The number of documented artificial intelligence failures and harms is climbing sharply, according to incident-tracking databases. One independent tracker reported an 81 percent increase in documented incidents from 2024 to 2025, logging 312 incidents in a single year, of which 23 were classified as critical.

The broader picture shows steady acceleration. The AI Incident Database recorded roughly 800 to 900 unique incidents through the first quarter of 2026, growing by 130 to 180 new entries per year. Total counts are rising 35 to 45 percent year over year, faster than the growth in AI deployment itself, a sign that harms are outpacing adoption.

Generative AI dominates recent additions, accounting for about 58 percent of 2025 entries. Among incident categories, misinformation, deepfakes, and content harms made up roughly 28 percent of recent cases, followed by discrimination and bias at about 22 percent and physical safety failures at about 14 percent.

Severity has held relatively steady. The share of incidents involving fatal or major harm sits at approximately 3 percent, with the large majority causing lesser but still meaningful damage such as privacy violations, financial loss, or reputational harm. The data indicate that as AI systems spread into more products and decisions, the count of documented failures is rising faster than the technology's footprint.

Source: AIAAIC -- https://www.aiaaic.org/aiaaic-repository