The United Nations Environment Programme (UNEP) is raising awareness about the growing environmental footprint of artificial intelligence (AI) ahead of the United Nations Environment Assembly in December. While AI holds promise for addressing environmental challenges, such as monitoring methane emissions and mapping ecosystem damage, research highlights its negative environmental impacts, including high energy use, water consumption, reliance on rare minerals, and electronic waste from data centres. UNEP emphasizes that the net effect of AI must be positive before large-scale deployment.
AI is a set of technologies capable of processing information and mimicking human cognition, with recent advances fueled by increased computing power and the expansion of data for training AI models. Its ability to detect patterns and predict outcomes has made it a powerful tool for environmental monitoring, efficiency improvements, and informed decision-making. UNEP uses AI, for example, to track methane leaks from oil and gas infrastructure, illustrating its potential to contribute to addressing climate change, biodiversity loss, and pollution.
However, large-scale AI deployment comes with significant environmental costs. Data centres require massive amounts of raw materials and rare earth elements for electronics and microchips, which are often mined unsustainably. They also generate electronic waste containing hazardous substances and consume large quantities of water for cooling, potentially surpassing the annual water use of entire countries. Moreover, AI operations require significant electricity, much of it from fossil fuels, contributing to greenhouse gas emissions. For instance, a single AI request via ChatGPT can consume ten times the energy of a typical Google Search, and data centres in AI-heavy regions could account for up to 35% of total energy use in some countries by 2026.
Beyond direct environmental impacts, AI applications can have unpredictable secondary effects. Self-driving cars could increase traffic emissions, while AI-generated misinformation might undermine climate action. The broader consequences of AI deployment remain uncertain, making environmental impacts a complex issue to manage.
Efforts to mitigate these impacts are emerging but remain limited. Over 190 countries have adopted non-binding ethical AI guidelines, and both the European Union and the United States are introducing environmental considerations into AI legislation. UNEP notes that most national AI strategies still overlook sustainability, leaving significant gaps in environmental safeguards.
UNEP’s recent issue note outlines five strategies to reduce AI’s environmental impact. These include developing standardized methods to measure AI’s footprint, mandating disclosure of environmental consequences for AI products, improving algorithmic efficiency, greening data centres through renewable energy and resource recycling, and integrating AI policies with broader environmental regulations. UNEP’s foresight work, summarized in Navigating New Horizons – A Global Foresight Report on Planetary Health and Human Wellbeing, highlights the importance of anticipating environmental challenges from technological innovations while addressing the triple planetary crisis of climate change, biodiversity loss, and pollution.
Overall, AI presents both opportunities and risks for the environment. While it can enhance monitoring, prediction, and efficiency, its energy demands, material consumption, and potential unintended effects make sustainable deployment essential. UNEP emphasizes strategic regulation, corporate responsibility, and informed policy integration as key measures to ensure AI contributes positively to planetary health.
