Generative AI for the power sector: Methodology, safety, trustworthiness, and policy

The integration of emerging energy technologies such as renewables, hydrogen fuel cells, battery storage, and increased grid interconnection into power grids introduces significant uncertainty in generation forecasting, grid stability, and demand-supply balancing. Thus, power grids are undergoing a monumental transformation to enable the transition toward sustainable, low-carbon energy systems. They are evolving into digitalized smart grids where cyber, physical, and social domains converge through intelligent components, advanced metering infrastructure, and wireless connectivity. In parallel, recent breakthroughs in artificial intelligence (AI), particularly Generative AI (GenAI), offer transformative technological capabilities useful for addressing these emerging challenges. While existing literature extensively explores individual applications of AI within the power sector, comprehensive assessments of GenAI's multidimensional impacts encompassing operational value, safety, trustworthiness, and regulatory frameworks remain limited. To bridge this gap, we introduce a 3-dimensional analytical framework focused explicitly on operational value creation, safety, and trustworthiness considerations, and tailored policy and regulatory implications. We propose a structured, replicable methodology for constructing GenAI use-cases, demonstrating its effectiveness using predictive maintenance as a representative example, with performance gains exceeding 31% compared to traditional AI methods. Further, we critically assess GenAI safety risks, establish a clear taxonomy with quantifiable metrics for trustworthy AI deployment, and derive targeted policy and regulatory recommendations based on emerging global practices relevant to the power sector.