A public catalogue of AI and other automated decision systems used by governments and companies, and of what happened to the people they reached. Each record is built from court rulings, inquiries, regulators' findings and journalism, and every claim is sourced.
For journalists, researchers, lawyers and policymakers. The data can be cited, downloaded and reused freely (CC-BY 4.0).
In-depth cases
6 full forensic case-files: sources, timeline, and harm coded on seven dimensions.
- S-001 · Fixed rulesSyRIDutch state inter-agency risk-scoring of welfare recipients, halted by the District Court of The Hague
- S-004 · Fixed rulesToeslagenaffaireDutch childcare-benefits fraud classification that wrongly accused tens of thousands of families
- S-005 · AI (machine learning)Optum Impact ProCommercial healthcare risk algorithm that under-estimated the health needs of Black patients
Registry
17 verified incidents: short records, each with its sources.
- R-1976Face recognition match leads to Robert Williams's wrongful arrest
- R-2168Rite Aid face recognition falsely flags customers
- R-2579Google's image scanning reports parents for medical photos
- R-3573ShotSpotter alerts used to justify stops and frisks
- R-3703Rikunabi sells predictions of which students will decline job offers
Work in progress: 76 further records are in review, from 4,934 articles screened. Records are published only once checked. How cases are documented →