The United Nations Environment Assembly on Friday approved the first resolution addressing environmental aspects of artificial intelligence (AI), but it did not include provisions to monitor AI systems throughout their lifecycle. Experts say this approach is essential to understanding AI's water, power and critical mineral consumption.
The resolution proposed by Kenya aims to harness the “opportunities and benefits of artificial intelligence systems by supporting the environment and minimizing its impact.”
It is also calling on the United Nations Environment Program (UNEP) to produce a report on the “environmental benefits, risks, and impacts of artificial intelligence.''
As negotiations progressed over a week in Nairobi, a draft resolution on AI called on the UNEP executive director to examine the environmental benefits, risks and impacts of artificial intelligence systems “throughout their lifecycle.”
But while governments including Kenya, Norway, Colombia and the European Union supported such language, an annotated draft text showed Saudi Arabia, Russia and the United Arab Emirates wanted it removed.
When the final resolution was passed on Friday, all traces of the AI lifecycle had been removed from the text. References to AI water and energy consumption, which had been featured in previous draft documents, were also removed.
“We can't talk about sustainable AI unless we address the entire lifecycle, from the traceability of critical minerals to the water used in data centers to how much renewable energy from developing countries feeds AI systems in wealthy regions,” said Faith Munyaro, Kenya's AI focal point.
Muñaro said that while adopting the resolution was an important first step, UNEA needed to move forward in negotiations to address “blind spots” and provide stronger language and clearer commitments on life-cycle accountability.
“We need to build sustainability into AI from extraction to disposal, or we risk repeating the same patterns of inequity seen in earlier technology transitions,” she told Climate Home News.
Direct financing cannot be expected.
As negotiations reached the halfway point on Wednesday, the AI resolution was essentially on the brink of collapse over funding issues, with Saudi Arabia and Iran arguing that money should flow primarily from developed countries to developing countries, and the UK and EU arguing that funding should come from all sources.
In the end, countries reached a compromise that avoided the obligation for wealthy countries to directly fund AI capabilities in the Global South. Instead, all countries are encouraged to “strengthen partnerships” that can mobilize finance, alongside “increased investment, including from the private sector and philanthropy,” in AI to support sustainable development.
AI is increasingly being used in the environmental field, has already been introduced in developing countries, and is in high demand for funding. For example, Sierra Leone's new NDC climate plan calls for approximately $7 million, including funding from donor countries, to build an AI-based climate and weather forecasting system to improve resilience. And in Kenya, AI is helping conservationists monitor forest degradation, initiate reforestation, and predict the carbon storage capacity of new forest areas.
Kenya's Munyaro said most data centers are concentrated in developed countries, but Africa lacks the expertise and funding to develop its own AI data systems. She added that without direct funding commitments, the burden would shift back to developing countries, potentially undermining such environmental projects.
Is AI good or bad for energy transition?
Soumya Joshi, research director at the Stockholm Environmental Institute (SEI), said AI has a significant impact on both climate and biodiversity and needs to be designed in a way that “doesn't repeat the same mistakes made before in the transition of mining technology”.
He said future discussions need to be informed by science and the environmental impacts along the entire AI value chain, including water, electricity, critical minerals and rare earths for making semiconductor chips, as well as pollution and what happens to AI systems at the end of their useful life.
Joshi said there is a need to prevent AI from increasing demand for electricity, reinforcing dependence on fossil fuels that undermines the transition to clean energy.
Earlier this year, UN Secretary-General António Guterres called on Big Tech to power all data centers with 100% renewable energy by 2030.
Data centers will account for approximately 1.5% of global electricity consumption in 2024. But that number is expected to more than double by 2030 as tech giants continue to build out the infrastructure needed to support power-hungry AI technologies.
Renewable energy sources combined with batteries are expected to provide half of the additional electricity, while increased demand from data centers will be a “significant” driver of growth in fossil gas and coal-fired power generation until the end of the century, according to the International Energy Agency (IEA).
As the Paris Agreement approaches its 10th anniversary, what are its achievements?
Geopolitics limits Nairobi outcome
The resolution on AI was largely seen by observers as a victory for UNEA, which unfolded in a tense political environment that limited progress on a range of important environmental issues.
The United States rejected the results, condemning what it called a “climate drama,” in line with President Donald Trump's administration's denial of climate science and his efforts to block climate action.
Behind the scenes, the governments of oil-rich Saudi Arabia and Turkiye, host of next year's COP31 climate talks, pushed to water down language on climate change, including the science of melting glaciers.
This rejection of established evidence prompted strong criticism in the final session of the conference from the small island states Fiji and Barbados, as well as the European Union and Australia. In the closing plenary, the EU representative said that EU Member States arrived at UNEA-7 with high hopes for the environment and multilateralism, but must accept the fact that Parliament “only achieved good results on some resolutions and less so on others.”
There was also widespread disappointment at the weak resolution on mining and transitional minerals, which agreed only to further consultations on international cooperation, rather than the creation of an expert group proposed by Colombia and Oman to identify new measures to make supply chains greener and more transparent.
But fears that some member states would use UNEA as an opportunity to resume their task of negotiating a global treaty on plastic pollution did not materialize, said Andrés del Castillo, a senior lawyer at the Center for International Environmental Law (CIEL).
Negotiations on a new deal were suspended in August after failing to reach an agreement with fossil fuel producing countries to block proposed caps on plastic production, a major market for petrochemicals. It is scheduled to reopen in February with a new chairman elected.
Referring to the Ministerial Declaration adopted in Nairobi on Friday, Del Castillo said countries “reaffirm our common determination to work constructively and proactively, with a sense of urgency and solidarity, and work toward the conclusion of bilateral agreements.” [plastics] negotiation”.
