| Canonical name | OECD AI Incidents and Hazards Monitor |
| Also known as | AIM, AI Incidents Monitor |
| Address | https://oecd.ai/en/incidents |
| Operator | the OECD.AI expert group on AI incidents (OECD.AI), with the support of the Patrick J. McGovern Foundation op |
| Founded | the expert group was formalised in January 2023 to develop it; LLM classification from November 2024 |
| Kind | monitor |
| Roles | harm, monitoring |
| Lifecycle | active |
| Observation | 2026-10-28: read |
| In its own words | “Automated media discourse monitor of AI incidents and hazards (Beta)” op |
| Unit | incident or hazard (an event in which an AI system leads, or could plausibly lead, to harm) op |
| Intake | automated: news articles aggregated by Event Registry; since November 2024, LLM classification (GPT-4o mini, then GPT-4o) and LLM-written metadata (o3-mini); daily op |
| Count | about 18,579 incidents and hazards (as of 2026-10-28) obs |
| Cadence | daily op |
| Note | its 14 'risk themes' are an analysis of its summaries in OECD AI Papers No. 53 (February 2026), which clustered them by k-means into 25 clusters and merged them to 14 |
op operator statement · 3p third-party report · obs our observation · rd our reading
The monitor calls itself an "Automated media discourse monitor of AI incidents and hazards (Beta)." An AI incident is an event in which the development, use or malfunction of an AI system directly or indirectly leads to harm; an AI hazard is an event that "could plausibly lead to an AI incident." The definitions follow the OECD's paper of May 2024.
Event Registry clusters news articles about one happening into an event, and only events tagged with AI concepts reach the monitor. Since November 2024 a small model (GPT-4o mini) filters, a larger one (GPT-4o) reclassifies, and o3-mini writes the title, summary, harm type, severity, stakeholders and country from the top three articles. New items are matched to existing events by embedding similarity. The pipeline runs daily on events one to four days old.
An event enters when reputable outlets report it. A harm no outlet reports has no event to cluster. The record's own wording is machine-written from the articles, so the monitor is itself a machine reception of the news: its summaries are compositions, and the articles are their sources.
It is the automated counterpart of the AI Incident Database, whose editors resolve each submission. A research study can enter it as an event: the Tow Center's 2025 study of AI search citation is filed as an incident.
| Admission requires harm? | yes: an incident, or a hazard that could plausibly lead to one op |
| Preserves original outputs? | no: titles, summaries and metadata are LLM-written from the top articles op |
Lee Sharks, ed. "OECD AI Incidents and Hazards Monitor." MMRS meta-aggregator, Machine-Mediated Reception Studies, compiled 2026-10-29. https://www.machinemediation.org/aggregators/oecd-ai-incidents-monitor/
Record https://www.machinemediation.org/aggregators/oecd-ai-incidents-monitor/ · entity https://www.machinemediation.org/aggregators/oecd-ai-incidents-monitor/#entity