The AI Incident Database (AIID) was introduced on 18 November 2020 by the Partnership on AI, with Sean McGregor as its primary author. It is a searchable, crowdsourced repository of cases where intelligent systems have caused real-world problems with safety, fairness, or other harms - drawing on news, trade, and academic reports across domains including transportation, energy, healthcare, law enforcement, and recruiting.
The project deliberately borrows from older fields that learned to log their failures. It echoes aviation’s distinction between “incidents,” where the risk of an accident rises, and “accidents,” which cause substantial damage - a discipline the founders credit with cutting aviation fatalities dramatically since 1970. It also mirrors the cybersecurity world’s Common Vulnerabilities and Exposures system, giving each entry an identifier so that practitioners can reference, compare, and learn from past failures rather than repeating them.
Governance of the database later moved to the Responsible AI Collaborative, a non-profit chartered specifically to maintain and expand it, with the Partnership on AI as a founding sponsor. By cataloging harms in a structured, citable way, the AIID became a standard feedstock for researchers, journalists, and policymakers studying how AI systems fail in practice - and a reminder that many AI risks are not hypothetical but already documented.