Abstract
Artificial intelligence (AI) is widely portrayed as an innovative, productivity-enhancing digital technology reshaping economies and societies. This framing is incomplete. AI also generates environmental, economic, social, political and informational externalities whose costs are not fully reflected in market or organizational decisions. Drawing on lessons from climate, nuclear and biotechnology governance, we conceptualize AI as a material, infrastructural and institutional system whose rapid scaling intensifies energy and water demand, labour-market disruption, concentration of computational resources, and informational and political risks. We argue that the comparison with climate change is structural rather than literal. Both involve externalized costs, collective-action problems, intergenerational asymmetries and fragmented governance. Prevailing approaches centred on efficiency gains, voluntary disclosure and firm-level ethics are therefore insufficient where rebound effects, concentrated market power and cross-border externalities persist. We propose a three-pillar governance framework centred on measurement and internalization of externalities, accountability and market structure, and international coordination and institutional capacity. The framework integrates environmental sustainability, economic power and societal risks within a common political-economy approach to governing AI.
•AI creates systemic externalities analogous to climate change.•Efficiency gains trigger rebound effects and rising resource use.•Market-based and voluntary governance approaches are insufficient.•AI requires institutional design beyond firm-level optimization.•Three-pillar framework for sustainable and equitable AI governance.