Navigating justice in AI lifecycles: ethical perspectives on infrastructural prerequisites and their environmental impacts

Abstract The expectation that AI can help address climate change is increasingly tempered by concern over the environmental impacts of its production, operation, and disposal processes and their related infrastructures. Rising consumption of energy, water, and rare materials contributes to environmental strain and raises justice questions on multiple aspects of its lifecycle. This paper examines the environmental implications related to the infrastructural prerequisites of AI through three justice perspectives: Environmental Justice, Climate Justice, and Multispecies Justice. By bringing the three frameworks together, the paper highlights the multidimensional character of the environmental challenges associated with the infrastructural prerequisites of AI lifecycles and how they are entangled with social and ecological inequalities. Drawing on illustrative cases, the paper demonstrates how different justice lenses bring into view distinct socio-technical and environmental interdependencies, broadening ethical analysis by extending it beyond distributive concerns to issues of procedure, the recognition of marginalised communities, and even non-human entities. We show that examining these cases through multiple perspectives avoids the limitations of relying on any single framework and enables a more comprehensive understanding of both immediate and long-term impacts, including effects on non-human entities. This reveals how environmental impacts are inseparably entangled with uneven exposures, differentiated responsibilities, and vulnerability.