[2N5-OS-28a] OS-28

[2N5-OS-28a-05] Application of the AI Ethical Impact Assessment to the Insurance Sector

〇Hidekazu Takahashi¹, Virginia Ghiara², Chiara Picardi³, Kyouko Ohhashi¹, Takao Mohri¹, Izumi Nitta¹ (1. Fujitsu Limited, 2. Fujitsu Research of America, Inc., 3. Fujitsu Research of Europe Ltd.)

Keywords:AI ethics, Impact assessment, Risk assessment, Responsible AI

The AI Ethics Impact Assessment proposed in 2022 is a method to extract potential ethical risks of AI systems according to the AI Ethical Model which was developed by interpreting AI ethical guidelines. This AI Ethical Model is a structured and concrete set of ethical requirements that AI systems must meet. Currently, AI laws are being developed in Europe to ensure the safe use of AI systems. According to the upcoming legal frameworks, AI systems are required to comply with industry regulations. The AI Ethical Model, however, only supports the "Ethics Guidelines for Trustworthy AI" to extract general ethical risks, and does not support sectorial regulations. In this paper, we present an AI Ethical Model, conforming to the guidelines published by the European Insurance and Pensions Supervisory Authority. It enables the application of the AI Ethics Impact Assessment to the insurance sector, identifying both general and sector-specific risks.