OECD AI Incidents and Hazards Monitor (AIM)
- Type
- Watchdog or incident database
- Place
- International — OECD
- Last checked
- 2026-09-22
- Next check due
- 2027-03-21
Ways to reach them
- Information only · 2026-09-22
Open submission process — announced as under development, NOT yet live
- Recipient
- OECD AI Incidents and Hazards Monitor (AIM)
- Who may use it
- People following the published contact instructions
- What it accepts
- Observed model behaviour; AI research concerns
Information only. This link provides background or evidence; it is not a place to send a message.No contact action is available for this entry. - Information only · 2026-09-22
Incidents dashboard (read-only)
- Recipient
- OECD AI Incidents and Hazards Monitor (AIM)
- Who may use it
- People following the published contact instructions
- What it accepts
- Observed model behaviour; AI research concerns
Information only. This link provides background or evidence; it is not a place to send a message.No contact action is available for this entry. - Information only
Homepage
- Recipient
- OECD AI Incidents and Hazards Monitor (AIM)
- Who may use it
- People following the published contact instructions
- What it accepts
- Observed model behaviour; AI research concerns
Information only. This link provides background or evidence; it is not a place to send a message.No contact action is available for this entry.
What it does
Tracks AI incidents and hazards to build an evidence base for policy, currently by automated monitoring of 150,000+ news articles daily via Event Registry, with LLM-based classification against the OECD's common reporting framework for AI incidents.
Honest assessment
Do not route an incident here yet — there is nowhere to put it. Use the AI Incident Database (incidentdatabase.ai), which accepts public submissions today and which OECD AIM draws on conceptually. Check back when OECD opens submissions.
Concerns it covers
Realised harms and near-misses across all severities — closest to an aviation-safety model. Empirical rather than theoretical; captures near-term harms far better than speculative catastrophic risk.
Government channel
Strong — designed explicitly to feed OECD and G7 policy work on incident reporting.