Incidents by Category and Autonomy
Every documented incident falls into one of five categories and one of three autonomy classifications - whether an AI system acted on its own, was directed to cause harm by a person, or behaved exactly as a deployed system was designed to. This grid shows where the corpus concentrates, which is a first signal for where regulation and oversight are most needed.
9 of 9 incidents in the archive are counted in this grid.
Cell color is shaded relative to this grid's own current highest count (3), not an absolute scale - the number in each cell is the real count. A shaded cell links to that category's listing, filtered to the matching autonomy classification for readers with JavaScript enabled; without JavaScript, or for a search crawler, the link shows the full, unfiltered category listing instead.