The emergence of data driven states marks a structural transformation in the relationship between power, institutions and public decision making. Political authority is no longer grounded exclusively in legal procedures, representation or coercive capacity, but increasingly in the systematic management of information flows, computational capability and the use of predictive models able to render social complexity readable and anticipatable. In this framework, artificial intelligence does not operate as a mere administrative tool, but as a cognitive infrastructure that redefines the ways in which power is legitimized and exercised.
Data assume a function that goes beyond decision support. They become a strategic infrastructure, comparable in importance to currency or energy. According to World Bank estimates, more than 70 percent of national governments now use advanced analytics systems for fiscal, health or infrastructure planning, while in high income countries public spending on data governance platforms is growing at an average annual rate exceeding 12 percent. Governing increasingly means possessing dynamic maps of society, real time indicators and simulations of collective behavior. Authority is thus exercised not only through decision, but through anticipation, transforming prediction into a form of preventive power.
This shift introduces a profound discontinuity in political judgment. Decisions based on statistical correlations tend to replace traditional causal reasoning. If a model shows that certain variables are systematically associated with an outcome, public action intervenes on that basis, even when causal chains remain opaque. During the pandemic, numerous governments adopted restrictive measures or allocated health resources on the basis of epidemiological models that few decision makers were able to interpret in detail. According to the OECD, in 2021 more than 60 percent of major emergency health decisions in member countries were supported by predictive models external to public administrations, developed by universities or private operators. Operational effectiveness increased, but at the cost of a growing distance between decision and understanding.
Politics, in this context, risks a progressive weakening of its deliberative dimension. When a choice appears as the technically inevitable outcome of a model, the space for dissent narrows. Contesting the decision means contesting data, algorithms and technical architectures that escape the ordinary language of public debate. The political decision maker thus tends to become a validator of technical outputs rather than a subject who openly assumes the risk of choice. Consensus is built through the promise of optimization, not through confrontation among alternative visions.
Max Weber described rationalization as the ambiguous destiny of modernity, capable of increasing efficiency while also trapping action within a purely instrumental logic. In data driven states, this intuition is renewed in computational form. Algorithmic rationality offers unprecedented power, but tends to marginalize political judgment as a normative act. Authority is legitimized because it appears rational and measurable, not because it is recognized as just or desirable. The distinction is subtle but structural, because it shifts the foundation of legitimacy from the political plane to the technical one.
The economic implications of this transformation are significant. States that base their governing capacity on data become increasingly dependent on the technological infrastructures that make them governable. Sovereignty acquires a computational dimension. According to the International Monetary Fund, more than 80 percent of cloud and advanced computing platforms used by European public administrations are provided by extra European operators. Those who control models, standards and computational architectures exercise an influence that precedes any formal decision, affecting spending priorities, industrial policies and macroeconomic risk management.
The corporate world offers an instructive parallel. Companies that have built their competitive advantage on intensive data use have achieved significant productivity gains, in some cases exceeding 20 percent according to McKinsey studies. However, the same analyses show that these organizations are particularly vulnerable when the context changes and the correlations on which models are based weaken. In such moments, human judgment becomes central again. Similarly, a state that relies excessively on prediction for its authority risks finding itself unarmed in the face of crises that fall outside historical patterns, such as geopolitical shocks or sudden social transformations.
Political authority in data driven states thus lies along a line of tension. On one side, the promise of more efficient governance, capable of allocating resources precisely and anticipating systemic problems. On the other, the risk of a de responsibility of power that takes refuge behind the apparent objectivity of numbers. The issue is not to reject data or artificial intelligence, but to reintegrate them within an institutional framework that recognizes the structural limits of prediction and preserves the space of judgment.
For public decision makers and economic leaders, this transformation suggests a convergent lesson. In a world governed by models, power does not derive only from the ability to predict, but from the ability to decide when prediction is not enough. Leadership, public and private, remains an act of assuming responsibility under conditions of uncertainty. If data constitute the new language of power, judgment remains its deep grammar, the element that allows language to produce meaning and not only results.
Global AI Observatory