When Artificial Intelligence Fails

Abstract

Diverse initiatives promote the responsible development, deployment and use of Artificial Intelligence (AI). AI incident databases have emerged as a valuable and timely learning resource and tool in AI governance. This article assesses the value of such databases and outlines how this value can be enhanced. It reviews four databases: the AI Incident Database, the AI, Algorithmic, and Automation Incidents and Controversies Repository, the AI Incident Tracker and Where in the World Is AI. The article provides a descriptive analysis of these databases, examines their objectives, and locates them within the landscape of initiatives that advance responsible AI. It reflects on their primary objective, i.e. learning from mistakes to avoid them in the future, and explores how they might benefit diverse stakeholders. The article supports the broader uptake of these databases and recommends four key actions to enhance their value.