MIT AI Incident Tracker (AI Risk Repository)

What it does

Nothing directly from the public. It ingests and classifies incidents from the AI Incident Database — 1,600+ incidents classified by risk domain, cause, harm type (10 categories including physical harm, financial loss and human rights), severity, entity and timing.

Honest assessment

Important for a playbook as a multiplier argument, not a separate action: it demonstrates that a single accepted AIID submission propagates into academic and policy-facing infrastructure read by researchers and governments. This is the concrete payoff that makes AIID submission worth the effort.

How to file

Submit to the AI Incident Database; if accepted there, your incident flows into this tracker automatically and is re-classified against academic taxonomies.

Format

N/A

Timing

Periodic updates (e.g. a June 2026 update was published).

What happens after

Classified and visualised in timelines, overviews, detailed records and national security impact assessments.

What it accepts

Nothing directly from the public. It ingests and classifies incidents from the AI Incident Database — 1,600+ incidents classified by risk domain, cause, harm type (10 categories including physical harm, financial loss and human rights), severity, entity and timing.

What it does not accept

Direct submissions. The page itself directs people to submit to AIID instead.

Operated by

MIT FutureTech / MIT AI Risk Repository, with Arcola AI; classification uses the MIT Risk Repository and CSET AI Harm Taxonomy