The AI Incident Database recorded 362 incidents during 2025, up from 233 in 2024. Cumulative entries reached approximately 800 to 900 unique incidents through the first quarter of 2026, with the database adding roughly 130 to 180 new incidents per year.

Year-over-year incident growth runs at approximately 35 to 45 percent, a rate that exceeds AI deployment growth. Analysts reading the series note that faster incident growth than deployment growth suggests detection and reporting practices are improving rather than that systems are failing at a higher per-deployment rate.

Incident categories cluster in three areas. Misinformation and content harms account for roughly 28 percent of logged entries, discrimination and bias for approximately 22 percent, and physical safety failures for about 14 percent.

The composition of new entries has changed. Generative AI incidents now represent approximately 58 percent of 2025 entries, reversing the pre-2023 pattern in which recommendation systems and computer vision applications produced most logged cases.

Media-reported content incidents show the steepest curve. Reported AI content incidents rose from about 50 in early 2020 to more than 200 by early 2024 and to nearly 500 by January 2026, a tenfold increase across the period.

The MIT AI Incident Tracker maintains a parallel record, and the Stanford AI Index responsible AI chapter draws on both series.

Source: AI Incident Database - https://incidentdatabase.ai/