Rand兰德：2025年计算治理博弈论模型中的策略与检测缺口研究报告（英文版）

This report documents research and analysis conducted as part of a study entitled “Broadly Capable AI [Artificial Intelligence] Threats and Mitigation (BCAIT): How Can a Cloud Compute Provider Know If a Big AI Training Run Is Happening on Their System?” sponsored by the Technology and Security Policy Center (TASP) in the RAND Global and Emerging Risks division. The purpose of the study was to investigate detection and monitoring mechanisms that cloud service providers could employ to identify