Ask what risk AI poses for climate, and there’s a standard short answer. Energy-hungry data centres.
This is a very narrow view of a much wider landscape.
Using a comprehensive AI risk taxonomy (left side of Map), we identify 30 ways in which the use and misuse of AI may jeopardise efforts to reduce carbon emissions. Our mapping of AI-climate risks ranges from adverse impacts on specific low-carbon technologies towards a zero-emission energy system (top right of Map) to a general weakening of climate policy and governance (bottom right of Map).
Each of the 30 links from left to right in the Map describe a general AI risk and a specific concern for climate mitigation. The top one (#S1) is on data centres. In an accompanying database of supporting evidence and examples, there are 29 more.
The full Map is dense, complex, and diffuse. This makes it harder to come up with clear risk mitigation strategies as part of ongoing efforts on AI governance. An important next step for us is to identify who should take on which risks to then identify how they should be tackled.
Our aim in setting out these AI risks for climate is to balance the prevailing ‘AI for good’ narrative and increase the prominence of climate change in AI governance frameworks. Read our longer report explaining the Map here.
