On 1 March 2025 the national public data resource registry platform officially went into operation. The interim measures on the administration of public data resource registration took effect the same day, with a validity of five years.
One standard, two tiers
The platform adopts a one-standard, two-tier architecture: the national registry platform is chiefly responsible for registration business of central and state organs, their directly subordinate institutions and central enterprises, and for the time being also accepts applications from some provinces that have not yet completed platform construction; provincial platforms run in step with the national platform. A number of localities have already carried out registration work on the national platform.
The point of the two-tier architecture is to unify the registration standard while retaining local administration, so that rules in different places do not conflict.
What registration solves
The difficulty with public data resources is unclear ownership and responsibility: the data is held by public bodies, but who may authorise its use and on what terms has long lacked a clear route. The registration regime tries to create identifiable ownership records and circulation vouchers for public data resources.
For AI companies this means that obtaining public data for model training may in future have to be done through a registration and authorisation chain, rather than on the strength of a cooperation agreement alone.
How it connects to other data-factor regimes
Registration is not an isolated regime. See our notes on data intellectual property registration: that regime addresses the confirmation of rights in data outputs, while the National Data Administration’s guide on building datasets addresses the supply quality of datasets.
See the action plan for building high-quality industry datasets: registration, rights confirmation and supply quality together make up the infrastructure of the data factor market.
What companies can prepare
First, map the public data sources available to the company, identifying which data comes from public bodies, how it was obtained and whether there is a chain of authorisation.
Second, write the lawful origin of data into training documentation. See our notes on data security risk assessment: proof of the lawful origin of training data is core material in filing and in responding to a dispute.
Third, watch the practical scope of registration and authorisation. In the early period of implementation, localities may differ on the scope of registration and the process, so the guidance actually issued should govern.
The institutional build-out of data as a factor will ultimately determine what training data AI companies can obtain and at what cost.
This article is based on material from the National Data Administration and the National Development and Reform Commission on the national public data resource registry platform and public information and is for reference only; it does not constitute legal advice.