An Incident-Based Ethical Risk Screening Matrix for AI-Based Information Systems Using Public AI Incident Records

This study developed an incident-based ethical risk screening matrix for AI-based information systems using 100 publicly reported AI incident records from 2020 to 2026. It used descriptive content analysis and criterion-based purposive sampling. Each incident was coded by year, system domain, ethical risk type, affected stakeholder, ethical principle violated, severity level, and evidence strength. The study aimed to identify common ethical risk patterns and translate them into a practical screening tool for IT managers and organizations. Results showed that the most common ethical risks were misinformation or deception, followed by misuse or malicious use, safety failure, privacy violation, and accountability failure. Generative AI and chatbots had the highest number of incidents. Notable risk exposure was also found in law enforcement and surveillance, finance, government and public service, education, and social media platforms. The most affected stakeholder group was the general public. Most cases were assessed as high severity, while some were assessed as critical severity. Another supporting pattern was the prevalence of deepfakes and synthetic media, mainly linked to impersonation, fraud, misinformation, privacy harm, and reputational damage. The study concludes that AI-based information systems need ethical screening early in adoption, deployment, or expansion. The proposed matrix helps users identify warning signs, ask targeted screening questions, and apply management actions. It also gives organizations a structured way to connect documented AI failures to practical review steps. The matrix is not a substitute for a full technical, legal, or cybersecurity review. It uses documented AI failures as practical evidence to support early risk identification and responsible IT management.