Trustworthy Artificial Intelligence in the Climate Transition: Mandating Life Cycle Accountability for Net-Zero Systems and Infrastructure Resilience

Abstract

Artificial Intelligence is increasingly becoming a major factor in the climate transition, but at the same time, its swiftly growing computational footprint might negate the net-zero gains if the impacts over its whole life cycle are not accounted for. The works of this paper propose the AI-Centric Green Accountability (AI-CGA) as a systems-level governance and assessment architecture that combines lifecycle carbon accounting, energy-aware model management, and climate-support decision-making within a trustworthy deployment of AI in climate matters. The paper draws on the present scenario of energy, transport, water, and urban systems and claims that AI-assisted prediction, optimization, and adaptive control can lead to a substantial reduction of operational emissions. However, the advantages of these technologies are being masked by unclear hardware supply chains, a lack of standard metrics for embodied carbon, compute intensity, or inference-scale impacts, and increasing training requirements. In order to fix these problems, the study sets forth a lifecycle accountability model that covers embodied emissions, operational compute, grid-linked carbon variability, and end-of-life burdens, along with verification pathways for policymakers and infrastructure operators. A research and implementation plan is presented, which includes the development of metrology, standardized reporting procedures, marginal grid-carbon integration, and multi-vector governance pilots that can demonstrate the net climate benefit. The findings provide a platform for AI-CGA that is expandable to regulating models with high-intensity computing, to making the alignment of AI innovation with the limitations of the planet, and ensure that climate-AI systems achieve verifiable decarbonization outcomes rather than shifting impacts to other areas of the value chain.