as much as a software one, even if the agreement only nods at electricity consumption rather than setting hard caps or standards. That framing matters because data‑center power use and grid capacity are fast becoming binding constraints on how quickly large models can be trained and deployed at scale.

For the race to AGI, this deal is more of a directional marker than an immediate brake. It tells platforms they will be held to higher standards around minors—age verification, safety by design, and illegal content controls—but leaves the energy question deliberately vague, largely at U.S. insistence. Over time, however, child‑safety norms and energy accounting could evolve into de facto global baselines that every major AI company must design around. The combination of reputational risk and the threat of stricter follow‑on rules will push labs to bake safety, consent and efficiency more deeply into model design, deployment tooling, and their cloud strategies.