A company sits on a valuable dataset but cannot prove it owns the rights to license it. That uncertainty is why a lot of Chinese data never enters the market. A new draft rule tries to give buyers and sellers something to stand on.
Key takeaways
- The National Data Administration (国家数据局) published the Data Property Rights Registration Guidelines (Trial) as an open draft for comment (公开征求意见稿) on 3 April 2026.
- The public comment window ran from 3 April to 19 April 2026 and drew 604 submissions; the rules are still a draft and are NOT yet in force.
- The draft defines data property rights as three separable rights — holding (持有权), use (使用权) and operation (经营权) — the so-called "structural separation" of rights.
- Next step is a pilot rollout, not immediate nationwide enactment; the document is explicitly a guidance document, not a binding law.
What the draft actually says
The draft applies within China to the registration of rights over data resources and data products of all forms. It defines "data property rights" as the property rights a right-holder enjoys over specific data, including the three rights above. "Holding" means holding or entrusting others to hold lawfully obtained data. "Use" means processing, aggregating or analyzing data to improve operations or create derivative data. "Operation" means providing data to others through transfer, licensing or contribution.
This three-way split follows the 2022 "Data Twenty Articles" (数据二十条) issued by the CPC Central Committee and the State Council, which proposed separating holding, processing-use and product-operation rights rather than fighting over who "owns" data outright.
How registration would work
The draft sets out a seven-step closed loop: application, acceptance, review, public notification, objection handling, information evidence-preservation, and certificate issuance. It specifies timelines and operating rules for each step and sets conditions for selecting registration institutions, which would be legal persons included in a national directory managed by the national data authority.
In other words, the draft is building the procedural plumbing for a unified national data-rights registry — a trusted layer that, in theory, lowers information asymmetry and gives later transfers, licensing and protection a clear anchor.
Why it matters
Today, many actors are reluctant to use or trade data because they cannot clearly answer "whose is it, and how are the rights split?" The draft is an attempt to make data a legible, registerable asset. Xinhua noted that local pilots in Shanghai, Fujian, Zhejiang and Guizhou already explored registration but differed on process, review standards and mechanisms — exactly the inconsistency a national guideline is meant to resolve.
For AI specifically, cleaner data-rights registration matters because training-data sourcing and licensing depend on being able to show provenance and authorized use.
How it fits the wider data agenda
The draft is one tile in a larger mosaic. The National Data Administration's 2025 work points already named "accelerating the building of a data property-rights system" as a priority, pushing authorization-based use of public, enterprise and personal data and piloting basic institutions in comprehensive data-element test zones. The registration guidelines translate that abstract mandate into procedure. They do not by themselves resolve harder questions — how personal-information rights interact with data-operation rights, or how cross-border data fits a national registry — which the draft leaves to later rules and the pending pilots.
The stage matters — this is a draft, not a law
The single most important point for readers: as of the run date, this is an open draft that has finished its comment period. It has not been enacted, has no effective date, and the administration says it will be advanced "through pilot methods." Treating it as "already in force" would be a hard error. The 604 comments are being studied; a finalized version may change.
Honest limitations
This article is based on the National Data Administration's official announcement and the Xinhua report carried on nda.gov.cn, plus the draft text. We did not see the consolidated feedback or any later finalized version, which was not available at writing. The draft's definitions are reproduced from the official text but we have not assessed how they will interact with existing personal-information or data-security rules. Whether pilots will launch in 2026 and in which provinces remains unconfirmed. Conversions and legal-effect claims are deliberately conservative: it is a guidance draft, not statute.
What readers can do now
- If you handle data commercially in China, monitor the finalized guideline and any pilot provinces before relying on registration for deals.
- If you are a compliance lead, map your data holdings against the three rights now — the split will shape future licensing language.
- If you follow AI policy, pair this draft with the broader "data elements" (数据要素) reforms, but do not assume registration is mandatory yet.