The PAI incident registry is built on anonymised case data contributed by practitioners. Your incident data helps the community understand what goes wrong in production AI, informs PSF domain evolution, and is published in aggregate in our annual incident pattern analysis.
All identifying information — organisation name, system name, model version, and any personnel details — is removed before data enters the registry. Contributions are reviewed by two PAI team members before publication. You will be contacted before any case data is used in a published analysis.
Send your incident report to research@productionai.institutewith the subject line “Incident contribution”. You can use free-form text; we will structure the data during review. If you prefer a structured template, reply to this email and we will send you one.
Contributors to published analyses are acknowledged by name or pseudonym at their choice. If you contributed to an incident that led to a significant change in PSF domain design, we will note this in the relevant framework note.