Restraint by Design: Rethinking Artificial Intelligence and Environmental Governance in an Age of Ecological Crisis

Abstract

Artificial intelligence is increasingly promoted as indispensable to addressing the climate and biodiversity crises, from optimizing energy grids to modelling deforestation. Yet the same computational infrastructure that promises environmental salvation generates substantial ecological costs of its own: escalating electricity demand, freshwater consumption for data-centre cooling, and resource-intensive hardware supply chains. This paper argues that the dominant “AI-for-climate” paradigm is constrained by an unexamined techno-solutionist logic that treats environmental problems as optimization problems awaiting sufficiently powerful models, while remaining largely silent on AI’s own material footprint and its tendency toward centralized, extractive control over natural systems. Engaging recent scholarship on AI’s environmental costs alongside classical Daoist ecological philosophy — particularly its account of non-coercive action and spontaneous natural patterning the paper develops an original interpretive framework of “restrained AI governance” that reorients environmental AI deployment away from maximal control toward proportionate, reversible, and ecologically situated intervention. The paper concludes with concrete governance implications for regulators, technologists, and environmental policymakers.