Anticipating AI Risks in Climate Research Agendas

Abstract Climate change has often been characterized as humanity's greatest threat. However, rapid advancements in artificial intelligence (AI) may cause technological risks, such as novel pathogens or autonomous drone warfare, to materialize more rapidly than many climate impacts. This raises fundamental questions for the climate community. How should climate professionals address the potential for AI systems to automate cognitive tasks? How might dramatic technological transformation over shorter timescales affect the relative value of near-term versus long-term research priorities? And what can the climate community contribute to managing this growing risk? Drawing on recent developments in AI capabilities, expert forecasting, and risk assessment, I suggest three broad scenarios (Business-as-Usual, AI Progress, and Doom) to guide our thinking about AI development and its implications for climate research. Rather than advocating for any single response, I explore how climate professionals, including researchers, communicators, educators, and graduate supervisors, might thoughtfully prepare for multiple futures while maintaining commitment to addressing the climate crisis. The analysis considers potential adjustments to research communication, graduate education, and collaborative engagement with the AI safety community, while acknowledging both the potential harms and benefits of this technology. These questions warrant serious consideration as the climate research community navigates unprecedented technological change alongside ongoing environmental challenges.