Accelerating the Building of National Security Capacity in the Field of Artificial Intelligence

At present, artificial intelligence has evolved from an important productive tool into a strategic technology that profoundly affects the overall development of the nation. Its scope of application is accelerating its penetration from the industrial sphere into key areas such as economic operation, public services, and social governance, leading to systemic changes in the connotation, structure, and means of realization of national security. The Fourth Plenary Session of the 20th Central Committee of the Communist Party of China proposed to "improve the national security system and strengthen the building of national security capacity in key areas." [1]1 The Fourth Plenary Session of the 20th Central Committee of the CPC was held in July 2024, focusing on further deepening reform comprehensively to advance Chinese-path modernization. Against this background, accelerating the building of national security capacity in the field of artificial intelligence, and integrating security requirements into the entire process of technological research and development, application promotion, and governance operation through institutionalized and systematic means, is not only an inevitable requirement for advancing the modernization of the national security system and capacity during the period of the 14th Five-Year Plan, but also an inherent need to promote the safe, orderly, and sustainable development of artificial intelligence and to better serve high-quality economic and social development. [2]2 The 14th Five-Year Plan (2021–2025) is China's national economic and social development plan for that period.

当前，人工智能已从重要生产工具演进为深刻影响国家发展全局的战略技术，其应用范围正从产业领域加速向经济运行、公共服务和社会治理等关键领域全面渗透，国家安全的内涵、结构和实现方式出现了系统性变化。党的二十届四中全会提出，要“健全国家安全体系，加强重点领域国家安全能力建设”。在此背景下，加快人工智能领域国家安全能力建设，通过制度化、体系化的方式将安全要求融入技术研发、应用推广与治理运行全过程，既是“十五五”时期推进国家安全体系和能力现代化的必然要求，也是推动人工智能安全、有序、可持续发展，更好服务经济社会高质量发展的内在需要。

The Contemporary Logic of Accelerating the Building of National Security Capacity in the Field of Artificial Intelligence

加快人工智能领域国家安全能力建设的时代逻辑

The development and application of artificial intelligence technology are deeply embedding key elements such as algorithms, data, and models into the national operation and governance system, making them a new type of infrastructure supporting economic and social development. This process not only reshapes the basic modes of information generation, decision-making, and execution feedback but also drives a fundamental transformation of the foundation on which national security operates. National security no longer depends solely on the allocation of traditional human resources and institutional arrangements; it increasingly relies on the capacity for continuous cognition, dynamic control, and adaptive adjustment of the overall behavior of complex intelligent technology systems. From macroeconomic regulation to public service provision, from social governance innovation to public security maintenance, artificial intelligence is comprehensively reconstructing the internal logic of national governance, prompting a shift in the governance paradigm from one centered on event response to one gradually centered on system operation.

人工智能技术的发展与应用，正在将算法、数据和模型等关键要素深度嵌入国家运行与治理体系，使之成为支撑经济社会发展的新型基础设施。这一进程不仅重塑了信息生成、决策制定与执行反馈的基本方式，更推动了国家安全运行基础的根本性变革。国家安全不再仅仅依赖于传统的人力资源配置与制度安排，而是日益取决于对复杂智能技术系统整体行为的持续认知、动态掌控与调适能力。从宏观经济调控到公共服务供给，从社会治理创新到公共安全维护，人工智能正在全方位重构国家治理的内在逻辑，促使治理范式从以事件响应为核心逐步向以系统运行为核心转变。

Against this background, the connotation of national security and its means of realization have also undergone structural adjustments. The focus of security has shifted from being mainly manifested as passive prevention against single risks or unexpected incidents, to full-time monitoring, dynamic optimization, and forward-looking guidance based on intelligent technology. The security of artificial intelligence systems is highly dependent on the organic matching between technological choices, institutional design, and governance capacity. Lacking systematic capacity support, during rapid expansion and iteration, technological systems can easily amplify the local deviations of existing governance strategies, thereby turning efficiency advantages into systemic risks and ultimately endangering the stability and sustainability of national security operations.

在此背景下，国家安全的内涵及其实现方式也发生了结构性调整。安全的重心已从主要体现为对单一风险或突发事件的被动防范，转向基于智能技术的全时监测、动态优化和前瞻引导。人工智能系统的安全性，高度依赖于技术选择、制度设计与治理能力之间的有机匹配。如果缺乏系统性的能力支撑，技术系统在快速扩展与迭代过程中，容易放大既有治理策略的局部偏差，致使效率优势转化为系统性风险，最终危及国家安全运行的稳定性与可持续性。

General Secretary Xi Jinping emphasized, "We must thoroughly implement the holistic approach to national security, and persist in consolidating security through development and seeking development through security." [3]3 The holistic approach to national security (总体国家安全观) is a strategic concept proposed by Xi Jinping that integrates political, economic, military, cultural, social, and other security domains. The holistic approach to national security emphasizes adhering to a systemic perspective and promoting the modernization of the national security system and capacity. From this strategic perspective, accelerating the building of national security capacity in the field of artificial intelligence is essentially complying with the inherent requirement of the transformation of the logic of national security from passive defense to proactive governance, and from post-event handling to pre-event prevention. By incorporating artificial intelligence security into the national security system for overall planning, and integrating scattered technological elements, governance resources, and institutional arrangements into an organic whole, it helps to transform the inherent uncertainty of technology into a governable object, ensuring the stable and orderly development of artificial intelligence within the overall framework of national governance. This is both an objective requirement of the changing stage of artificial intelligence technology development and an inevitable choice for the proactive upgrading of national security governance methods.

习近平总书记强调，“要深入贯彻总体国家安全观，坚持在发展中固安全、在安全中谋发展”。总体国家安全观强调坚持系统观念，推动国家安全体系和能力现代化。从这一战略视角看，加快人工智能领域国家安全能力建设，本质上是顺应国家安全逻辑从被动防御向主动治理、从事后处置向事前预防转型的内在要求。通过将人工智能安全纳入国家安全体系进行整体谋划，把分散的技术要素、治理资源与制度安排整合为有机整体，有助于将技术内在的不确定性转化为可治理对象，确保人工智能在国家治理总体框架内实现稳健、有序发展。这既是人工智能技术发展阶段变化的客观要求，也是国家安全治理方式主动升级的必然选择。

Consolidating the Technological Support for Building National Security Capacity in the Field of Artificial Intelligence

夯实人工智能领域国家安全能力建设的技术支撑

Whether the national security capacity system in the field of artificial intelligence can operate effectively depends first on whether there is a stable, reliable, and independently controllable technological foundation. As a technological form highly dependent on algorithm models, computing power resources, and data elements, the security of artificial intelligence systems is not entirely determined by the performance indicators of individual technologies, but also by the capacity for continuous control, transparent understanding, and credible verification of the key technological links in the entire chain. Only when a stable control capability is formed at the technological level can the national security system maintain sufficient strategic focus and operational resilience in a complex and changing international environment and rapidly iterating technological evolution.

人工智能领域国家安全能力体系能否有效运行，首先取决于是否具备稳定、可靠、自主可控的技术基础。作为一种高度依赖算法模型、算力资源和数据要素的技术形态，人工智能系统的安全性不完全取决于单项技术的性能指标，还取决于对全链条关键技术环节的持续掌控、透明理解与可信验证能力。只有在技术层面形成稳定的掌控能力，国家安全体系才能在复杂多变的国际环境和快速迭代的技术演进中保持足够的战略定力与运行韧性。

From the perspective of capacity building, technological controllability is not a simple restriction on technological development, but a structural capacity arrangement focused on the overall situation and the long term. Its core lies in enhancing the systematic control and leading capacity over key links such as core algorithms, advanced computing power, high-quality data, and basic software by improving the layout of the national technology system, ensuring that artificial intelligence systems remain behaviorally predictable, processually traceable, and resultantly verifiable in diverse application scenarios. This controllability is reflected both in the strategic overall allocation of key resources and in the institutional guidance of the direction of technological evolution through standards, norms, and ethical guidelines. Embedding security and trustworthiness requirements deeply in the initial stage of technological research and development and system design can achieve pre-emptive risk identification and source governance, thereby significantly enhancing the overall robustness of artificial intelligence technology applications.

从能力建设角度看，技术可控并不是对技术发展的简单限制，而是一种着眼于全局和长远的结构性能力安排。其核心在于通过完善国家技术体系布局，增强对核心算法、先进算力、高质量数据及基础软件等关键环节的系统性掌控与主导能力，确保人工智能系统在多样化的应用场景中始终保持行为可预期、过程可追溯、结果可校验。这种可控性，既体现在对关键资源的战略性统筹配置上，也体现在通过标准、规范与伦理指南对技术演进方向进行制度性引导上。在技术研发与系统设计的初始阶段即深度嵌入安全与可信要求，可实现风险的前置识别与源头治理，从而显著提升人工智能技术应用的整体稳健性。

At the same time, with the deep application of artificial intelligence in areas such as critical infrastructure, public decision support, and social credit systems, the credibility, fairness, and explainability of the technological system itself are gradually becoming important components of national security capacity. Establishing and improving normalized mechanisms covering algorithm design, model training, output evaluation, and security testing, so that artificial intelligence systems possess characteristics of logical understandability, decision traceability, and error correctability during operation, helps to enhance the governance主体的 ability to manage technological complexity. This model, which supports subjective governance judgments with objective technical norms, is conducive to promoting the transformation of artificial intelligence security governance from mainly relying on external mandatory intervention to gradually relying on technology's inherent constraint mechanisms and standardized operation.

与此同时，随着人工智能在关键基础设施、公共决策支持和社会信用体系等领域的深入应用，技术系统本身的可信性、公平性与可解释性逐渐成为国家安全能力的重要组成部分。建立健全覆盖算法设计、模型训练、输出评估与安全测试等常态化机制，人工智能系统在运行过程中具备逻辑可理解、决策可追溯、错误可校正的特征，有助于增强治理主体对技术复杂性的驾驭能力。这种以客观技术规范支撑主观治理判断的模式，有利于推动人工智能安全治理从主要依赖外部强制性干预，逐步转向依靠技术内生的约束机制与规范运行。

It is necessary to further recognize that the technological foundation of artificial intelligence security capacity does not exist in isolation but is embedded within the national overall scientific and technological innovation system and industrial ecosystem. General Secretary Xi Jinping pointed out: "We must vigorously strengthen modern engineering and technological science research that integrates multiple disciplines, driving the development of basic sciences and engineering technology to form a complete modern science and technology system." [4]4 This quotation is from Xi Jinping's speech at a symposium on scientists in September 2020. Only by strengthening the connection between basic research, applied research, and industrial development, and promoting cross-disciplinary and cross-field collaborative technological breakthroughs and resource integration, can we enhance the overall resilience, adaptive capacity, and sustained innovation vitality of China's artificial intelligence technology system at a higher level, providing continuous and reliable technological support for the modernization of the national security system and capacity.

需要进一步看到，人工智能安全能力的技术基座并非孤立存在，而是嵌于国家整体科技创新体系和产业生态之中。习近平总书记指出：“要大力加强多学科融合的现代工程和技术科学研究，带动基础科学和工程技术发展，形成完整的现代科学技术体系。”只有通过强化基础研究、应用研究与产业化开发之间的衔接，推动跨学科、跨领域的技术协同攻关与资源整合，才能在更高层面提升我国人工智能技术体系的整体韧性、自适应能力与持续创新活力，为国家安全体系与能力现代化提供持续、可靠的技术支撑。

Improving the Institutional Guarantees for Building National Security Capacity in the Field of Artificial Intelligence

健全人工智能领域国家安全能力建设的制度保障

If a solid technological foundation constitutes the hard support for building national security capacity in the field of artificial intelligence, then institutions and governance mechanisms determine whether these capacities can operate stably and continue to be effective. Artificial intelligence security does not originate from some isolated technological breakthrough, but is a comprehensive governance issue triggered by the deep embedding of complex technological systems into the national governance structure. Therefore, the key to constructing a national security capacity system in the field of artificial intelligence lies not in the simple accumulation of security rules and institutional numbers, but in transforming scattered security requirements into sustainable governance capacity through high-level institutional coordination, mechanism innovation, and process reengineering, and truly embedding them into the national governance system and operational processes.

如果说坚实的技术基础构成了人工智能领域国家安全能力建设的硬支撑，那么制度和治理机制则决定着这些能力能否稳定运行并持续发挥效能。人工智能安全并非源于某个孤立的技术突破，而是复杂技术系统深度嵌入国家治理结构后引发的综合治理命题。因此，构建人工智能领域国家安全能力体系，其关键不在于安全规则与机构数量的简单叠加，而在于通过高水平的制度协同、机制创新与流程再造，使分散的安全要求转化为可持续运行的治理能力，并真正嵌入国家治理体系和运行过程之中。

First, in the field of artificial intelligence, the effective generation of security capacity depends on clear governance boundaries and a stable responsibility structure. This requires, under the guidance of the holistic approach to national security, further clarifying the boundaries of government regulation, corporate principal responsibility, industry self-discipline, and social supervision. By simultaneously embedding and integrally deploying security objectives in science and technology planning, industrial policies, industry standards, and public governance practices, security requirements are pre-positioned and integrated into key links such as technological layout, project decision-making, and application promotion. This institutional embedding does not mean starting from scratch, but forming stable interfaces within the existing governance system, enabling artificial intelligence security requirements to continuously play a role in daily decision-making, execution, and supervision processes, thereby gradually transforming into normalized governance capacity.

首先，在人工智能领域，安全能力的有效生成依赖清晰的治理边界和稳定的责任结构。这要求在总体国家安全观引领下，进一步明确政府监管、企业主体责任、行业自律与社会监督的边界，通过在科技规划、产业政策、行业标准与公共治理实践中同步嵌入、一体部署安全目标，将安全要求前置融入技术布局、项目决策和应用推广等关键环节。这种制度嵌入并非另起炉灶，而是在既有治理体系中形成稳定接口，使人工智能安全要求能够在日常决策、执行和监督过程中持续发挥作用，从而逐步转化为常态化的治理能力。

Second, the cross-domain, cross-level, and cross-actor characteristics of artificial intelligence technology and its applications determine that its security governance inherently entails a high degree of complexity and a need for coordination. In the absence of effective coordination, it is easy to produce overlapping responsibilities, information silos, and disjointed actions, thereby weakening the overall security defense line. By improving normalized cross-departmental, cross-regional, and cross-domain coordination mechanisms and information-sharing mechanisms, and by promoting coordinated interaction among governance actors in policy formulation, risk assessment, compliance oversight, and emergency response, it is possible to reduce the institutional friction arising from fragmented governance and to steadily upgrade the AI security capability system from scattered departmental responses to a nationwide, unified, systematic, and intensive operation.

其次，人工智能技术及其应用的跨领域、跨层级、跨主体特征，决定了其安全治理天然具有高度复杂性和协同需求。若缺乏有效协同，容易导致职责重叠、信息壁垒与行动脱节，削弱整体安全防线。通过健全跨部门、跨地区、跨领域的常态化协同机制与信息共享机制，推动治理主体在政策制定、风险评估、合规监管与应急响应中的协同联动，有助于减少治理碎片化带来的制度摩擦，推动人工智能安全能力体系从各部门的分散应对，稳步提升为全国一盘棋的体系化、集约化运行。

Finally, in the face of the distinctive characteristics of rapid iteration in AI technology, dynamic migration of application scenarios, and continuous evolution of risk forms, the institutional system must possess the wisdom and flexibility for agile response and dynamic adaptation. Therefore, the design of governance mechanisms should closely revolve around key lifecycle stages such as algorithm updates, model optimization, system upgrades, and scenario expansion, and actively explore and establish an adaptive governance framework that can evolve and adjust flexibly in tandem with technological development — for example, piloting "sandbox regulation," [5]5 "Sandbox regulation" (沙盒监管) refers to a regulatory approach, originally used in the financial technology sector, in which authorities allow controlled, time-limited testing of innovative products or services under relaxed rules within a defined "sandbox" environment, enabling real-world experimentation while managing risks. promoting advance standard research, and adopting tiered compliance measures. This kind of dynamic institutional adaptation does not lower the security baseline; rather, through continuous learning about the laws of technological operation and precise calibration of governance strategies, it achieves a higher level and more effective risk prevention and control.

最后，面对人工智能技术快速迭代、应用场景动态迁移、风险形态持续演化的鲜明特点，制度体系必须具备敏捷响应与动态调适的智慧与弹性。因此，治理机制的设计应紧密围绕算法更新、模型优化、系统升级与场景拓展等关键生命周期环节，积极探索并建立能够随技术发展同步演进、灵活调整的适应性治理框架，例如试点“沙盒监管”、推行标准预研、采用阶梯式合规等手段。这种制度上的动态适配，并非降低安全底线，而是通过对技术运行规律的持续学习与治理策略的精准校准，实现更高水平、更有效率的风险防控。

(Author: Zhang Wei, Professor at the School of Public Administration, Sichuan University)

（作者：张 伟，系四川大学公共管理学院教授）