The modern constitution is Latour's term for the settlement that governs who may speak for what in modern societies. Scientists speak for Nature, with the authority of empirical objectivity. Politicians speak for Society, with the authority of democratic representation. Nature is the realm of facts, Society the realm of values. The constitution was never explicitly written or ratified, but it organizes how knowledge, power, and legitimacy are distributed across Western institutions. AI has broken it. The technology is simultaneously technical and social, factual and political, a hybrid whose governance demands deliberative structures that the constitution's purified categories cannot provide.
The constitution's stability depends on its two clauses working in tandem. Nature is presented as transcendent — existing independently of human politics, discovered but not constructed. Society is presented as immanent — constituted by human agreement, governed by human choice, open to deliberative revision. Scientists, as the authorized spokespersons for Nature, produce findings that political deliberation must accept as constraints. Politicians, as the authorized spokespersons for Society, produce laws that technical actors must operate within. The division of labor appears rational and necessary. It is also, Latour argued, a fiction that prevents the recognition of hybrid phenomena that cut across both domains.
Climate change was the hybrid that first forced the constitution into visible crisis. It is simultaneously a physical phenomenon (governed by atmospheric chemistry) and a political question (concerning distribution, responsibility, and action). Neither scientists nor politicians can legitimately speak for it alone. Climate governance requires deliberative structures that bring empirical understanding and democratic accountability into sustained confrontation — and the absence of such structures is why climate response has been so structurally inadequate for so long.
AI is the same kind of hybrid, operating at faster timescales and with more intimate effects on daily cognition. Its capabilities are technical facts about optimization and training. Its deployments are political choices about who benefits, whose perspectives are encoded, and what kinds of work the technology privileges. The current governance landscape reproduces the constitutional divide with remarkable fidelity: technical experts speak for the technology, legislators speak for society, and the hybrid gap between them — where the mechanism meets the effect — remains ungoverned. The EU AI Act addresses one side. Civil liberties advocates address the other. Neither addresses the hybrid.
The constitution's most consequential contemporary mechanism is the corporate governance of AI: frameworks like Constitutional AI that embed value choices directly into training processes. This is hybrid governance in practice — value choices made through technical procedures, by a small number of engineers at a handful of companies, affecting every user of the system. The constitution's categories cannot see this. It looks like technical work when examined from the political side; it looks like ethical guidance when examined from the technical side. The political reality — that value choices of civilizational consequence are being made inside training loops without democratic accountability — remains invisible to both.
The term was developed in We Have Never Been Modern (1991), drawing on Latour's historical reading of the 1660 debate between Robert Boyle and Thomas Hobbes about the proper distribution of authority between experimental science and sovereign political power. Boyle won the scientific half (facts belong to experimental demonstration); Hobbes won the political half (order belongs to sovereign authority). The settlement, Latour argued, constituted the modern division of knowledge and power that Western societies have inhabited ever since.
The analysis was developed further in Politics of Nature (2004), where Latour proposed an alternative constitution adequate to ecological politics — one in which the boundary between science and politics was not eliminated but reorganized so that hybrids could be governed through deliberative processes involving both empirical inquiry and democratic participation.
Tacit but enforced. The constitution operates without explicit articulation. Challenges to it are treated as category mistakes rather than legitimate political contests.
Boyle and Hobbes. The foundational settlement divided authority between experimental science and sovereign politics. Subsequent Western institutions inherited the division.
Hybrids as constitutional crisis. Climate change, pandemics, and AI are entities that refuse the division. Their governance requires new constitutional arrangements, not mere application of the old ones.
Corporate governance as quiet constitution. Constitutional AI and similar practices embed value choices in technical processes. This is already hybrid governance — just not democratically accountable hybrid governance.
Alternative constitution required. The path forward is not to abolish the division between science and politics but to redesign it so that hybrids can be deliberated by structures involving both kinds of authority and accountability.
Critics object that the modern constitution, whatever its philosophical defects, has produced remarkable achievements — scientific knowledge, democratic governance, technological capability — and that abandoning it risks losing those gains. Latour's reply is that he is not proposing abandonment. He is proposing extension: the current constitution handles pure domains reasonably well but cannot govern hybrids. What is required is a deliberative architecture that adds to rather than replaces existing structures, opening new governance spaces where the hybrids can be addressed on their own terms. Whether such an architecture can be built without corroding the effectiveness of the existing institutions remains the open political question of the post-2020 moment.