Rand兰德：2025年直面AI-网络安全交汇挑战：应对新兴AI网络攻击威胁的政府决策结构化方法研究报告（英文版）

The emergence of increasingly capable artificial intelligence (AI) models and agents in recent years has raised concerns about the potential harms they can bring. Such models threaten to uplift malicious actors—providing them with capabilities they would likely not possess on their own or expanding their capacity to do harm.1 These concerns extend to diverse fields, such as biosecurity, proliferation of weapons of mass destruction, and AI-enabled offensive cyberspace operations (OCOs).2 In th