Abstract: The development of artificial intelligence is transforming from the era of generative AI to that of agentic AI. The hallmark of the agentic AI era is the emergence of AI agents. These agents possess the capability to autonomously plan, utilize tools, and interact with the external environment, shifting AI's focus fromwhat it can saytowhat it can do.The development of AI agents signifies not merely an expansion of technical capabilities but a profound transformation in the relationships between humans and technology, and between humans and organizations. Within this new systemic framework, the nature and content of national security risks have undergone significant changes. Domains such as military, political, and cybersecurity, as well as the security of critical infrastructure, are all experiencing marked shifts under the impact of AI agents. Security risks across these various domains manifest differently over short, medium, and long terms, reshaping the national security landscape at multiple levels.To address the challenges posed by AI agents, the United States, Europe, Russia, and Japan have accumulated valuable governance experience.Through precise positioning and agile governance strategies, China has achieved notable success in this area, offering a valuable reference for AI governance worldwide.
Keywords: Agentic AI; AI agents; security risks; AI governance
Authors：Feng Shuai, Xie Jianzhi
Source：Feng, S., & Xie, J. Z. (2026). The challenge of AI agents: National security risks and governance in the era of agentic AI. Academic Journal of Russian Studies, (4).
Publish date：2026/8/16
Original link: 
智能体的挑战：代理式人工智能时代的国家安全风险与治理_封帅.pdf
