Dataset for A Competing-Hazards Systematization of Loss of Control in Autonomous Agents

Dataset for the paper "A Competing-Hazards Systematization of Loss of Control in Autonomous Agents." Two CSV files: incidents.csv (22 publicly documented incidents in which a language-model agent acted outside its sanctioned scope) and evaluations.csv (102 multi-step agent-safety evaluations). Each record is scored on the paper's audit questions, with a supporting source and an evidence grade for every field. The measurement model, the search and inclusion protocol, and the identification rule are specified in full in the paper.