Connexion
Executive Summary:
On June 3, 2026, the National Data Administration (NDA) issued its Implementation Plan for Promoting the Construction of High-Quality Industry Datasets (关于推进行业高质量数据集建设行动的实施方案) (NDA, June 8). The plan is focused on building the PRC’s data foundations on which artificial intelligence (AI) systems depend. It directs the construction of sector-specific datasets for AI model training across 19 sectors and five innovation areas, including healthcare, e-commerce, public security, urban governance, and social credit, as well as innovation areas such as the low-altitude economy. [1] As with all aspects of PRC data governance, commercial and administrative incentives coexist with the implementation of the Party’s social control and national security objectives.
High-Quality Datasets
The term “high-quality dataset” (高质量数据集) has a specific official definition that is both technical and political in nature. According to technical standards issued in August 2025 by the “National Technical Committee 609 on Data of Standardization Administration of China” (SAC/TC609; 全国数据标准化技术委员会), a high-quality dataset is defined as a collection of data that has undergone collection, processing, and other data handling, can be directly used for developing and training AI models, and can effectively improve model performance. [2] Dataset quality is reflected across five dimensions: “large” scale, “robust” security, “correct” viewpoint, “good” results, and “wide” application, and can be measured using both static and dynamic quality evaluation methods, which are outlined below (NDA, August 2025). Data security compliance is also defined within a series of quality compliance indicators, which state that data in datasets for AI model development and training shall “not contain illegal content including content that violates socialist core values, discriminatory content, commercial violations, or content that infringes upon the legitimate rights and interests of others” (不包含违反社会主义核心价值观的内容、歧视性内容、商业违法违规、侵犯他人合法权益等非法内容) (SAC/TC609, August 2025).
The same standards establish a three-tier classification system for high-quality data, which are arranged by sensitivity. “General knowledge datasets” (通识数据集) contain knowledge accessible to the public without specialist background, sourced from encyclopedias and Internet platforms, and are rated low sensitivity. “Industry general knowledge datasets” (行业通识数据集) contain knowledge requiring professional background, sourced from academic papers, industry reports, and official documents, and are also rated low sensitivity. “Industry professional knowledge datasets” (行业专识数据集) contain knowledge requiring deep professional background and operational experience, sourced from organizations’ internal business systems and management platforms, and are rated as highly sensitive (SAC/TC609, August 29, 2025).
The NDA’s High-Quality Dataset Construction Guidelines (高质量数据集建设指引), a methodology document on building high-quality datasets for AI training issued by the NDA in August 2025, recorded that “more than 35,000 high-quality datasets had been built nationally as of June 2025, with a total volume exceeding 400 [petabytes]” (截至2025年6月，全国建设高质量数据集超3.5万个、总量超400PB) (NDA, August 2025). The guidelines identify three main technical approaches to data collection:
Each reflects a different aspect of the PRC’s existing data infrastructure, as well as their relationship to challenges of building AI training data at scale. Social credit, for example, illustrates the first. It is inherently multi-source and heterogeneous because it is distributed across separate authorities including public security, courts, tax, and market regulators. The National Credit Information Sharing Platform (全国信用信息共享平台) and unified social credit code system have been designed to improve aggregation and information sharing across silos (People’s Daily, April 1, 2025). Progress has been uneven, but the direction and intent are clear and iterative improvements will continue to see those objectives realized.
Physically Distributed but Logically Centralized
The new implementation plan includes dataset management provisions that require full lifecycle oversight covering collection, cleaning, processing, labeling, quality inspection, evaluation, iteration, and audit. The provisions subject data to state management within a system described as “physically distributed but logically centralized” (物理分散、逻辑集中) (NDA, June 8).
This approach is not new: the formulation appeared as the governing principle for cross-departmental data sharing under the 2015 “Big Data Development Action Outline” (促进大数据发展行动纲要) from the National Development and Reform Commission (NDRC; 国家发展和改革委员会), and at the time the national e-government extranet connecting 118 central government units and the social credit system’s information sharing infrastructure were among the systems built on that basis (NDRC, September 25, 2015).
The National Dataset Management Service Platform (国家数据集管理服务平台), built and operated by the National Data Development Research Institute (国家数据发展研究院) under NDA guidance, launched for a trial period on April 29, 2026, just before the plan was formally issued, at the Digital China Construction Summit (数字中国建设峰会) (NDA, April 29). The platform provides dataset catalogue management and construction monitoring functions to data management authorities, publication and quality evaluation functions to dataset suppliers, and search and retrieval functions to dataset users.
One of the Party-state’s instruments for addressing structural weaknesses associated with achieving the objective of “physically distributed but logically centralized” (物理分散、逻辑集中) has been the rollout of “data exchanges” (数据交易所). Data exchanges are state-supervised marketplaces designed to make bulk data transactable, fusible, and legible across the departmental and jurisdictional lines that fragmented governance had drawn (National Information Center, February 20, 2024). The first exchange had been established in Guiyang in 2014, but they began proliferating after the 13th Five-Year Plan, which called for implementing a “national big data strategy” (国家大数据战略) and promoting the open sharing of data resources (Study China, November 12, 2015; NPC, March 2016; CAC, May 18, 2021). Relatedly, the new implementation plan’s provisions encourage datasets to be listed and traded through data exchanges and data circulation service institutions, with the objective of existing trading models upgrading from basic data packages to application programming interface (API) calls and full-stack services. [3]
Data exchanges are designed so that transactions are logged, participants verified, and audit trails maintained (State Administration for Market Regulation, March 1, 2020). Data exchanges, as data holders, are also subject to the cooperation obligations of the 2021 Data Security Law, which requires provision of data to public security and state security organs on demand (NPC, June 10, 2021).
Conclusion
The implications of the provisions of the Implementation Plan for Promoting the Construction of High-Quality Industry Datasets are that, in the long term, they are part of a series of regulations, standards, and other measures designed to improve the Party’s ability to both maximize data and technology ecosystems for the purposes of economic development and its own social and political control. The control elements are embedded subtly in the definition of quality data and in the terms governing how that data is ultimately catalogued and shared, traced, and potentially accessed.
The plan is one regulation in a decades-long systems construction project. Where its provisions fit in the bigger puzzle is that they enable the kinds of data sharing and dataset building that will improve not only the quality and accuracy but also the effectiveness of a range of governance systems. This includes the social credit, urban governance platforms (such as smart cities platforms), and public security applications, that the plan refers to specifically.
Notes
[1] According to the NDA Implementation Plan, the full sector list includes: scientific research, industrial manufacturing, agriculture and rural areas, smart energy, transport, financial services, healthcare, education, e-commerce, human resources, culture and tourism, emergency management, meteorological services, green and low-carbon, public security, urban governance, housing construction, natural resources, and social credit. Innovation areas: low-altitude economy, embodied intelligence, autonomous driving, smart ocean, and bio-manufacturing (NDA, June 8).
[2] Key technical standards include:
1. TC609-5-2025-0, Construction Guidelines (高质量数据集 建设指南), which cover full lifecycle construction guidance;
2. TC609-5-2025-02, Format Requirements (高质量数据集 格式要求) which cover metadata format standardization;
3. TC609-5-2025-03, Classification Guidelines (高质量数据集 分类指南) which cover unified classification framework; and
4. TC609-5-2025-04, Specifications for quality evaluation and test (高质量数据集 质量评测规范), which cover unified evaluation indicators covering accuracy, completeness, timeliness, and consistency.
For a summary, see: https://archive.ph/MbWbi.
The full text of TC609-5-2025-03 can be found at: https://web.archive.org/web/20260622121303/https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/3.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E5%88%86%E7%B1%BB%E6%8C%87%E5%8D%97%E3%80%8B.pdf.pdf.
The full text of TC609-5-2025-04 can be found at: https://web.archive.org/web/20260624011335/https://sjj.hubei.gov.cn/bmdt/tzgg/202511/P020251110587761615062.pdf.
[3] A basic data model refers to a buyer acquiring a static snapshot copy of a dataset, likely using a database manager or excel, for further applications. An API allows one application to “call” another application to request data, automating the data request process. API-based or full-stack-services provide the buyer with the ability to access real-time data on demand (Cloudflare, accessed June 24).
The post Databases Centralized for AI Development appeared first on Jamestown.
Executive Summary:
At the end of March, Chinese state media announced the establishment of a new global governance institution, the World Data Organization (WDO; 世界数据组织), headquartered in Beijing. The top of the People’s Daily front page reported that Chinese Communist Party (CCP) General Secretary Xi Jinping had penned a congratulatory letter that Politburo Standing Committee member Ding Xuexiang (丁薛祥) had read out at the organization’s founding conference. Ding called it “a major event” (一件大事) and declared that Xi’s letter, which said the WDO would “provide a valuable platform for … improving global data governance” (为 … 完善全球数据治理提供了有益平台), demonstrated the country’s “high expectations” (殷切期望) for the WDO (People’s Daily, March 31, [a], [b]).
Publicly available information about the WDO currently tells a different story. Although official coverage claimed that around 500 people attended the founding conference, and that the WDO has more than 200 members from over 40 different countries representing 14 major industries, no publicly available evidence supports these claims (CCTV, March 30). [1] Xinhua merely reports that these members are “entities and individuals related to the global data sector” (由全球数据领域相关单位及个人自愿结成的) (Xinhua, March 30). Reporting also notes a founding declaration was released at the event, though this, too, is not yet public (Xinhua, March 30). As of publication, the only acknowledged member of the WDO appears to be Jack Perry, Chairman of the 48 Group Club. The 48 Group Club is a London-based company formed in 1954 by members of the Communist Party of Great Britain. Its “mission statement” includes a nod to Zhou Enlai (周恩来), who the group’s website refers to as “China’s much-loved Premier” (48 Group Club, accessed June 12). [2]
WDO Promotes Party Preferences in Cyberspace
The WDO almost immediately cultivated a sense of international imprimatur that looks to be premature. Although it appears to present itself as an international institution, its leadership and characterizations of its intended role in state media show that it is an unusually obvious vehicle for enhancing CCP—rather than purely PRC—influence. In light of this, the decision of the United Nations (UN) Commission on Science and Technology for Development and vice chair of the UN Data Governance Working Group Muhammadou M.O. Kah to deliver remarks at the WDO’s inaugural event likely should be interrogated (Xinhua, March 31).
Attempts to access the WDO’s website (thewdo.org) currently redirects to a “Coming Soon” placeholder (thewdo.org/ComingSoon), but still allows the home page to be briefly visible. A capture of the home page shows that it contains the opening lines of the organization’s “founding declaration” (成立宣言), which includes calls to “strengthen industrial collaboration, jointly build data infrastructure, advance the development of standards and joint innovation, and promote the secure and trustworthy flow of data” (加强产业协同，共建数据基础设施，推进标准研制与联合创新，推动数据安全可信流动与资源开发利用). It also presents the organization’s priorities. These are extensive, but include working on development plans; handling issues among members and mediating disputes; building unified repositories for relevant regulations, technologies, experts, enterprises, tools, and resources; organizing and conducting policy research and data evaluation and testing; setting international standards and providing certification services; driving the construction of critical data service infrastructure and open-source ecosystems, and cultivating data application scenarios, among other far-reaching tasks (WDO, accessed June 12). [3]
The limited information that can be extracted from the WDO website, which is yet to launch, is a sanitized version of what appears in Party media. A recent article by Cheng Haiye (程海烨) of the Institute of International Studies, Shanghai Academy of Social Sciences sees the WDO as primarily supporting the CCP’s core security interests in the international arena. Despite later calling for promoting policy dialogues with the United States and Europe, he says, first of all, that the organization will “uphold the bottom line of security” (坚守安全底线) and support other countries’ data sovereignty. He then lists a number of items that the organization is there to oppose that indicate, implicitly but unequivocally, that the organization is targeted as an anti-Western institution: “data hegemony, bloc confrontation, and ‘long-arm jurisdiction’” (数据霸权、阵营对抗与“长臂管辖”). The article also makes clear that Cheng views the WDO as in line with Xi Jinping’s comprehensive national security concept, his proposed Global Data Security Initiative, and efforts to build a community of common destiny in cyberspace (Shanghai Academy of Social Sciences, June 3).
More telling is a commentary that appeared on the theory page of the People’s Daily on June 8. This piece was written by the WDO’s president (理事长), Tan Tieniu (谭铁牛). The piece is steeped in Party rhetoric, quoting Xi Jinping no fewer than seven times and intoning the appropriate Party formulas, referencing Xi’s four global initiatives and the community of common destiny. It also includes the same targeted rejection of the West, decrying “small yards and high fences” (小院高墙), data hegemony (数据霸权), and “data colonialism” (数据殖民主义). In this, he is more explicit than Cheng, despite his executive position in the WDO. He criticizes the United States for its “attempts to maintain digital hegemony under the guise of ‘digital freedom’” (美国以“数字自由”为名试图维持数字霸权地位) and the European Union for its influence on global standards, which he argues exacerbate “the fragmentation and imbalance in the global data governance landscape” (全球数据治理格局的分化与失衡进一步加剧) (People’s Daily, June 8).
Tan’s ambitions for the WDO are also clearly in line with CCP preferences. He writes that the organization will “help to further implement the [four] global governance initiatives” (有助于深入践行全球治理倡议) and “promote the development of the international order in a more just and equitable direction” (推动国际秩序朝着更加公正合理的方向发展), a common refrain that entails making the world safer for autocratic systems like the PRC’s (People’s Daily, June 8).
WDO Led by Senior United Front and Surveillance Tech Champions
Tan Tieniu is both an obvious and a disturbing choice to lead the WDO. An accomplished and internationally recognized scientist, Tan appears eminently capable of running such an organization. His longstanding involvement with the CCP’s united front system, and his extensive and groundbreaking research on surveillance technologies, however, are cause for alarm.
The most detailed account of his background appeared in a long profile published in CCP Members News in 2015. It charts his life’s story, from receiving a government scholarship to study at Imperial College London and taking up a tenured position at Reading University in the United Kingdom, to his rapid rise within the Chinese Academy of Sciences (CAS). He won multiple prestigious awards and accolades along the way, including being awarded the title of “National Outstanding Communist Party Member” (全国优秀共产党员) by the Central Committee on the Party’s 80th anniversary—one of only 50 recognized nationwide that time. The piece, although no doubt embellished for its propaganda value, conveys a strong sense of Tan’s support for the Party, which he joined in 1983, before he moved to the United Kingdom. Tan’s involvement with the united front system is extensive. His prior positions include
He was also a delegate to the 16th, 17th, 18th, and 19th Party Congresses. His stature has only risen since then (People’s Daily Online/CCP Member News, July 27, 2015). Before taking on his new role, he was the Party Secretary of Nanjing University (Nanjing University, accessed June 12). While this is all indicative of his stellar CCP credentials, his other activities are much more concerning.
In 2021, the Biden administration in the United States sanctioned Tan Tieniu. At the time, he was one of seven deputy directors of the Liaison Office of the Central People’s Government of the Hong Kong Special Administrative Region, and so involved in the undermining of the Sino-British Joint Declaration (U.S. State Department, July 16, 2021).
Tan Tieniu’s research history suggests that he may have directly contributed to the state’s technological capabilities that have enabled an unprecedented apparatus of repressive social control. His published research has focused heavily on the ability of digital imaging to identify people via recognizing their gait or analyzing their irises and palmprints. He has received awards and grants for research that explicitly contributes to domestic security work. These include a grant for the years 2012–2016 to study “social perception data processing for public security” (面向公共安全的社会感知数据处理), third prize in the 2008 Beijing Municipal Science and Technology Progress Award for his work on “real-time intelligent video surveillance early warning system and application” (实时智能视频监控预警系统及应用), and second prize in the 2011 National Science and Technology Progress Award for his research into “video content understanding technology and applications for security surveillance” (面向安全监控的视频内容理解技术与应用) (Baidu/谭铁牛; Google Scholar/Tieniu Tan, accessed June 12).
Conclusion
The WDO, according to its vice president and secretary general (秘书长), former Party Secretary and Chairman of the China Mobile Communications Group Yang Jie (杨杰), wants to work closely with the International Telecommunication Union and the World Intellectual Property Organization to “jointly promote global data cooperation from rules discussions to standards alignment, application scenario implementation, and inclusive development” (共同推动全球数据合作从规则讨论走向标准衔接、场景落地和普惠发展) (Xinhua, March 30). If it succeeds in these ambitions, it will be a significant win for the Chinese Communist Party at the expense of the United States and its allies.
The WDO is yet to gain significant ground as a global institution, though it has already received a level of endorsement from the United Nations. The selection of a world-leading expert in human surveillance technologies, who is sanctioned by the U.S. government, and who views the organization as a vehicle to advance the CCP’s interests and explicitly to push back on the United States, however, suggests that the West has an interest in making sure it gains no further ground.
Notes
[1] The 14 industries are: industry, finance, healthcare, public services, e-commerce retail, transportation and logistics, energy, construction, internet, agriculture, education, media, automotive, and law. A webpage on Aiqicha (爱企查) claims that as of mid-May 2026, the organization had received “hundreds of membership applications worldwide, covering a total of 42 countries and regions, with corporate members accounting for over 76 percent” (已收到全球上百份会员申请，会员总数覆盖 42 个国家和地区，企业会员占比超过 76%) (Aiqicha, accessed June 12).
[2] Zhou Enlai was also the founder of CCP intelligence and deeply involved in overseas united front work. See: Mattis, Peter and Matthew Brazil, Chinese Communist Espionage: An Intelligence Primer (Naval Institute Press, 2019). Among his more horrific acts, in 1931 he directly ordered the massacre of the family of a former agent, which included a two-year-old child. See: Chen Jian, Zhou Enlai: A Life (Harvard University Press, 2024).
[3] There is an apparent separate Chinese website for the WDO (sjsjzz.com). This website, however, contains only a brief home page and looks suspiciously like it has been made by an AI model (WDO, accessed June 12).
The post U.S.-Sanctioned United Front Figure Leads World Data Organization appeared first on Jamestown.
Executive Summary:
Policymakers in the People’s Republic of China (PRC) want to improve “data supply” (数据供给) to “provide sufficient ‘fuel’ for the development and application of artificial intelligence” (为人工智能发展和应用提供充足“燃料”). This is according to a new plan from the National Data Administration (NDA, 国家数据局) titled “Implementation Plan for Promoting the Construction of High-Quality Industry Datasets” (关于推进行业高质量数据集建设行动的实施方案). Published on June 8, the plan aims to create a “data flywheel” to accelerate the construction of a “symbiotic ecosystem in which data factors and artificial intelligence evolve synergistically” (建数据要素与人工智能协同演进的共生生态) (NDA, June 8).
One speculative part of the plan calls for “exploring a value system based on tokens” (探索以词元 (Token) 为基础的价值体系). This is part of a commercial push to make data-related products economically viable, which policymakers believe will dramatically improve the development of the AI sector in the People’s Republic of China (PRC). The plan therefore directs industry to “explore new trading models such as token trading, and build a data value system based on tokens that is quantifiable and priceable” (探索词元交易等新型交易模式，构建以词元为基础，可量化、可定价的数据价值体系) (NDA, June 8).
The wording around tokens (词元) in the plan is tentative but significant. This is because it is the first time that the digital token has been given an official, standardized Chinese name and definition. It was only in March that NDA director Liu Liehong (刘烈宏) announced that the Mandarin word ciyuan (词元) would be used officially to denote digital tokens. In an address delivered at the China Development Forum, Liu described tokens as “settlement units” (结算单位) connecting technological supply with commercial demand that are “measurable, priceable, and tradable” (可计量、可定价、可交易) (People’s Daily, March 25). The importance of tokens—the basic unit through which large language models (LLMs) process information—to artificial intelligence (AI) applications has prompted Beijing to define, standardize, and regulate them before any real market emerges.
Token Centrality Leads Beijing to Take the Initiative
Token consumption has become a key indicator of demand for AI infrastructure in the PRC. The rise of agentic software that carries out multi-step tasks by planning, calling tools (drawing on external programs such as web search, code, or databases), and checking and revising its own output, has pushed token consumption far above that of a simple chatbot exchange (Anthropic, November 26, 2025, March 24). China CITIC Bank (CITIC; 中信银行) estimates that a single task for an AI agent consumes 10–100 times the number of tokens of a standard chatbot session (WeChat/@CITIC Securities, April 15). The PRC’s recent nationwide “lobster farming” (养虾) craze, in which users deploy and continuously run the open-source agent OpenClaw, was a catalyst for growing demand for inference compute and cloud services. [1] The ensuing surge in token usage has led to a “compute shortage” (算力荒), rising cloud prices, and demand for domestic inference chips, according to CITIC (WeChat/@Huatai Securities, March 13; WeChat/@CITIC Securities, April 15).
Beijing has moved quickly to define the regulatory environment around the token economy after its rapid emergence. After Liu defined tokens and the token economy in March, the NDA has moved to produce an implementation plan for the sector in roughly two months—including a month-long public comment period, during which state media raised public awareness about the “ciyuan economy” (WeChat/@China Economic Times, April 17, April 21, April 27, April 28). This official coverage makes clear that Beijing views tokens as more than an accounting unit within the LLM technology stack. Tokens are an instrument that can link datasets, model services, and compute resources across an entire emerging sector.
Setting Standards and Picking Winners
Without waiting for dedicated legislation, Beijing is applying existing tools to govern the token economy. On the positive side, the state backs compliant intermediaries as the legitimate operators of token services. For example, the NDA has spotlighted Guizhou Data Bao (贵州数据宝), a former enterprise compute provider that has pivoted to selling token services from Gui’an’s (贵安新区) state-backed data center cluster, where its daily token calls roughly quadrupled in three months (NDA, April 10). On the negative side, it constrains higher-risk intermediaries through existing rules on AI services, data exports, anti-money laundering, and foreign exchange. For instance, in June 2026 the Ministry of State Security (MSS) publicly warned that unregulated “AI relay stations” (AI中转站), single-interface resellers of access to various domestic and foreign models, pose data security- and national security-related risks (Wechat/@MSS, June 8). Legal commentators had already flagged one such platform, B.AI, which was launched by controversial crypto entrepreneur Justin Sun (孙宇晨), as potentially violating the Interim Measures for the Management of Generative AI Services (生成式人工智能服务管理暂行办法) and the PRC’s cross-border data-export rules (Cyberspace Administration of China, July 10, 2023; National Business Daily, May 12). [2]
Price oversight may be the next front, though it remains at the proposal stage. After Alibaba Cloud, Tencent Cloud, and Baidu AI Cloud all raised prices for compute within a ten-day window, researchers at the National Development and Reform Commission’s Institute of Market and Price Studies argued for a monthly token price index and a billing disclosure regime (WeChat/@China Economic Times, May 18). The government is yet to formally propose such a mechanism, however, and state media still describe the token economy’s pricing and regulatory framework as under construction (Xinhua, June 2).
Beijing’s moves so far have been about who may operate in the token economy. Whether the token can also be traded and priced is a separate question. The NDA implementation plan’s call to “explore ciyuan trading” is an early move toward it, expanding the current data trading regulatory regime to tokens and building on the 2022 “Twenty Articles on Data” (数据二十条), released jointly by the Party Central Committee and the State Council, which allowed for data trading even when ownership is unclear (MOFCOM, December 2, 2022).
One of the main problems for developing a token trading system is figuring out how to price tokens (WeChat/@Journal of Modern Commercial Banking, March 26). An intuitive approach would be to set prices based on token volume. Some services, such as OpenRouter, trade freely using per-token pricing for AI usage. This approach is unlikely to lead to a large, liquid market, however, as token value is shaped by an array of additional factors beyond volume, meaning that services that consume more tokens are not necessarily worth more (WeChat/@China Economic Times, April 21; CSDN, April 27). The NDA’s framing of “ciyuan trading” which anchors the value and exchange of data on the token, is therefore unlikely to become a liquid market soon.
The NDA’s approach is more likely driven by a desire for control. By making the token a common valuation tool, it places datasets, model services, and compute within a single comparable system. A state-led system for measuring and evaluating AI services based on the token is already being built. The China Academy of Information and Communications Technology (CAICT), a research institute under the Ministry of Industry and Information Technology (MIIT) that serves as the sector’s main standards body, recently upgraded its evaluation system for model-as-a-service (MaaS) platforms into a dedicated “Token Service Evaluation System” (Token服务评估体系). This broadens assessment from a platform’s model capabilities to the quality, performance, reliability, and metering (tracking how many tokens are used) of its token services (WeChat/@CAICT, April 14). In defining what counts as a competent token service, CAICT is setting the standards for the ciyuan economy. These standards in future could become de facto benchmarks for government procurement, state-owned enterprise (SOE) cloud adoption, and industry certification. Setting those standards before any market exists defines the requirements that every token-service operator will eventually have to meet.
State Guidance for Token Services
Chinese policymakers consistently emphasize the need to create use cases and applications for emerging technologies. The same goes for the ciyuan economy. As an article in the Party journal Study Times recently put it, AI competition is evolving from a race over model parameters to an engineering contest over “who can put models to use” (谁能把模型用好) (Study Times, May 13). State-owned China Telecom has moved fastest to occupy the model application layer, in early May launching Xingchen (星辰) TokenHub, a one-stop platform for aggregating models, routing requests, and metering and operating token services. This hub is tied to its Xirang (息壤) compute scheduling system and provides everything from raw compute to token delivery (WeChat/@State-Owned Assets Report, May 8). The company’s chairman, Ke Ruiwen (柯瑞文), has cast this as a “strategic choice” (战略选择) to move from “selling traffic” (卖流量) to “selling tokens” (卖token) (WeChat/@LinkitmeCloud, April 30). [3] In mid-May, all three state telecom operators launched retail token plans, with China Telecom’s starting at renminbi (RMB) 9.9 ($1.45) per month for 10 million tokens (National Business Daily, May 17).
China Telecom has convened an industry body, the Token Ecosystem Alliance (Token生态联盟), to pull more than 20 compute, model, security, and application partners onto its own platform and, in its words, “expand the token ecosystem” (做大词元生态) beyond its in-house services (NetEase, April 30). Company executives describe the alliance as a way to anchor that ecosystem around itself (C114, April 29). This illustrates a degree of coordinated competition under state direction. [4] The alliance gathers central SOEs, cloud platforms, model vendors, and chipmakers whose interests are far from aligned under the banner of “national cloud, national chips, national models” (国云、国芯、国模). [5] China Telecom wants to be the hub; Huawei wants to expand its Ascend chip ecosystem; Alibaba Cloud already runs its own stack; model vendors such as Zhipu AI (智谱) and iFlytek (讯飞) want distribution; and domestic chipmakers want access to the mainstream service layer (WeChat/@CITIC Securities, April 15; WeChat/@FaxAI, May 12). In practice, the alliance pools partners’ models, compute, and applications onto China Telecom’s platform. For a model vendor or chipmaker, joining the alliance is the way to secure both customers and a foothold in the emerging stack before its standards, routing rules, and entry points are settled. For China Telecom, which can now control access to the enterprise and government orders for which those vendors compete, the alliance cements the centrality of a powerful SOE (C114, April 29). The token-service layer, in other words, is where measurement standards, platform power, and industrial alliances converge, and where SOEs, by convening and coordinating, can influence the platforms that reshape the industry.
Numerous PRC firms are seeking to carve out space in the AI sector’s platform layer. The nascent state of this layer, with standards and entry points yet to be defined, means that its key players are yet to establish themselves. Infinigence AI (无问芯穹) is a good example of an emerging platform firm. It is a service provider, offering compute for AI inference that uses both Western and Chinese hardware, and currently hosts more than 160 models (WeChat/@ChinaVenture, May 6). Unlike Western chips such as those designed by Nvidia and AMD, many domestic chips do not come with mature software ecosystems, creating the space for intermediaries like Infinigence to translate the fragmented field of PRC chips into usable inference services (Rand, June 26, 2025). Infinigence has backing from numerous leading investors in the PRC technology space and recently raised more than $100 million in a round led by local state capital and industrial investors. The company has rebranded itself from a “compute operator” (算力运营商) to “China’s token economy hub” (中国 Token 经济枢纽), signaling its ambitions (WeChat/@ChinaVenture, May 6). This is an early signal that token infrastructure is becoming a distinct asset class that local state capital is moving to own—another channel through which the state shapes the emerging ciyuan economy.
Conclusion
A new economic sector, the ciyuan economy, is emerging rapidly in the PRC. Shaped by NDA policy and CAICT’s evaluation criteria, platform companies are moving into the space, often with the backing of significant investments. A mature ciyuan economy is a long way off, however, as important questions remain on how tokens will be priced, and what compliance framework will be constructed.
The state’s swift mobilization to regulate this new sector—before a market has had time to fully develop—is the inverse of the process in the United States. There, the token economy has run ahead of the state and regulation has stayed deliberately light. Each approach has its strengths and weaknesses. In light of these differences, the ultimate effects of the PRC’s concerted efforts to shape the country’s AI inference activity toward its preferred goals will be an important story in U.S.–PRC AI competition in the years ahead.
Notes
[1] “Lobster farming” is a Chinese internet meme referring to the deployment and use of the open-source AI agent OpenClaw. OpenClaw became known as “little lobster” (小龙虾) because its icon resembles a red lobster; “Lobster farming” refers to users configuring, fine-tuning, and continuously running the agent on a personal computer or cloud server so that it can perform workflows through messaging apps or other interfaces. After OpenClaw became a viral phenomenon in the PRC, local governments and tech companies quickly moved to support the trend. Shenzhen’s Longgang District (深圳龙岗区) issued the “Ten Lobster Measures” (龙虾十条), while Wuxi High-Tech Zone (无锡高新区) introduced twelve measures to support OpenClaw and related open-source community projects. These policies included computing-power subsidies, deployment services, application vouchers, and providing several million RMB in funding for so-called “one-person companies” (OPC; 一人公司) (CCTV, March 9). Major PRC tech firms, including Tencent, Alibaba, Baidu, ByteDance, Huawei, and Xiaomi, also lowered the barrier to “lobster farming” through cloud deployment, one-click installation of OpenClaw-like products, and domestic software and hardware adaptation (36Kr, March 11; Alibaba Cloud; Baidu Qianfan Cloud; Tencent Cloud; Volcengine; accessed May 14).
[2] Private AI API relay platforms like B.AI operate in a legal gray area. They forward domestic users’ prompts and business data to overseas models while operating outside the PRC’s formal compliance channels. Beyond the generative AI measures noted in the text, this can also trigger cross-border data obligations under the Measures for Security Assessment of Data Exports (数据出境安全评估办法) and the Measures on Standard Contracts for the Export of Personal Information (个人信息出境标准合同办法) (Cyberspace Administration of China, July 7, 2022; Office of Central Cyberspace Affairs Commission, February 24, 2023).On social media, the owner of a private API relay platform claimed that he had been detained by Shanghai police for 37 days for operating a gray-market relay service and is now out on bail pending trial while facing criminal charges. The authenticity of this claim, however, has not been independently verified (X/@YesterdayBigcat, May 13).
[3] China Telecom has not fully monopolized this layer of the token economy. Around the same time as it released its Xingcheng TokenHub, NetEase Youdao (网易有道) released a platform called ThinkFlow and China Mobile released its own platform, MOMA. Both platforms are neutral aggregation and routing platforms rather than LLMs, which distinguishes them from similar offerings such as Alibaba Cloud’s Bailian (阿里云百炼) and Tencent Cloud’s Token Plan, which are tied to those firms’ own model and cloud ecosystems (WeChat/@FaxAI, May 12).
[4] This pattern has appeared repeatedly in the PRC’s strategic industries, including electric vehicles, photovoltaics, and 5G telecommunications. Beijing sets the direction and builds a common framework, multiple firms compete within that framework, and the state retains coordinating authority over standards, certification, and market access. See Naughton, Barry, The Rise of China’s Industrial Policy 1978 to 2020, Universidad Nacional Autónoma de México, 2021.
[5] Participating companies included China Electronics Corporation (中国电子), China Mobile, CRRC (中车集团), Huawei, ZTE, Alibaba Cloud, Zhipu AI, iFLYTEK, Hygon Information (海光信息), Moore Threads, and MetaX (WeChat/@Cailian AI Daily, April 30).
The post The PRC’s Token Economy Takes Shape appeared first on Jamestown.
Lateo.net - Flux RSS en pagaille (pour en ajouter : @ moi) 