China - Beijing - AI Innovation Measures (2023)

Overview

The Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence, officially issued by the Beijing Municipal People's Government Office, represent a strategic legislative effort to solidify Beijing's position as a global leader in artificial intelligence innovation. Enacted on May 23, 2023, and publicly released on May 30, 2023, these measures are designed to provide a comprehensive framework for fostering the growth and healthy development of the general AI sector within the city. The overarching goal is to fully leverage the government's guiding role and the catalytic effect of innovation platforms, thereby integrating innovation resources, optimizing the allocation of essential factors, and cultivating a robust innovation ecosystem. This initiative is explicitly aimed at achieving innovation leadership and ensuring the rational and healthy advancement of general artificial intelligence in Beijing, while also emphasizing the critical importance of risk prevention in this rapidly evolving field. The measures are a direct response to national directives, including the call from the 20th National Congress of the Communist Party of China to build a modernized industrial system and develop strategic emerging industries like AI, as well as the April 2023 Politburo meeting's emphasis on general AI development, innovation ecosystems, and risk mitigation.

This policy instrument is not merely a set of recommendations but a binding regulation that outlines concrete steps and strategic directions for various stakeholders, including government agencies, research institutions, and enterprises. It seeks to address key challenges and bottlenecks in the AI industry's development, particularly concerning computing power, data, algorithms, applications, and regulatory oversight. By focusing on these core elements, the measures aim to break through existing barriers and propel Beijing's AI innovation forward. The document underscores Beijing's existing strengths in AI, such as its leading research institutions and innovative enterprises that have already produced significant original achievements in large models, brain-inspired intelligence, and embodied intelligence. The measures are intended to build upon these foundational advantages, further accelerating the city's journey towards becoming a globally influential hub for AI innovation and application. The emphasis on general AI reflects a forward-looking approach, recognizing the transformative potential of advanced AI models like those based on the Transformer architecture, which have demonstrated significant capabilities in content generation and multi-turn dialogue, paving the way for more generalized AI systems.

Definitions

While the Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence do not contain a dedicated, explicit glossary section, the document implicitly defines and elaborates upon several core concepts through its thematic structure and operational objectives. Central to the measures is the concept of "General Artificial Intelligence" (通用人工智能), which refers to AI systems capable of performing a wide range of intellectual tasks, akin to human cognitive abilities, rather than being limited to specific functions. This encompasses advanced AI models such as "large language models" (大型语言模型), "large visual models" (大型视觉模型), "multimodal large models" (多模态大模型), "scientific computing large models" (科学计算大模型), and "brain-inspired neural networks" (脑启发神经网络), which are explicitly mentioned as targets for research and development. These terms collectively describe the sophisticated, high-parameter models that are at the forefront of AI innovation and are central to achieving general AI capabilities.

Another fundamental concept is "Computing Power Resources" (算力资源), which refers to the computational infrastructure and capabilities essential for training, deploying, and operating complex AI models. The measures highlight the need to enhance the integrated supply and efficient deployment of these resources, including the establishment of public computing centers and multi-cloud computing power scheduling platforms. Similarly, "High-Quality Data Elements" (高质量数据要素) are crucial, denoting the vast, diverse, and meticulously curated datasets required for training robust and unbiased AI models. This includes "Chinese pre-training data" (中文预训练数据) and "multimodal data sources" (多模态数据来源) such as text, images, audio, and video, which are vital for developing comprehensive large models. The document also implicitly defines an "Innovation Ecosystem" (创新生态) as a collaborative environment fostered by government guidance and innovation platforms, integrating resources like talent, technology, and capital to facilitate AI development. Finally, the concept of a "Prudent and Inclusive Regulatory Environment" (包容审慎的监管环境) is introduced, advocating for a balanced approach that encourages innovation while mitigating risks associated with AI development and deployment.

Governance and Institutional Framework

The governance and institutional framework for the Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence is primarily spearheaded by the Beijing Municipal People's Government Office, which formally issued the document. This central role signifies the city government's commitment to providing top-down guidance and strategic direction for AI development. The measures themselves are a notice from the Beijing Municipal People's Government Office, indicating that they carry the weight of official government policy and mandate implementation across relevant municipal departments and districts. The document explicitly states that it was issued "经市政府同意" (with the approval of the Municipal Government), underscoring its authoritative nature and broad applicability within Beijing's administrative structure.

Beyond the issuing authority, the implementation of these measures involves a collaborative effort among various municipal entities. While the specific coordinating bodies are not exhaustively listed in the provided snippets, the emphasis on "充分发挥政府引导作用和创新平台催化作用" (fully leveraging the government's guiding role and the catalytic effect of innovation platforms) suggests a multi-stakeholder approach. This implies that agencies responsible for science and technology, economic development, and information technology, such as the Beijing Municipal Science & Technology Commission, the Administrative Commission of Zhongguancun Science Park, and the Beijing Municipal Bureau of Economy and Information Technology, would play significant roles in operationalizing the measures. For instance, the measures advocate for organizing commercial computing power to meet urgent demands and efficiently promoting the construction of new computing power infrastructure, tasks that would require coordination among various government departments and private sector partners. The vision includes the establishment of "Beijing Artificial Intelligence Public Computing Centers" and "Beijing Digital Economy Computing Centers" in districts like Haidian and Chaoyang, indicating a decentralized yet coordinated implementation strategy involving district-level governments.

Key Focus Areas

The Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence delineate five principal areas of focus, each designed to bolster Beijing's comprehensive AI capabilities and foster a thriving innovation ecosystem. The first key area is the "Enhancement of Integrated Supply Capacity for Computing Power Resources" (提升算力资源统筹供给能力). This involves strategically leveraging Beijing's existing computing power advantages, implementing a 'computing power partner program' (算力伙伴计划) to collaborate with cloud vendors, and rapidly consolidating available computing resources. The objective is to establish clear technical standards, hardware and software service requirements, and scale of computing power supply, ensuring that innovative entities have access to diverse, high-quality, and inclusive computing power to support AI technology innovation and product development. Furthermore, the measures aim to efficiently promote the construction of new computing power infrastructure, including the establishment of large-scale advanced computing power supply capabilities through centers in Haidian and Chaoyang districts, supporting the development of models with billions of parameters.

The second critical area is the "Enhancement of High-Quality Data Element Supply Capacity" (提升高质量数据要素供给能力). This involves organizing relevant institutions to integrate and clean Chinese pre-training data, forming secure and compliant open basic training datasets. The measures also advocate for continuously expanding multimodal data sources and building high-quality corpora for large models, encompassing text, images, audio, and video, while supporting data circulation and transactions through legally established data trading institutions. The third focus area is the "Systematic Construction of General Artificial Intelligence Technology Systems, including Large Models" (系统构建大模型等通用人工智能技术体系). This entails guiding innovative entities to strengthen research into general AI technology systems, continuously exploring new pathways for general intelligent agents, brain-inspired intelligence, and embodied intelligence. The fourth area emphasizes "Promoting Innovative Application of General Artificial Intelligence Technology in Scenarios" (推动通用人工智能技术创新场景应用). This involves increasing the supply of innovative application scenarios and promoting the demonstrative application of general AI in various sectors such as government affairs, healthcare, scientific intelligence, finance, autonomous driving, and urban governance. Finally, the measures highlight "Exploring the Creation of an Inclusive and Prudent Regulatory Environment" (探索营造包容审慎的监管环境), which encourages innovative entities to utilize secure and trustworthy software, tools, computing, and data resources. It also supports independent innovation, promotion, application, and international cooperation in foundational AI technologies like algorithms and frameworks, advocating for pilot programs for inclusive and prudent regulation in areas like the Zhongguancun National Independent Innovation Demonstration Zone.

Implementation Framework

The implementation framework for the Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence is designed to be multi-faceted, leveraging both governmental guidance and market mechanisms to achieve its ambitious goals. A cornerstone of this framework is the establishment of the "computing power partner program" (算力伙伴计划), which aims to consolidate existing computing resources and collaborate with major cloud service providers. This program is critical for ensuring a stable and diversified supply of high-quality computing power, which is a fundamental requirement for the development and training of advanced AI models. Through this initiative, Beijing seeks to provide innovative entities with accessible and affordable computing resources, thereby lowering barriers to entry and accelerating research and development. Furthermore, the measures call for the efficient promotion of new computing infrastructure construction, including the development of large-scale public computing centers in key districts, which will serve as foundational hubs for AI innovation.

In parallel with computing power, the measures outline a robust strategy for enhancing the supply of high-quality data elements. This involves organizing institutions to curate and clean vast datasets, specifically focusing on Chinese pre-training data and multimodal data sources. The establishment of "data training bases" (数据训练基地) is envisioned to facilitate high-level data openness and improve the scale and quality of AI data annotation libraries. These bases will serve as critical infrastructure for data processing, enabling the creation of secure and compliant open basic training datasets essential for model development. The framework also emphasizes fostering an innovation ecosystem by integrating various resources and strengthening factor allocation, which includes talent, technology, and capital. This involves encouraging collaboration between research institutions, universities, and enterprises to collectively advance AI technology. The measures also promote the creation of a unified multi-cloud computing power scheduling platform to manage and operate heterogeneous computing environments, ensuring seamless and efficient execution of AI computing tasks across different cloud environments. This integrated approach to resource management and ecosystem building is central to the operationalization of the measures, aiming to create a synergistic environment where innovation can flourish.

Monitoring and Evaluation

The Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence, while primarily focused on promotional strategies, implicitly establish a framework for monitoring and evaluation to ensure the effectiveness and progress of the outlined initiatives. Although the provided snippets do not detail explicit metrics or a dedicated oversight body for evaluation, the very nature of "promotion measures" implies an ongoing assessment of their impact on the AI industry. The stated goals of enhancing computing power supply, improving data element quality, building general AI technology systems, and promoting application scenarios inherently require continuous monitoring to gauge success. For instance, the success of the "computing power partner program" would likely be evaluated based on the increase in available computing power, the number of innovative entities utilizing these resources, and the efficiency of the multi-cloud scheduling platform. The development of public computing centers would be monitored for their capacity, utilization rates, and their contribution to supporting large model research and development.

Similarly, the effectiveness of measures related to data elements would be assessed by the volume and quality of curated datasets, the establishment and operationalization of data training bases, and the extent of data circulation and transaction through authorized platforms. Progress in building general AI technology systems, particularly large models, would be evaluated by the number and sophistication of models developed, breakthroughs in foundational AI technologies, and the exploration of new pathways in areas like embodied intelligence. The promotion of innovative application scenarios would be monitored by the diversity and impact of AI applications across various sectors, including government affairs, healthcare, and finance. Furthermore, the commitment to "exploring the creation of an inclusive and prudent regulatory environment" suggests an ongoing review of regulatory approaches to ensure they foster innovation while effectively managing risks. This continuous feedback loop, though not explicitly detailed in terms of reporting mechanisms, is essential for adapting the measures to the dynamic landscape of AI development and ensuring that Beijing remains at the forefront of AI innovation. The broader context of the "Beijing Implementation Plan for Accelerating Construction of Sources of Artificial Intelligence Innovation with Global Influence (2023-2025)" further suggests that these measures are part of a larger, coordinated effort with defined objectives that would necessitate regular performance reviews.

Penalties, Liability, and Appeals

The Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence are primarily designed as a promotional and facilitative policy instrument, focusing on incentivizing and supporting the growth of the AI industry rather than imposing strict penalties or establishing detailed liability frameworks. Consequently, the document does not explicitly outline specific penalties for non-compliance, nor does it detail mechanisms for liability or appeals in the traditional legal sense. The emphasis is on fostering an environment conducive to innovation, with a strong focus on collaboration and resource provision. Instead of punitive measures, the policy encourages adherence through the promise of support, resources, and a favorable operating environment for entities engaged in AI development. The underlying philosophy appears to be one of guidance and enablement, aiming to steer the industry towards responsible and innovative practices through positive reinforcement and strategic support rather than through sanctions.

However, the measures do touch upon the concept of a "prudent and inclusive regulatory environment" (包容审慎的监管环境), which implies a recognition of the inherent risks associated with advanced AI technologies. This regulatory approach seeks to balance the need for innovation with the imperative of risk prevention. It encourages innovative entities to utilize "secure and trustworthy software, tools, computing, and data resources", suggesting that while direct penalties are not specified, there is an expectation of responsible conduct and adherence to evolving safety and ethical guidelines. The mention of exploring "pilot programs for inclusive and prudent regulation" in areas like the Zhongguancun National Independent Innovation Demonstration Zone indicates a forward-looking approach to governance, where regulatory tools and processes are innovated to address emerging challenges without stifling technological advancement. While the measures themselves do not detail liability provisions, it is understood that broader national and municipal laws and regulations concerning data security, cybersecurity, intellectual property, and product liability would still apply to AI-related activities in Beijing. Therefore, while this specific document focuses on promotion, the legal and ethical responsibilities of AI developers and deployers remain governed by the wider legal landscape.

Relationship to Other Instruments

The Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence are not a standalone policy but are strategically positioned within a broader framework of national and municipal initiatives aimed at advancing artificial intelligence. The document explicitly states that its formulation is "为贯彻落实《北京市加快建设具有全球影响力的人工智能创新策源地实施方案（2023-2025年）》" (to implement the 'Beijing Implementation Plan for Accelerating Construction of Sources of Artificial Intelligence Innovation with Global Influence (2023-2025)'). This direct linkage establishes the Measures as a key operational component of a larger, more comprehensive strategic plan for AI development in Beijing. The 'Implementation Plan' serves as the high-level strategic roadmap, setting ambitious goals for Beijing to become a globally influential AI innovation hub, while these 'Measures' provide the specific, actionable steps and policy levers to achieve those overarching objectives, particularly focusing on general AI.

Furthermore, the Measures align with and are informed by broader national strategies, including directives from the 20th National Congress of the Communist Party of China and the Politburo meetings that emphasize the importance of developing strategic emerging industries, fostering innovation ecosystems, and addressing risks in AI. This indicates that Beijing's AI policy is integrated into China's national AI agenda, contributing to the country's overall technological advancement. The Measures also implicitly interact with other regulations concerning data governance, cybersecurity, and intellectual property, even if not explicitly cited. For example, the focus on "secure and compliant open basic training datasets" and "data circulation, transaction" implies adherence to existing or forthcoming data protection and data security laws. Similarly, the promotion of AI technology and applications would operate within the existing legal frameworks for intellectual property rights. The emphasis on an "inclusive and prudent regulatory environment" suggests an awareness of the need to harmonize these promotional measures with broader regulatory oversight, ensuring that innovation proceeds responsibly and in line with national policy directions.

International Alignment

The Beijing Measures for Promoting the Innovative Development of General Artificial Intelligence implicitly and explicitly aim for international alignment, primarily by positioning Beijing as a "globally influential" hub for AI innovation. The foundational document that these measures implement, the "Beijing Implementation Plan for Accelerating Construction of Sources of Artificial Intelligence Innovation with Global Influence (2023-2025)," clearly articulates an ambition for global leadership. This overarching goal necessitates an awareness of international best practices, technological advancements, and ethical considerations in the global AI landscape. While the measures do not extensively detail specific international collaborations or mutual recognition agreements, the pursuit of "innovation leadership" and the development of "cutting-edge AI models that achieve domestic or international leadership status" inherently require benchmarking against global standards and fostering an environment that can attract international talent and investment.

The document's emphasis on "international cooperation" in foundational AI technologies like algorithms and frameworks further underscores a desire for engagement with the global AI community. By encouraging such cooperation, Beijing seeks to both contribute to and benefit from the worldwide advancements in AI research and development. The focus on developing general AI, large models, and related infrastructure is in line with global trends in AI research and industrial development, indicating a strategic alignment with the cutting edge of the field. Furthermore, the exploration of an "inclusive and prudent regulatory environment" suggests an effort to develop governance approaches that are both effective locally and potentially adaptable or comparable to international regulatory discussions on AI ethics, safety, and responsible development. While the primary focus remains on domestic innovation and industrial promotion, the aspiration for global influence and leadership intrinsically links these measures to the broader international discourse and competitive landscape of artificial intelligence.

Implementation Timeline

Compliance Checklist

Sources and References