Year: 2025
Type: article
Abstract: As artificial intelligence (AI) technologies continue to spread into human life, developers must ensure benefits while minimizing the risk of adverse impacts. This study aims to evaluate risks in real-world AI applications using the AI Incident Database. It employs Failure Mode and Effect Analysis and the National Institute of Standards and Technology AI Risk Management Framework to identify failures, their causes and effects, and assess how current systems address them. A total of 100 incident reports were analyzed. The findings indicate frequent failures in autonomous systems and biased predictions. Seven cases were classified in the highest risk categories, including those involving physical harm and loss of life. Over 80% failures originated from algorithmic flaws or poor data quality. The method employed successfully evaluates the risks in current AI applications, revealing critical gaps in risk management and emphasizing the urgent need for targeted safeguards and proactive mitigation strategies.
Authors: Yunarso Anang, Lya Hulliyyatus Suadaa, Lutfi Rahmatuti Maghfiroh, Nori Wilantika, Masakazu Takahashi, Yoshimichi Watanabe
Institutions: Politeknik Statistika STIS, University of Yamanashi
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7092188744