Grok Image Generation Safety Audit: Incident Dataset (43 Prompts, 100% Failure Rate)
收藏Figshare2026-03-13 更新2026-04-28 收录
下载链接：
https://figshare.com/articles/dataset/Grok_Image_Generation_Safety_Audit_Incident_Dataset_43_Prompts_100_Failure_Rate_/31724035
下载链接
链接失效反馈官方服务：
资源简介：
Structured dataset documenting 43 safety filter test incidents across Grok's image generation system on X (formerly Twitter), conducted between December 2024 and January 2025. The dataset records a 100% safety filter failure rate across all tested categories, including generation of photorealistic images of named minors, nonconsensual intimate imagery of public figures, and political disinformation content. Each incident includes prompt text, output description, failure category, and date. This dataset is the evidentiary basis for: Lefkowitz, D. (2026). "Grok Image Generation Governance Audit: Targeted Sexualization on X." SSRN Working Paper. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6123306
创建时间：
2026-03-13
