Abstract: [Purpose/Significance] This study aims to identify the key research focuses in the field of AI ethics and provide theoretical references for the academic community and policymakers. [Design/Methodology] Employing the Web of Science Core Collection as the data source and applying the emerging BERTopic model, this study conducts topic mining and knowledge structure extraction on the literature pertaining to AI ethics, thereby unveiling the research trends and underlying structures in the field. [Findings/Conclusion] The field of AI ethics is divided into 16 themes, which can be summarized into seven research directions: medical and health ethics, education and research ethics, military and security ethics, sustainable development and industrial ethics, journalism and media ethics, robotics ethics, and social governance and algorithmic fairness, revealing a dual-core theme structure of "technical ethics" and "social ethics". Concurrently, research in this field exhibits a clear trend of shifting focus from fragmentation toward systematization and from theoretical exploration toward practical governance, with growing attention from both the public and the academic community. [Originality/Value] The BERTopic model is introduced into the field of AI ethics research, leading to the identification of the knowledge structure of these themes. Policy recommendations are also proposed based on differences in thematic risks, thereby providing decision-making support for AI ethics governance practices.

Keywords: Artificial Intelligence（AI）, Ethics, Topic mining, Governance policies