Public trackers that log documented AI system failures show incident counts rising sharply. The AI Incident Database (AIID), maintained by the Responsible AI Collaborative and used alongside repositories such as AIAAIC, had logged approximately 800 to 900 unique incidents through the first quarter of 2026, with roughly 130 to 180 new entries added per year.

Separately, the OECD's AI Incidents and Hazards Monitor, which scans media coverage rather than curating individual case files, tracked monthly reports of AI content-related incidents climbing from about 50 per month in early 2020 to nearly 500 per month by January 2026, a roughly tenfold increase over six years, with the sharpest jump occurring in the most recent twelve months.

Among cataloged AIID incidents, misinformation and content harms account for approximately 28 percent of cases, discrimination and bias issues make up about 22 percent, and physical safety failures represent close to 14 percent. Generative AI systems now drive the majority of newly logged incidents, accounting for an estimated 58 percent of 2025 entries, compared with roughly 5 percent in 2021.

Researchers who maintain these trackers caution that the figures represent a lower bound. Public databases capture incidents that receive media coverage or are self-reported, and they systematically undercount internal-only failures and incidents in regions with less developed AI journalism coverage.

Source: AIAAIC Repository and OECD AI Incidents and Hazards Monitor, chart via Statista - https://www.aiaaic.org/aiaaic-repository