The AI Incident Database recorded 362 documented incidents in 2025, up from 233 in 2024, a 55% year over year increase in reported artificial intelligence harms and near misses.

Between the start of May and the end of July 2026, the database added 148 new incident identifiers, covering entries 1471 through 1618. That pace works out to roughly 49 new incidents a month across the three month window.

Cumulative totals reached approximately 800 to 900 unique incidents through the first quarter of 2026, with the collection growing at roughly 130 to 180 new incidents annually before the 2026 acceleration.

Two categories dominate the volume. Deepfake and synthetic media incidents accounted for 179 of 346 entries in one measured period, or 52% of the total. That grouping includes voice impersonation used in business email compromise, video manipulation in disinformation campaigns and image deepfakes used in extortion and fraud.

Three public databases track AI incidents systematically and overlap partially. The AI Incident Database is maintained by the Responsible AI Collaborative. The OECD AI Incidents and Hazards Monitor tracks roughly 5,000 to 7,000 incident reports compiled through news monitoring, capturing broader media coverage than the AIID entries. The MIT AI Risk Repository provides a third classification framework.

Counting methodology affects the totals substantially. AIID assigns one incident identifier per distinct event and links multiple reports to it, while media monitoring approaches count reports rather than events, which produces the order of magnitude difference between the databases.

Source: AI Incident Database - https://incidentdatabase.ai/blog/incident-report-2026-may-june-july