An independent AI incident database maintained by INHUMAIN.AI recorded 847 verified AI-related incidents between 2020 and 2026, drawing from sources including the AIAAIC Repository, the AI Incident Database, government enforcement filings, court records and media reporting. The tracker logged 312 incidents in 2025 alone, up from 172 in 2024, an 81 percent year-over-year increase, and up from 42 incidents recorded in 2020.

Of the 2025 total, 23 incidents were classified as "critical" severity, a category reserved for events involving severe physical harm, death, large-scale discrimination or precedent-setting failures. Bias and discrimination made up the largest single category of 2025 incidents at 21.5 percent, followed by privacy violations at 16.7 percent and deepfake-related harms at 15.4 percent, which the tracker identified as its fastest-growing category.

North America recorded the most incidents by region in 2025, with 128, followed by Europe with 78 and Asia-Pacific with 61.

The tracker's outcome analysis found that of all 847 documented incidents since 2020, 412, or 48.6 percent, resulted in no known consequence for the organization or party responsible. Regulatory investigations or fines followed in 11.5 percent of cases, civil litigation was filed in 10.5 percent, and criminal convictions were recorded in 1.1 percent. The database's methodology requires each incident to be verified against at least two independent sources before inclusion in its summary statistics, and unconfirmed reports are excluded.