The number of documented AI incidents is rising faster than AI deployment itself, according to figures compiled from the AI Incident Database. The database logged 362 incidents in 2025, up from 233 in 2024, a 55 percent increase in a single year. The cumulative count reached roughly 800 to 900 unique incidents by the first quarter of 2026.

Generative AI now drives the majority of new entries, accounting for about 58 percent of incidents logged in 2025. The largest incident categories are misinformation and content harms at roughly 28 percent, discrimination and bias at about 22 percent, and physical safety failures at around 14 percent. The mix shows that most recorded harms involve content and decision-making rather than physical systems, though physical failures remain a meaningful share.

A separate tracker maintained by the OECD points in the same direction. Its monitor of media-reported AI content incidents rose from about 50 in early 2020 to over 200 in early 2024 and nearly 500 by January 2026, a roughly tenfold increase across the period.

The data describes a steady climb in documented cases where AI systems caused or nearly caused harm. As organizations deploy generative tools more widely, the incident counts suggest that monitoring, disclosure, and access controls are not keeping pace with adoption, leaving a growing record of public failures for researchers and regulators to study.

Source: AI Incident Database - https://incidentdatabase.ai/blog/incident-report-2025-november-december-2026-january/