The Children's AI Harm Registry

The Children's AI Harm Registry is a structured, civilian-maintained dataset documenting cases in which artificial intelligence systems caused or contributed to harm affecting people under the age of 18. Version 2.1 contains 446 documented cases spanning 2016 to 2026 across 46 named countries plus multi-jurisdictional cases, and a separate Aggregate and Systemic Incidents layer of 22 population-scale findings. Severity: Fatal 51, Severe 279, Moderate 109, Low 6. Verification: 58 Verified (2+ independent sources), 388 Documented (single source). What changed in v2.1: 16 cases added that surfaced between the v2.0 source cutoff (May 22, 2026) and September 15, 2026, plus four labeled pre-cutoff gaps. Additions include a coroner-sourced UK fatality naming ChatGPT and verified school deepfake cases in Argentina, France and Germany. Ten aggregate entries added, including the Italian Garante fine of Character.AI, Florida v. OpenAI and Altman, the 42-state attorney general subpoena to OpenAI, and the Amsterdam District Court injunction against Grok. No v2.0 row was modified, reordered or renumbered. Files: Childrens_AI_Harm_Registry_v2.1.xlsx (About, Registry, Aggregate Incidents tabs; read the About tab before citing), childrens_ai_harm_registry_v2.1.csv, aggregate_incidents_v2.1.csv, and registry_v2.1_additions.csv (the 16 rows added in this version). Correction: the v2.0 workbook cited the v1.0 DOI in its About tab. Cite each version by its own DOI. Maintainer: Chidinma Omai, Raised to Resist.