The Global Landscape of Environmental AI Regulation: From the Cost of Reasoning to a Right to Green AI

Review

This paper examines how environmental regulation should adapt to the rapidly growing energy and resource footprint of generative and reasoning-based AI systems. It argues that current regulatory approaches focus mainly on facility-level disclosures and training costs, while largely ignoring inference-level impacts where much of the environmental burden may occur. Using comparative legal analysis across several jurisdictions, the authors highlight a global transparency gap and propose a governance framework centered on model-level disclosure, standardized reporting of inference costs, and user-facing rights to avoid unnecessary generative AI.

The paper’s central contribution is its insistence that AI environmental governance must move beyond datacenter reporting toward model-level transparency. The discussion of inference costs, especially for reasoning models, highlights an emerging regulatory blind spot. The comparative typology of regulatory approaches also provides a useful analytical scaffold for understanding how jurisdictions differ in ambition, transparency requirements, and enforcement mechanisms.

Methodologically, the paper functions primarily as normative legal-policy analysis supported by secondary technical evidence and case examples. This approach is appropriate for regulatory scholarship, but the empirical claims about environmental magnitude rely heavily on estimates and proxy calculations rather than direct measurements. While the authors acknowledge these limitations, some policy conclusions are presented with greater certainty than the underlying evidence supports.

The paper also raises ambitious normative proposals such as a “right to green AI” and a right to access digital infrastructure without unnecessary generative AI components. These ideas are intellectually interesting but require clearer definitions and operational boundaries to be practically implementable, particularly in hybrid systems where generative and non-generative functionality are intertwined.

Overall, the paper is timely and agenda-setting. Its central argument, that environmental AI regulation must incorporate inference-level transparency and model-level accountability, is compelling. Strengthening methodological clarity, moderating empirical claims, and further developing implementation pathways would significantly enhance its impact and suitability for publication.

Key Insight

Effective environmental governance of AI will require regulation to shift from facility-level reporting toward model-level transparency, especially around inference costs and reasoning-heavy systems.