Beijing’s AI Policy Framework Ranks Among Nation’s Most Effective
A comprehensive evaluation of Beijing’s artificial intelligence (AI) policy landscape reveals a robust and strategically designed governance framework that has positioned the capital at the forefront of China’s AI revolution. The findings, published in the October 2021 issue of Global Science, Technology and Economy Outlook, demonstrate that Beijing’s AI policies are not only well-structured but also highly effective in fostering innovation, industrial growth, and technological leadership. Conducted by Ren Shasha, an assistant researcher at Beijing Electronic Holding Co., Ltd., the study employs a rigorous quantitative methodology to assess five key AI policy documents issued between 2017 and 2019. Using the Policy Modeling Consistency (PMC) index model—a comprehensive analytical framework that evaluates policy design across multiple dimensions—the research offers a data-driven assessment of Beijing’s policy strengths and identifies critical areas for refinement.
The study’s central conclusion is that Beijing’s AI policy architecture is among the most advanced in the country, with four out of five evaluated policies scoring in the “excellent” or “perfect” range according to the PMC index. The highest-rated policy, Guiding Opinions on Accelerating Technological Innovation to Cultivate the Artificial Intelligence Industry in Beijing (P1), achieved a near-perfect score of 9.00, earning the classification of “perfect.” Three other policies—Zhongguancun National Independent Innovation Demonstration Zone Artificial Intelligence Industry Cultivation Action Plan (2017–2020) (P2), Beijing Action Plan for Promoting the Integrated Development of Artificial Intelligence and Education (P3), and Work Plan on Promoting Artificial Intelligence Industry Development through Public Data Opening (P4)—received “excellent” ratings, with PMC scores of 8.20, 7.25, and 7.40, respectively. Only one policy, Fifteen Measures to Accelerate the Innovative Development of Zhongguancun Science City’s Artificial Intelligence (P5), fell into the “acceptable” category, with a score of 6.98. The average PMC index across all five policies stands at 7.77, indicating a strong overall performance and a well-coordinated policy environment that supports AI development from multiple angles.
This high level of policy effectiveness is particularly significant given Beijing’s pivotal role in China’s national AI strategy. As one of the country’s primary innovation hubs, Beijing hosts over 1,000 AI enterprises, accounting for approximately 26.5% of the national total as of 2019. The city has been designated a key node in China’s broader ambition to become a global leader in AI by 2030, as outlined in the State Council’s New Generation Artificial Intelligence Development Plan. In this context, the quality and coherence of local policy frameworks are essential for translating national goals into tangible outcomes. Ren Shasha’s analysis confirms that Beijing has made substantial progress in aligning its policy instruments with strategic objectives, creating a supportive ecosystem for research, commercialization, and industry-wide transformation.
The PMC index model used in the study is particularly well-suited for evaluating complex policy landscapes because it incorporates ten first-level variables and 47 second-level indicators, ensuring a multidimensional assessment. These variables include policy nature (e.g., predictive, advisory, regulatory), policy duration (short-, medium-, and long-term), policy perspective (macro vs. micro), policy evaluation (clarity of goals, feasibility of plans), policy domains (economic, social, political, technological, environmental), policy focus (technology innovation, talent development, market guidance), incentive mechanisms (funding, legal support, talent incentives), policy functions (technology breakthroughs, deep applications, demonstration leadership), policy recipients (government agencies, universities, enterprises), and policy transparency. Each variable is scored on a binary or normalized scale, and the composite PMC index is calculated to reflect the overall quality of policy design.
One of the most notable findings of the study is the high performance of Beijing’s policies in areas such as policy evaluation, policy function, and policy transparency. All five policies received perfect or near-perfect scores in these categories, indicating that they are clearly articulated, functionally comprehensive, and publicly accessible. For instance, the policies consistently define specific development goals, provide detailed implementation plans, and are grounded in sound rationale. They also emphasize multiple functional objectives, including technological breakthroughs, scientific innovation, deep application, and demonstration leadership. Moreover, all policies are publicly available, ensuring transparency and accountability in governance.
Equally impressive is the breadth of policy domains covered. The evaluated policies engage with economic, social, technological, and environmental dimensions, reflecting a holistic approach to AI development. For example, P1 and P2 emphasize technological innovation and industrial ecosystem building, while P3 focuses on the integration of AI into education, and P4 addresses the role of public data in driving AI applications. This multidomain engagement ensures that AI development is not siloed within a single sector but is instead integrated into broader societal and economic transformation.
However, the study also identifies several areas where Beijing’s AI policies could be further strengthened. The most significant shortcomings lie in policy nature, policy duration, and policy recipient inclusiveness. Specifically, the policies exhibit a notable lack of advisory and diagnostic elements. While they are strong in predictive, guiding, and descriptive functions, only one policy (P1) includes advisory content, and only one (P4) contains diagnostic language that identifies existing challenges, such as insufficient public data supply or weak AI integration in big data projects. The absence of diagnostic assessments limits the ability of policymakers to tailor interventions to specific bottlenecks, while the lack of advisory components reduces opportunities for stakeholder engagement and feedback.
Another critical gap is the limited emphasis on medium- and long-term development goals. Four of the five policies—P1, P2, P4, and P5—focus exclusively on short-term objectives (1–3 years), with no explicit provisions for medium (3–5 years) or long-term (5+ years) planning. Only P3 includes a multi-phase development timeline that spans short, medium, and long-term horizons. This short-term orientation may hinder strategic continuity and long-term investment, especially in a field like AI, where breakthroughs often require sustained research and development over extended periods. In contrast, national-level strategies, such as the State Council’s AI development plan, outline clear milestones for 2020, 2025, and 2030, suggesting that local policies should align with these longer-term visions.
Additionally, the scope of policy recipients is somewhat narrow, particularly in policies issued by sub-municipal or specialized agencies. For example, P2, issued by the Zhongguancun Science and Technology Park Management Committee, primarily targets entities within the Zhongguancun area, while P5, issued by the Haidian District People’s Government, is focused on local enterprises and institutions. Although these localized policies are valuable for targeted development, they may not fully address the needs of a city-wide or cross-sectoral AI ecosystem. In contrast, P1, issued by the Beijing Municipal Party Committee and Municipal Government, has a broader reach and higher inclusiveness, covering government departments, educational institutions, and both in- and out-of-park enterprises.
Despite these limitations, the overall policy framework demonstrates a high degree of sophistication and strategic coherence. The top-performing policy, P1, exemplifies best practices in policy design. It achieves high scores across nearly all dimensions, with particular strengths in policy evaluation, policy function, and policy transparency. Its only notable weakness is in policy duration, where it scores below average due to its focus on short-term goals ending in 2020. Nevertheless, its comprehensive approach to technology innovation, talent cultivation, industrial layout, and market guidance makes it a model for other jurisdictions.
P2, while slightly lower in overall score, reflects the specialized nature of zone-based innovation policies. Its lower score in policy recipients is a direct result of its geographic and institutional focus, but it compensates with strong performance in policy focus and incentive mechanisms, particularly in supporting technology transfer and industry-academia collaboration. P3, though rated “excellent,” shows room for improvement in economic and political engagement, as well as in incentive diversity. Its narrow focus on education limits its cross-sectoral impact, but it serves as an important precedent for domain-specific AI integration policies.
P4, the public data openness initiative, is notable for its innovative approach to leveraging government data assets to stimulate AI development. However, it underperforms in policy focus, particularly in areas such as market guidance, standard setting, intellectual property protection, and enterprise cultivation. These gaps suggest that while data access is a critical enabler, it must be accompanied by complementary policies that support commercialization, competition, and regulatory clarity. P5, the lowest-scoring policy, suffers from a narrow policy perspective, limited policy nature, and short-term orientation. As a district-level measure, it lacks the macro-level guidance needed to align with city-wide or national strategies.
In light of these findings, Ren Shasha proposes three key recommendations for improving Beijing’s AI policy framework. First, policymakers should enhance the advisory and diagnostic components of future policies. This includes conducting systematic assessments of current challenges, engaging with stakeholders to gather feedback, and incorporating evidence-based recommendations into policy formulation. Second, greater emphasis should be placed on medium- and long-term planning. Aligning local policies with national timelines (e.g., 2025 and 2030 milestones) would ensure strategic continuity and attract long-term investment. Third, the scope of policy recipients should be expanded to include a wider range of stakeholders, particularly in sectors such as healthcare, transportation, and manufacturing, where AI has transformative potential but remains under-regulated.
These recommendations are not merely academic suggestions but practical imperatives for sustaining Beijing’s leadership in AI. As global competition in artificial intelligence intensifies, the quality of policy governance will increasingly determine which regions can attract top talent, secure venture capital, and drive innovation at scale. Beijing’s current policy framework provides a solid foundation, but continued refinement is necessary to maintain its competitive edge.
Moreover, the study’s methodology offers a replicable model for evaluating AI policies in other cities and regions. The PMC index approach, which has been applied to sectors such as new energy vehicles, robotics, and big data, proves highly effective in capturing the multidimensional nature of technology policy. By combining text mining with structured variable analysis, it enables researchers and policymakers to move beyond qualitative descriptions and toward quantifiable, evidence-based assessments. This is particularly valuable in an era where AI policy is often shaped by hype and speculation rather than rigorous analysis.
In conclusion, Ren Shasha’s research provides a timely and insightful evaluation of Beijing’s AI policy landscape. It confirms that the city has established one of the most effective AI governance frameworks in China, characterized by clear goals, comprehensive functions, and broad domain coverage. At the same time, it highlights opportunities for improvement in policy diagnostics, long-term planning, and inclusiveness. As Beijing continues to evolve as a global AI hub, these insights will be invaluable for shaping policies that are not only technically sound but also socially inclusive and strategically sustainable.
Beijing’s AI Policy Framework Ranks Among Nation’s Most Effective
Ren Shasha, Beijing Electronic Holding Co., Ltd., Global Science, Technology and Economy Outlook, DOI: 10.3772/j.issn.1009-8623.2021.10.009