Potential futures for the IPCC’s approach to artificial intelligence

Abstract During the first six assessment cycles of the Intergovernmental Panel on Climate Change (IPCC), generative artificial intelligence (AI) tools were not widely available. This has rapidly changed. The IPCC will need to make choices about how to use artificial intelligence powered applications in literature identification, literature assessment, and communicating the contents of the report. These are not merely technical decisions, but sociopolitical challenges, as the IPCC has a social function of creating a shared basis of reality for addressing climate change. To explore these challenges, this paper describes four scenarios of potential AI development: rapid adoption of AI agents, perceptions of superior AI reliability, caution among the research community, and public backlash to AI. If the IPCC can navigate these emergent challenges, there is an opportunity for it to serve as a learning opportunity for other institutions grappling with how to do scientific assessment in a world with generative AI.