Governance for Generative AI in Clinical Development: A Cross‑Sectional Survey in Japan

Abstract

Introduction The development of generative artificial intelligence (AI) has been driven by advances in AI and machine learning, leading to innovative applications across various fields such as natural language processing and image generation. Particularly in the clinical development industry, generative AI has contributed to streamlining data analysis and research support, thereby fostering progress in personalized medicine. However, the current situation is characterized by a lag in governance and regulatory compliance related to its use. Aim This study aims to investigate the current status of generative AI utilization and governance within organizations in Japan's clinical development industry, clarifying the extent of adoption and identifying associated challenges. Methods Between May 21 and June 13, 2025, an online survey was conducted targeting pharmaceutical companies, contract research organizations (CROs), academic research organizations (AROs), and system vendors. A total of 35 items were collected regarding organizational attributes, the status of generative AI use, governance measures, and training activities. Data were aggregated and analyzed using keyword analysis and word cloud visualization to identify salient features. Results Respondents and organizations involved included AROs (49 respondents across 36 organizations), pharmaceutical companies (48 respondents across 21 organizations), CROs (33 respondents across 13 organizations), and system vendors (2 respondents across 2 organizations). All respondents were individuals with responsible positions to make decisions regarding the use of generative AI within their respective organizations or departments in clinical development. Approximately 76% of respondents reported obtaining approval for generative AI use, with tools such as OpenAI's ChatGPT series and Microsoft Copilot being predominantly used. The status of governance varied between organizations; more than 83.3% of pharmaceutical companies (40 respondents), 60.1% of CROs (20 respondents), and 16.3% of AROs (8 respondents) had some form of governance documentation related to generative AI use. However, the development of operational-level standard operating procedures (SOPs) was insufficient across all organizations—only 8.3% of pharmaceutical companies, 6.1% of CROs, and none of the AROs and system vendors had such documents fully in place. Generative AI was mainly used for translation of documents, brainstorming, and document creation and maintenance, with expectations for future applications including advanced data analysis and programming tasks. Benefits cited included increased operational efficiency, automation, and creative support, while barriers such as security and privacy concerns and risks of misinformation were also noted. The focus of education and training centered on AI literacy and safe usage practices, emphasizing the need for strengthened security education within organizations. Conclusion We investigated and summarized the current status of generative AI utilization and governance for clinical development in Japan. A key finding was the lack of organizational governance documents related to utilization of generative AI and education and training. There is an urgent need to establish such governance frameworks along with more practical education and training methods.

Ähnliche Arbeiten