How Should Governments Regulate to Minimise the Negative Climate Impacts of Generative AI?
Author: Rory Hannigan
The below article is one of a series of blogs based on a Policy Brief shortlisted as a finalist for the 2026 Chronos Sustainability Prizes at the LSE.
Setting the Scene
GenAI's adoption is spreading rapidly across consumer and commercial markets. Within two months of its launch, ChatGPT had reached 100 million active users, and GenAI attracted $33.9 billion in global private funding in 2024. Environmental regulation has not evolved at the pace of GenAI's growth. Existing frameworks, including the EU AI Act, energy efficiency directives, and national reporting requirements, were not designed to address AI-specific externalities.
This surge of investment in, and usage of, GenAI carries significant environmental impacts. The Policy Brief identified four pathways through which the growth in GenAI creates significant environmental impacts:
Through electricity-related greenhouse gas emissions from AI training and inference.
Through upstream resource extraction for AI hardware.
Through local environmental pressures from data-centre operations, particularly water use.
Through downstream impacts from AI-enabled applications in high-emitting sectors such as oil and gas.
Key Findings
Not all of these risks are suitable for domestic intervention. Upstream supply-chain impacts require international coordination, while downstream emissions depend primarily on sectoral activity and are better addressed through industry-specific climate policy.
Effective governance at the domestic level therefore requires focusing on those impacts and risks that are geographically fixed, infrastructure-linked and subject to national regulatory authority. This leaves three priority issues:
The rapid growth in AI-driven electricity demand;
Localised water stress from data-centre siting and cooling; and
The absence of standardised measurement, reporting, and verification (MRV) for AI-related environmental impacts.
Key Recommendations
The brief recommends three priority actions for national governments:
Establish mandatory and standardised MRV for AI processes and data-centre operations. This would require operators to report predicted (for proposed facilities) and actual (for existing facilities) annual and peak electricity use, carbon intensity, water withdrawals, cooling method, and installed compute capacity. Transparent MRV is a prerequisite for effective grid planning, water management, and alignment with decarbonisation targets.
Integrate projections for AI-driven electricity and water demand into infrastructure planning and decarbonisation strategies. This would involve attributing demand to AI training and inference workloads rather than data-centre capacity alone, and adopting siting criteria that direct new infrastructure toward regions with sufficient grid capacity and water availability.
Use targeted standards for cooling and water use at large-scale AI infrastructure. Stress-based permitting and cooling-technology standards can mitigate cumulative pressure on local water systems, while avoiding blanket restrictions on AI deployment
The trajectory of GenAI remains uncertain. Regardless of whether this wave represents a lasting technological shift or a speculative bubble, the infrastructure supporting it will remain capital-intensive and tightly tied to national energy and water systems. Given the economic importance of GenAI, governments should design policies and regulations that support and enable the development of AI-related infrastructure, while shape its trajectory in a way that effectively manages and mitigates environmentally harmful outcomes.
--
Notes
Rory Hannigan's policy brief titled "How Should Governments Regulate to Minimise the Negative Climate Impacts of GenAI?" was shortlisted for the Chronos Sustainability Postgraduate Prize 2026. The full policy brief can be downloaded here.Rory is completing his MSc in Environmental Policy and Regulation at the LSE, with a focus on sustainable finance and the climate transition. He previously studied law, where his research focused on carbon markets, and has experience working in sustainable finance consultancy.
Read more on the Chronos Sustainability Prizes at the LSE here
