This report investigates "AI for Science," referring to the application of artificial intelligence (AI) in scientific research, and provides an overview of the latest trends as of late 2025.

The development and application of AI are profoundly transforming industrial structures and societal systems. As AI technologies permeate all domains of society, Digital Transformation (DX) is entering a new phase, often described as AI Transformation (AX).

Scientific research is no exception to this trend. AI is increasingly embedded at every stage of the research process, reshaping approaches to hypothesis generation, experimental procedures, data analysis, and knowledge integration, and significantly accelerating both the pace and scale of scientific discoveries. Historically, scientific research has experienced four major paradigm shifts: the First Paradigm (Empirical/Experimental Science), the Second Paradigm (Theoretical Science), the Third Paradigm (Computational Science), and the Fourth Paradigm (Data-Driven Science). Against this backdrop, AX in scientific research―commonly referred to as "AI for Science"―is increasingly regarded as a potential "fifth paradigm" following these. This emerging paradigm has the potential to fundamentally transform the nature of science by integrating and reorganizing the four paradigms on an AI foundation.

This report is organized into five chapters.

Chapter 1 outlines the background of AX in scientific research and presents a conceptual framework for AI for Science, incorporating recent policy trends. It envisions a mutually reinforcing relationship between the application of AI to scientific research (i.e., "AI for Science"; "AI → scientific research") and the contributions of scientific research to advances in AI (i.e., "Science for AI"; "scientific research → AI") . In addition, the chapter broadens the scope of AI for Science by examining how AI is reshaping the broader science, technology, and innovation ecosystem. It then analyzes bibliometric data on AI for Science-related papers, assesses Japan's global position based on trends in AI-related paper counts, and provides an overview of domestic and international policy trends to clarify national strategies and policy directions of various countries.

Chapters 2 through 4 focus on specific research trends in AI for Science. Chapter 2 provides an overview of the field as a whole, identifies cross-cutting trends, and discusses key considerations for advancing AI for Science, including research actors, institutional frameworks, and research environments. It addresses foundational elements such as knowledge and data infrastructure and large-scale AI models, as well as the impacts of AI for Science on scientific practice itself and the study of these impacts through metascience. Chapter 3 introduces trends in the life sciences, materials, environmental and energy, and information science, highlighting the contributions of AI to each field and the direction of progress (AI → fields). Chapter 4 adopts the reverse perspective, focusing on contributions from individual scientific fields to AI (fields → AI), with particular attention to semiconductors, computational infrastructure, knowledge and data platforms, and foundational models.

Finally, Chapter 5 synthesizes the findings and insights from Chapters 1 through 4 and discusses their implications for policy design aimed at further advancing AI for Science. This report seeks to contribute to broader discussions on how to holistically assess scientific research processes, AI research infrastructure, societal implementation, and metascience. Such an integrated perspective is expected to support agile and mutually reinforcing actions that facilitate the continued transformation and evolution of scientific research.

• (in Japanese) Trends in AI for Science 2026: How AI Transformation Is Reshaping Science, Technology, and Innovation - The Beyond Disciplines Collection -