KidAID: A Dataset of Real World AI Incidents Involving Children and Adolescents

KidAID (Kids AI Incidents Database) is a curated dataset of real-world incidents in which AI systems measurably impacted children and adolescents, spanning child-facing products, general-audience platforms used by minors, and public-sector deployments affecting youth. Incidents are systematically collected from credible sources and normalized into structured fields such as system/deployer/developer, sector and setting, harm category, failure mode, severity, and evidence type. The dataset is designed to support governance and safety research, including trend analysis, risk pathway modeling, sociotechnical failure clustering, evaluation of safeguards, and benchmarking of child-specific safety interventions. It provides a foundation for reproducible measurement of child-related AI harms and for developing practical mitigations across the AI lifecycle.

儿童AI事件数据库（KidAID，Kids AI Incidents Database）是一套经过精心整理的真实世界事件数据集，收录了人工智能系统对儿童及青少年产生可量化影响的各类事件，覆盖面向儿童的产品、未成年人使用的通用受众平台，以及影响青年群体的公共部门人工智能部署场景。所有事件均从可信来源中系统性采集，并被规整为标准化结构化字段，涵盖系统/部署方/开发方、所属领域与应用场景、伤害类别、失效模式、严重程度以及证据类型等维度。该数据集旨在为人工智能治理与安全研究提供支撑，研究方向涵盖趋势分析、风险路径建模、社会技术失效聚类、安全保障措施评估以及针对儿童的安全干预措施的基准测试等。它为可复现地衡量与儿童相关的人工智能伤害，以及在人工智能全生命周期中制定切实可行的伤害缓解措施，提供了坚实的研究基础。