On 3 April 2026 the National Data Administration published a notice opening public consultation on the Guide on Data Property-Rights Registration (Trial) (draft for comment), with comments open until 19 April. The Guide aims to establish and improve the data property-rights system, build a unified national registration system and help cultivate a nationally integrated data market.
Data needs an identity document too
A recurring obstacle in the marketisation of data elements has been that it is hard to say who data “belongs to”. The same dataset may be jointly generated, held and used by multiple parties, with a complex chain of rights, making it difficult in a transaction to say who can dispose of it or license it.
Registration is meant to solve this. The Guide sets out a systematic approach to registration-body management, registration procedures and the use of registration certificates, issuing an “identity document” for data as an intangible asset and clarifying who holds it, how it may be used and who may use it. For buyers, a registration certificate reduces due-diligence cost; for holders, it is evidence for asserting rights and conducting transactions.
Registration is disclosure, not title
It is worth distinguishing data property-rights registration from real-estate registration or patent grant. It is closer to a disclosure and evidence mechanism: registration does not necessarily create rights, but records rights claims, holding status and licensing scope in a unified system so they can be checked and verified.
The practical effect cuts both ways. On the upside, transaction certainty improves; on the downside, if the registered content does not match reality, the registrant bears the consequences. Before registering, a company should therefore get its internal data ledger, source evidence and licensing chain in order — otherwise registration may lock in an error.
What it means for AI companies
AI companies typically draw data from three sources: their own business data, publicly collected data, and third-party purchased or licensed data. The rights basis for each differs fundamentally.
First, training data must have an explainable source. Data property-rights registration pushes for traceability, raising the bar on the lawful sourcing of training data — see our data-compliance service.
Second, data assets can become a subject of financing and capital contribution. For AI companies whose core asset is data, the evidentiary value of registration will increasingly show in financing, equity contribution and M&A — see our M&A and exit service and financing and equity service.
Third, companies in public-data authorised operation should watch the relationship between registration and licensing. The National Data Administration’s parallel push on high-quality industry datasets and property-rights registration are two links in the same chain — see our analysis of the action plan for high-quality industrial datasets.
The registration system is still at the trial stage and will be adjusted with pilots. But the direction is clear: the more valuable data becomes, the more its source and ownership must be explainable. Rather than waiting for the system to settle, companies are better off building the ledger now — see our data-compliance service.
This article is based on public information concerning the National Data Administration’s notice of 3 April 2026 opening public consultation on the Guide on Data Property-Rights Registration (Trial) (draft for comment). It is for general reference only and does not constitute legal advice.