沙利文发布《构建AI主权，促进AI普惠》立场报告
Frost & Sullivan Releases the Policy Position Paper " AI Sovereignty For an Inclusive Intelligent Society"
人工智能主权正从学术概念走向全球政策实践。过去两年，全球已有一百余个国家发布国家人工智能战略，自主、安全与供应链韧性成为各国共同关切。沙利文在本份立场报告中提出，人工智能主权无关封闭，其本质是在开放体系中守住选择权，即一个国家对关键资源与关键规则是否仍保有最终决定权。报告将这一能力拆解为三个层次、六个要素：底层的能源、算力与数据构成物理与资源架构，中层的模型与应用承担技术转化与价值实现，上层的人工智能发展主权则决定技术路线、优先领域与合作伙伴的自主选择。
在此基础上，报告构建了AISDI人工智能主权发展指数，以六个维度、三十项核心指标与三层加权体系，对全球195个国家完成了系统测算。加权体系中，基础资源占50%，中层技术占30%，上层治理占20%。评估结果显示，拉开国家差距的并非资源禀赋的多寡，而是三层能力之间的耦合程度：多数国家的战略与法规建设快于电力、算力与数据基础设施的积累，因而长期停留在进阶级。据此，报告将各国划分为基础级、进阶级与成熟级三个成熟度层级，并指出人工智能主权并无单一路径，只有与自身成熟度相匹配的优先次序。
面向落地，报告提出画像定位、工具选择、分期配置与动态调适的四步路径生成方法，并配套覆盖六大要素的116项行动工具箱，从对内建设能力与对外塑造规则两个方向，为不同层级国家提供可直接调用的政策措施。报告同时建立风险管控与价值创造的双重机制，五类核心风险各自对应明确的应对路径，公共价值、产业价值与生态溢出三级回报则为主权建设的持续演进提供支撑。报告最终指出，自主可控并非终点，人工智能主权的更高目标是人工智能普惠：主权是普惠的前提，普惠是主权更高的目标，而连接二者的是合作。
AI sovereignty is moving rapidly from an academic concept to global policy practice. Over the past two years, more than one hundred countries have issued national AI strategies, making autonomy, security and supply chain resilience a shared concern. In this position paper, Frost & Sullivan argues that AI sovereignty is not about isolation. Its essence is the preservation of choice within an open system, namely whether a country retains ultimate decision-making authority over critical resources and rules. The report resolves this capability into six elements across three layers: energy, computing power and data form the foundational physical and resource architecture; models and applications carry technological conversion and value realisation at the intermediate layer; and AI development sovereignty at the upper layer governs the independent choice of technology pathways, priority sectors and international partners.
Building on this framework, the report establishes the AI Sovereignty Development Index, or AISDI, comprising six dimensions, thirty core indicators and a three-layer weighting system, and applies it to all 195 countries worldwide. Foundational resources account for 50 percent of the weighting, intermediate technologies for 30 percent, and upper-layer governance for 20 percent. The assessment shows that what separates countries is not the abundance of their resource endowments, but the degree to which the three layers of capability are coupled. In most countries, strategies and regulations advance faster than power, computing and data infrastructure accumulate, leaving them at the advanced level for an extended period. On this basis, the report classifies countries into basic, advanced and mature tiers, and concludes that there is no single path to AI sovereignty, only a sequence of priorities matched to each country’s own maturity.
For implementation, the report proposes a four-step pathway generation method covering profile positioning, tool selection, phasing configuration and dynamic adaptation, supported by a 116-item action toolkit spanning all six elements. The toolkit offers directly callable policy measures for countries at every tier, organised along two directions: internal capability building and external rule-shaping. The report also establishes a dual mechanism of risk control and value creation, in which five categories of core risk are each matched to a defined response, while returns at three levels, comprising public value, industrial value and ecosystem spillovers, sustain the continued evolution of sovereignty building. The report concludes that autonomous control is not the destination. The higher objective of AI sovereignty is AI equity: sovereignty is the prerequisite for equity, equity is the higher goal of sovereignty, and cooperation is what connects the two.
点击以下链接，查看节选版 报告：
如果您对人工智能主权及全球人工智能治理有进一步的研究需求，请联系我们：
沙利文 李女士
E-mail: livia.li@frostchina.com
