On 7 February 2026 the National Data Administration, the Ministry of Industry and Information Technology, the Ministry of Public Security and the China Securities Regulatory Commission issued the Opinions on Cultivating Data-Circulation Service Institutions and Accelerating the Market-Based Valuation of Data Elements (Guoshu Zhengce [2026] No. 6). For the first time, the document makes clear that China will cultivate three types of data-circulation service institutions.
The positioning of the three types
Data exchanges (centres) emphasise comprehensive service functions, taking on roles such as transaction matching, rule-making and compliance review.
Data-circulation platform enterprises emphasise specialised development, providing data processing, matching and delivery services in specific industries or scenarios.
Data traders focus on stepping up the development of data products and services, and may expand data-acquisition channels through self-collection and sale, purchase for resale and other means, develop data products, provide collection, processing and circulation-transaction services, and explore new models such as high-quality datasets and data-as-a-service.
Why this matters for AI companies
AI model performance depends on data, and the compliance risk of data comes mainly from “source”. For a long time, companies purchasing or building datasets faced a practical difficulty: the authorisation chain of the data source was hard to prove completely.
The standardisation of data-circulation service institutions addresses exactly this problem. When data circulates through institutions with a defined functional positioning and compliance duties, its source legality, scope of authorisation and use boundaries gain a clearer vehicle. For AI companies, this means “data procurement” can gradually move from “private transactions” to “traceable transactions”.
Four implications for AI companies
First, the burden of proof on data sources is not reduced. Even when purchasing through a data trader, a company must still clarify the scope of authorisation and use restrictions in the contract and retain the authorisation chain. See our data compliance service and commercial contracts service.
Second, data traders themselves become compliance subjects. Data traders participating in the development and use of public data resources must comply with public-data management requirements; where personal information is involved, they must perform personal information protection duties. See the PI outbound certification measures.
Third, high-quality datasets are where policy meets the market. Data traders developing high-quality datasets align with the National Data Administration’s industry dataset-building direction; see the action plan for high-quality industry datasets and the National Data Administration’s dataset guidance.
Fourth, data property rights registration will become circulation infrastructure. Data property rights registration and the cultivation of data-circulation service institutions form a complete chain; see the guidance on data property rights registration.
Recommendations
Companies can proceed in three steps: first, map existing data sources, marking which come from public channels, which from procurement and which from user generation, and assess the completeness of each one’s authorisation; second, assess whether “cleaner” data can be obtained through data exchanges or data traders; third, make data-source compliance a project-initiation checklist item in model development.
The standardisation of the data-element market will, over the long run, reduce AI companies’ compliance uncertainty, but in the short term it raises the standardisation cost of “data procurement”. Whether a company can adapt to this shift will affect the sustainability of its model training.
This article is based on the Opinions on Cultivating Data-Circulation Service Institutions and Accelerating the Market-Based Valuation of Data Elements issued by the National Data Administration and three other departments, and public reporting. It is for general reference only and does not constitute legal advice.