Incident databases collect public reports of AI failures and attacks so you can learn from other organizations' mistakes. Use them to build threat scenarios and to benchmark your own incident taxonomy. Some regimes, such as the EU AI Act, also require reporting serious incidents to authorities.
Read in this order
- OfficialAI Incident Database (AIID)Start here: the main public database of AI incidents.
- Open sourceMITRE ATLAS data, including case studiesReal attack case studies mapped to adversary techniques.
- OfficialArticle 73: Reporting of serious incidents (EU AI Act Service Desk)When a serious incident must be reported under the EU AI Act.