Privacy after Publicness: A Publicness-Inference-Power Theory of AI Governance
TL;DR
A Publicness-Inference-Power (PIP) theory of AI governance is developed, which rejects two opposing simplifications: that public disclosure extinguishes privacy, and that conventional data-protection rules adequately govern all downstream uses of public data.
Abstract
Privacy law and AI governance are often treated as adjacent but distinct regulatory domains. This separation becomes unstable when artificial intelligence systems transform publicly accessible or voluntarily disclosed information into sensitive attributes, predictions, rankings, and consequential decisions. This article develops a Publicness-Inference-Power (PIP) theory of AI governance. The theory rejects two opposing simplifications: that public disclosure extinguishes privacy, and that conventional data-protection rules adequately govern all downstream uses of public data. It argues instead that publicness changes the locus of regulatory concern. The principal risk moves from unauthorized access to the computational transformation of information into institutional power. The PIP model identifies five linked stages: publicness, aggregation, inference, decision, and power. At each stage, information acquires new meaning, creates new asymmetries, and may generate harms that cannot be explained by collection or disclosure alone. The article uses doctrinal and comparative analysis of the European Union General Data Protection Regulation, the EU Artificial Intelligence Act, India's Digital Personal Data Protection Act, the California Consumer Privacy Act, and major international AI-governance frameworks. It shows that current regimes contain fragments of an inference-oriented approach but do not yet provide a coherent governance architecture for derived data and consequential use. The article proposes six theoretical propositions and an operational governance model based on inference registers, contextual-purpose boundaries, provenance, validation, contestability, and action controls. The contribution is a shift from data-status governance - asking whether information is public or private - to transformation-and-power governance - asking what an AI system derives, how confidently it derives it, for what purpose, and with what effects on persons and institutions.