AI Incident Database (Responsible AI Collaborative)

What it does

Public, open-source, citable index of realised AI harms and near-harms, modelled on aviation and cybersecurity incident databases. Community-submitted incidents are editorially reviewed and indexed with reports, entities and classifications. Recent entries span AI medical transcription misuse, deepfake tooling, chatbots implicated in violence planning, and law-enforcement technology abuse.

Honest assessment

Small, volunteer-heavy, and genuinely reads its submissions — one of the few places in this entire list where an unaffiliated person's report reliably gets human attention and can end up in the permanent public record. If you have a specific realised harm, start here. Editorial review takes time and marginal submissions do get rejected.

Concerns it covers

Realised harms of any severity, empirically grounded. Strongest on near-term harms; increasingly captures misuse and agentic failures. Not a theoretical-risk organisation.

Government channel

Indirect but real — AIID entries are cited by OECD AIM, academic literature, journalists and regulators. Getting an incident indexed here makes it citable everywhere else.