The growing use of AI is adding emissions and environmental pressures that drive climate change.

AI is being leveraged by the fossil fuel industry to identify new oil and gas reserves and optimise drilling. Meanwhile, large-scale AI is generating unprecedented electricity and water demand, while mining raw materials and manufacturing specialised hardware create additional ecological burdens.

A growing narrative argues that the environmental footprint of consumer-facing AI models such as ChatGPT is justified by their potential to accelerate climate solutions. In reality, most AI applications that meaningfully support climate action do not depend on large-scale generative models. Most are smaller, targeted and far less computationally intensive.

Examples of AI-based climate applications span a wide range of sectors, with many already delivering results and others still experimental.

AI can transform how we observe and monitor our changing planet. AI weather models can produce global 10-day forecasts in seconds on standard hardware. Climate TRACE and MapBiomas use computer vision to analyse satellite data, creating detailed emissions and land-use inventories and enabling near real-time deforestation alerts. Decadal climate emulation models can simulate how policies might affect the planet, but more research is required before they can replace traditional physics-based models.

AI can help reduce greenhouse gas emissions. It is already being used to forecast renewable energy supply and demand, optimise traffic signals and freight routes, and identify energy waste in buildings. Applications in cement and steel are advancing but have yet to demonstrate scalable impact. AI-enabled materials science could unlock breakthroughs in batteries or carbon-capture catalysts, but remains largely at the research stage.

AI can help societies prepare for warming already underway. It is being used to model crop yields and vegetation stress, forecast food insecurity, identify neighbourhoods vulnerable to heat and flooding, and track hurricanes for early warning. Other applications, including precisely linking disease burden to climate signals, remain at an earlier stage.

AI is never a neutral tool but a reflection of the choices, values and inherent biases of its creators.

There is a severe lack of data from the Global South, leading to models that are unrepresentative of populations most vulnerable to climate change. Citizen-generated data can also favour dense populations and roads, leaving remote areas underrepresented. Initiatives such as AfriClimate AI are building local monitoring systems for data-scarce environments, while Climate Change AI’s Data Gaps project identifies data bottlenecks and needed datasets.

True climate justice in AI means asking who benefits from AI insights and who bears the environmental and labor costs, ensuring affected communities are involved in project co-design. An AI project mapping wildlife corridors, for example, might rely on data from local communities, conservation organisations or indigenous groups, who should gain access to its insights. Developers must also consider misuse, such as whether endangered species locations could be exploited by poachers. Responsible development therefore means safeguards for how data and outputs are shared, and being clear about who is accountable for unintended consequences and the decisions AI informs.

AI is not a prerequisite for all climate challenges.

A recurring foundational question needs to be: Is AI really the best solution for this problem? Many barriers depend on political will, infrastructure and traditional engineering. Treating the climate crisis primarily as a technical problem that can be optimized with more data and sophisticated models, risks ignoring structural drivers of ecological breakdown. But in many domain-specific cases, AI can be a powerful accelerant. We should avoid promising capabilities that do not yet exist, while scaling tools already showing results and drawing attention to data and research gaps.

Education and collaboration are key. Climate scientists and policymakers need to understand what AI can and cannot do, while technologists need to explore the diversity and complexity of the climate challenges their skills can help tackle. That’s the mission of organisations such as Climate Change AI, which catalyzes work at the intersection of climate change and machine learning through a global interdisciplinary community. By bridging expertise, closing data gaps and focusing on problem-specific solutions, we can move beyond the hype of “AI for Good” and into the gritty reality of building a resilient, inclusive and sustainable future.

__________________________________________________________________________

Elisabeth Anna Resch is the Director of Partnerships at the University of Cambridge SupTech Lab and Digital Transformation Solutions - with the mission to accelerate the public sector’s digital transformation to build more resilient, inclusive and sustainable economies. She initiated global initiatives such as SupTech Week, innovation fellowships, and public-private secondment programs.

With over a decade of experience at the United Nations Global Compact and the World Bank, Elisabeth has mobilized multi-stakeholder partnerships and leadership initiatives that champion women’s empowerment, responsible innovation and social accountability across sectors. In 2015, Elisabeth contributed to the adoption of the Sustainable Development Goals while working at the Executive Office of the UN Secretary-General.