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

Growth in reported incidents runs 35 to 45 percent year over year, outpacing aggregate growth in AI deployment. That differential means incident reporting is rising faster than the underlying installed base, which reflects both wider deployment into consequential settings and more systematic documentation.

Recent submission volume illustrates the pace. Between the start of November 2025 and the end of January 2026, the database added 108 new incident identifiers, running from Incident 1254 through Incident 1361.

Harm categories cluster tightly. Misinformation, deepfakes, and content harms account for approximately 28 percent of recent incidents. Discrimination and bias represent about 22 percent. Physical safety failures make up roughly 14 percent. Generative AI systems dominate recent additions, accounting for 58 percent of 2025 entries.

Scope varies across trackers. The OECD AI Incidents and Hazards Monitor covers 5,000 to 7,000 broader media coverage entries, with substantial overlap against the AI Incident Database. The difference reflects methodology, since the OECD monitor captures media reporting while the incident database applies a narrower verification standard before assigning an identifier.

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