Sustainability-focused generative AI risk mitigation strategies

The rapid rise of generative AI (GenAI) has sparked the sustainability community to explore its potential applications, such as climate impact modeling and renewable energy optimization. However, deploying these GenAIpowered solutions in enterprise environments raises risk concerns. In particular, chatbots and similar GenAI applications face risks of misinformation and disinformation stemming from knowledge sources, user prompts, and the response generation process. While traditional probabilistic analysis methods often struggle to effectively assess risks in GenAI applications, the Risk-Reducing Design and Operations Toolkit (RDOT) provides a qualitative complement for addressing these challenges. In this study, we propose a framework that applies the RDOT methodology specifically to GenAI applications in the sustainability domain, drawing lessons learned from an internal enterprise GenAI application development. We outline mechanisms for structured risk identification, testing, evaluation, and specific risk mitigation techniques. By embedding these techniques in the development and testing process, we enhance the reliability of sustainability-focused GenAI solutions. We found that 34 (out of 111 or 31%) of the RDOT strategies have already been utilized in the internal GenAI application with 10 of them showing particular value in sustainability-focused GenAI application development. Another 17 (15%) were not utilized but are highly promising. Our finding addresses a gap in current practices, providing sustainability practitioners with a systematic way to navigate the challenges of deploying GenAI technologies in real-world settings.