butterflylabs/ai-incidents

Butterfly Labs AI事件数据库是一个持续更新的开放数据集，汇集了来自AI事件数据库（AIID）和16+研究、政策及民间社会RSS源的AI安全事件数据。涵盖语言模型、自主系统、机器人技术、深度伪造、监控、偏见、隐私和无障碍性失败等多个领域。每条记录包含标题、摘要、来源URL、类别、严重性、涉嫌的开发者/部署者/受害方以及ISO日期字段。数据集大小为2641条记录，采用CC-BY-4.0许可协议。

A continuously-updated, open dataset of AI safety incidents aggregated from the AI Incident Database (AIID) and 16+ research, policy, and civil-society RSS feeds. Covers language models, autonomous systems, robotics, deepfakes, surveillance, bias, privacy, and accessibility failures. Each record includes title, summary, source URL, category, severity, alleged developers/deployers/harmed parties, and ISO date fields. The dataset contains 2641 records and is licensed under CC-BY-4.0.