AI Incident Database

Anyone can report an incident — nothing appears here until a moderator reviews it. Published entries get a permanent identifier (RAI-YYYY-NNNN) and are hash-chained so the database's integrity can be independently checked at GET /api/incident-db/verify. Need to check a specific model/provider pair programmatically before deploying it — e.g. from CI/CD? See GET /api/incident-db/check (PRO plan required for production use).