Multiple public databases have emerged to systematically track documented cases of artificial intelligence systems causing harm, malfunctioning, or generating public controversy, providing researchers and companies with a growing evidentiary record of how AI deployments go wrong in practice. The AI, Algorithmic, and Automation Incidents and Controversies repository, known as AIAAIC, began in 2019 as a private research project and has since evolved into a public interest database that catalogs incidents spanning privacy violations, biased outcomes, safety failures, and reputational controversies tied to AI and automated systems.
Separately, the AI Incident Database maintained by the Responsible AI Collaborative has logged more than 1,400 documented incidents, while the OECD's AI Incidents and Hazards Monitor has recorded more than 9,000 incidents and over 5,000 identified hazards across its broader classification system, which includes near misses and emerging risk patterns in addition to confirmed harms. Stanford's AI Index Report separately found that documented AI incidents rose to 362 in the most recent tracking period, up from 233 the year before, reflecting both increased AI deployment and improved detection and reporting of problems as they occur.
Researchers who study these incident databases note that the three major trackers use different classification methods and severity thresholds, meaning the same underlying incident can appear differently across each system, and that comparing raw incident counts between databases requires care. Even so, the consistent upward trend across all three trackers points to a real increase in the volume of documented AI-related harm as generative AI and autonomous AI agents have moved from experimental deployments into everyday business and consumer products.
Academic researchers have also called for greater standardization across these databases, arguing that a common taxonomy and reporting schema would make it easier for companies, regulators, and the public to identify recurring failure patterns and target specific categories of risk, such as data exposure, physical safety, or reputational harm from AI-generated content, for closer scrutiny.
Source: AIAAIC – https://www.aiaaic.org/aiaaic-repository