Uniform AI rules, or the quiet coup of efficiency?
By Raymond Aron
One can understand the temptation: if fifty states legislate AI in fifty different idioms, innovation becomes a customs house, and the nation risks turning its own market into a labyrinth. The Trump Administration’s project, a federal framework that broadly preempts state law, is thus presented as minimally burdensome and strategically necessary. In a world of technological rivalry, the argument from efficiency is not frivolous.
But democracies are not administered like data centers. In regimes of multiple parties and divided sovereignties, the permanent problem is to reconcile constitutional restraint with democratic contestation, and then to reconcile democracy with governmental effectiveness. @Les régimes de partis multiples, héritiers des régimes constitutionnels ou libéraux, cherchent à maintenir les valeurs du libéralisme dans des sociétés démocratisées et sont travaillés par une double dialectique structurante : tension entre constitution et démocratie, et tension entre démocratie et efficacité gouvernementale. The new initiative is a textbook attempt to resolve the second tension by compressing the first.
What is striking is less the idea of a national standard than the method: executive order, agency policy statements, litigation task forces, and, above all, funding conditions. When broadband grants become a lever to discourage “onerous” state laws, we leave the serene domain of persuasion and enter that of administrative constraint. The modern state’s tasks expand faster than the habits of deliberation, and this expansion is the most serious reason many doubt the viability of pluralist constitutional regimes. @Après avoir écarté plusieurs arguments déterministes (usure démagogique, civilisation de masse, pression des intérêts, totalitarisation des partis), Aron estime que le seul argument de fond sérieux contre la viabilité des régimes constitutionnels‑pluralistes tient à l’ampleur croissante des « tâches de l’État moderne », qui pourraient dépasser les capacités d’un tel régime. The AI file illustrates it with almost pedagogical clarity.
The Administration preserves state authority for children’s safety, for compute and data center infrastructure, and for state procurement. This carve out is politically astute, but it also admits what the rhetoric denies: uniformity is impossible where vital interests, local conflicts, and public fears are concentrated. If one centralizes everything else, one risks producing the familiar spectacle of dispersed responsibility and contested legitimacy, with governors, courts, agencies, and firms each claiming to serve the public while paralyzing one another. @Le problème central des régimes constitutionnels‑pluralistes n’est pas seulement leur caractère oligarchique, mais aussi le risque structurel d’instabilité, de dispersion du pouvoir et d’impuissance gouvernementale, risque aggravé par la priorité donnée à la conservation du pouvoir, par la prolifération de freins institutionnels et par la dépendance des gouvernants à l’égard des groupes de pression, de la presse et des règles administratives.
A national AI framework may be necessary. Yet if it is born primarily from executive impatience, it will invite the very litigation and distrust it claims to prevent. What, precisely, will make this framework legitimate in the eyes of those who must obey it?