The White Paper on: AI and Environment was prepared by the Nature & People Foundation, Fudan University (China), and the Pahlé India Foundation (India), with the participation of the UN Internet Governance Forum’s Dynamic Coalition on Environment. The study was presented on July 6 in Geneva at the first meeting of the Global Dialogue on Artificial Intelligence, established by the UN General Assembly.

Comprehensive study

The authors describe their work as a comprehensive study examining two aspects of AI development. On the one hand, new technologies have the potential to accelerate the fight against climate change, biodiversity loss, land degradation, and water scarcity. On the other hand, the rapid growth of computing power requires enormous amounts of electricity, water for cooling data centers, and natural resources for equipment production.

“Whether artificial intelligence ultimately becomes nature’s ally or its adversary will depend less on the technology itself than on the engineering solutions, transparency requirements, and policy frameworks the world adopts in the coming years,” Columbia University professor and director of the Sustainability Solutions Network Jeffrey Sachs emphasizes in the foreword to the publication.

The report’s authors note that UN Secretary-General António Guterres previously called on leading AI companies to openly disclose data on the environmental impact of their technologies. However, despite growing interest in regulating artificial intelligence, the environmental consequences of its development remain insufficiently studied.

Human assistant

The report cites numerous examples of how AI is already helping nature. Algorithms improve the accuracy of climate models and weather forecasts, enable faster detection of forest fires, track greenhouse gas emissions, monitor the status of rare animal species, predict droughts and floods, detect leaks in water supply systems, and optimize irrigation in agriculture.

Ecological footprint

But the technology also has a downside. As generative AI becomes more widespread, the primary environmental burden gradually shifts from training the models to their daily use by billions of people. According to the report, by 2025, all AI systems worldwide could generate 32.6–79.7 million tons of CO₂ emissions—roughly the same amount as New York City produces in a year.

Water consumption

However, information on the actual consumption of electricity, water, and other resources is still clearly insufficient. The authors cite this lack of transparency as one of the main problems. It is estimated that by 2025, AI could consume between 312 and 765 billion liters of water—roughly the same amount as humanity spends annually on bottled water. By 2027, this figure could reach a volume comparable to half the UK’s annual water consumption. Furthermore, data centers are often built in regions where fresh water is already scarce, placing additional strain on local resources.

Expert recommendations

Researchers compared approaches to AI regulation in Russia, India, China, the United States, and the European Union. They concluded that none of the existing models fully considers the environmental impacts of artificial intelligence development. The proposed immediate steps include introducing uniform international environmental reporting standards for AI, assessing its impact throughout its lifecycle, making energy and water use data more transparent, and more actively coordinating international efforts.

The study’s main conclusion is quite simple: the environmental footprint of artificial intelligence is not an inevitable price of technological progress. What will happen in ten or twenty years depends on the decisions governments, companies, and international organizations make today.

The first report of the Independent International Scientific Panel on AI, established under the auspices of the United Nations, was published earlier .