中国国安部长：AI构成政治安全和泄密风险

中国国家安全部长陈一新日前发表署名文章，指出人工智能存在六大风险隐患1。这些风险包括系统性影响政治安全环境、颠覆性升级网络攻防态势、规模性放大窃密泄密风险、AI垄断性加剧科技发展失衡、结构性冲击社会治理体系以及根本性变革军事斗争形态1。

陈一新强调，敌对势力可利用深度伪造影音、AI生成图文、智慧水军等技术对中国实施舆论战和认知战1，这些行为直接威胁中国的政治、制度和意识形态安全。他指出，网络攻防已进入"漏洞工业化、攻防全自动、AI对AI"的新阶段，以美国Claude Mythos及GPT-5.5-Cyber模型为代表1。

为应对这些挑战，陈一新主张中国应以高水平安全、高效能治理推动AI产业健康有序发展1，并加快构建完善的AI安全监管平台，定期发布安全风险防范指南1。

Chen Yixin, head of China's Ministry of State Security, published a signed article on September 13 outlining six major risks posed by artificial intelligence to China's national security 1. According to Chen, AI poses systemic threats to the political security environment, including capabilities for large-scale information theft and unauthorized disclosure 1.

Chen identified hostile forces and ill-intentioned actors as potential users of AI technologies to conduct "information warfare and cognitive warfare" against China through deepfake audio and video, AI-generated images and text, and coordinated online campaigns 1. He stressed that such operations directly threaten China's political system, institutional framework, and ideological security 1. Beyond political risks, Chen outlined five additional concerns: disruption of cybersecurity defenses, expansion of espionage and data breaches, technological inequality exacerbated by AI monopolization, structural damage to social governance systems, and fundamental transformation of military conflict 1.

To address these challenges, Chen called for accelerating the development of an AI security risk prevention and control system 1. He emphasized the need to balance "high-level security with high-efficiency governance" to ensure AI industry development proceeds in a "healthy and orderly manner" 1. The article referenced American AI models such as Claude Mythos and GPT-5.5-Cyber as examples of how cyber attacks have entered a new phase characterized by "industrial-scale vulnerabilities, fully automated attack and defense operations, and AI versus AI confrontations" 1. China should establish comprehensive AI security regulatory platforms and periodically release risk prevention guidelines, according to Chen's proposal 1.