Digital ambition under ecological constraints: Governing AI in climate-vulnerable South Africa

Abstract

Artificial intelligence (AI) is promoted as a driver of development and sustainability. However, its deployment remains embedded in energy-, water-, material-, and land-intensive infrastructures operating under significant ecological and resource constraints. Drawing on environmental policy integration, sustainability transitions, socio-metabolic theory, and socio-ecological-technical systems (SETS) research, the study examines the extent to which environmental considerations are integrated into AI governance. Using qualitative comparative policy analysis, complemented by contextual documentary evidence on South Africa's emerging digital infrastructure, the study identifies a structural governance gap: AI policy has evolved largely in parallel with environmental and infrastructure governance systems. This misalignment limits the state's capacity to anticipate resource pressures, manage infrastructural lock-in, and steer AI deployment toward sustainable trajectories. Rather than proposing new regulatory regimes, the article advances a transition-oriented Green AI governance framework that leverages existing mechanisms for environmental screening and authorisation, sustainable public procurement, infrastructure and resource-capacity planning, resource transparency and disclosure, and institutional coordination. The South African case highlights a broader challenge facing climate-vulnerable contexts, where digital transformation risks outpacing governance capacity. The findings contribute to sustainability transitions scholarship by demonstrating how emerging digital technologies can be incrementally embedded within existing governance systems to align innovation with biophysical limits.