AI can help forecast renewable energy, optimize power grids, and detect ecological hazards before they become disasters, but only if the world agrees on how to govern the technology responsibly. A new report from the Columbia Center on Sustainable Investment (CCSI) and Hitachi's Research and Development Group finds that while AI holds genuine transformative potential for addressing climate and sustainability challenges, our ability to govern this technology has not kept pace with its rapid advancement.

Why Does AI Governance Matter More Than Raw Computing Power?

The report examined five domains critical to sustainable development: planetary environment, energy systems, industry and labor, finance, democracy, and societal resilience. Researchers conducted a systematic review of existing literature and interviewed leading experts from the International Labour Organization, University College London, the University of Oxford, the University of Waterloo, and the United Nations University.

The central finding is straightforward but sobering: AI's resource-intensive infrastructure is straining water supplies and accelerating electronic waste, while the technology itself remains opaque even to the engineers who build it. Unlike most technologies, the exact functioning of AI models operates as a "blackbox," making it difficult to predict or control outcomes.

"Technology is what we make it. Just as the Montreal Protocol showed that nations can agree on binding limits to protect a global commons, the report argues that AI demands the same kind of coordinated international action before the window to act closes," said Lara Fornabaio, lead researcher at CCSI.

Lara Fornabaio, Lead Researcher at Columbia Center on Sustainable Investment

The report also warns that operational gains in finance are unlikely to overcome deeper structural barriers that restrict capital flows, a distinction that is consistently obscured in mainstream AI optimism. Additionally, disparities in AI access risk deepening income inequality across nations and communities.

What Does a Global AI Governance Framework Look Like?

Rather than proposing a single solution, the CCSI and Hitachi report outlines a three-phase roadmap for international coordination. This approach recognizes that AI governance requires binding agreements similar to environmental treaties, but tailored to the unique challenges posed by rapidly evolving artificial intelligence.

• Phase One: Establish Scientific Baseline: Create a U.N.-mandated independent scientific panel to develop shared understanding of AI's capabilities and risks across all sectors, ensuring nations work from the same evidence base.
• Phase Two: Interim Safety Framework: Implement binding international restrictions on the most dangerous categories of AI research, similar to how the Montreal Protocol restricted ozone-depleting chemicals.
• Phase Three: Global Convention: Adopt a comprehensive framework convention on AI that sets universal obligations while allowing individual states flexibility to tailor implementation to their national priorities and circumstances.

Fornabaio emphasized that the challenge is not technological but organizational. "Unlike most technologies however, the exact functioning of AI models is a blackbox, even to the engineers who develop it. What is needed now is the collective ability to shape and govern this technology, despite how rapidly the technology is set to evolve," she explained.

Fornabaio

How Can Nations Begin Implementing AI Governance Today?

The report's recommendations suggest concrete steps that policymakers and international bodies can take immediately, even as AI technology continues to advance. These actions focus on building the institutional capacity and political will needed for coordinated global action.

• Convene Scientific Experts: Establish independent panels at the United Nations to assess AI capabilities and risks, creating a shared factual foundation that transcends national interests and political disagreements.
• Identify High-Risk Research Areas: Work with researchers and industry leaders to define which categories of AI development pose the greatest environmental and social risks, then negotiate binding restrictions on those activities.
• Design Flexible Implementation Pathways: Create international agreements that set clear global goals while allowing countries to implement solutions suited to their economic capacity, existing infrastructure, and cultural values.
• Monitor Resource Consumption: Establish transparency requirements for AI data centers regarding water usage, energy consumption, and carbon emissions, similar to environmental reporting standards for other industries.

The timing of this report reflects growing recognition that AI's climate applications cannot be separated from AI's environmental costs. While AI can help predict renewable energy generation and optimize smart grids, the computing infrastructure required to run these systems consumes enormous amounts of water and electricity.

What Role Does Environmental Ethics Play in AI Governance?

Beyond policy and science, the report's findings align with recent statements from religious and ethical leaders emphasizing that AI development must remain accountable to environmental stewardship. Pope Leo recently published an encyclical on artificial intelligence that specifically highlighted the environmental impact of data centers, calling attention to the enormous amounts of water, energy, and infrastructure they require, as well as their significant influence on carbon emissions.

The Pope's encyclical, titled "Magnifica Humanitas, on Safeguarding the Human Person in the Time of Artificial Intelligence," frames environmental protection as inseparable from human dignity. He writes that "the creative intelligence of humanity is a gift that can alleviate suffering and open up new possibilities, but it must remain ordered toward the common good, justice, the care of the vulnerable, and creation".

This ethical framing complements the CCSI report's governance recommendations by emphasizing that AI governance is ultimately about values, not just technical standards. The question is not simply whether AI can help solve climate change, but whether the process of developing and deploying AI respects both human communities and the natural world.

The convergence of scientific, policy, and ethical perspectives suggests that the next phase of AI development will be defined not by raw computational power, but by the institutions and agreements nations build to ensure that power serves humanity's long-term interests rather than short-term gains.