On 1–2 April 2024 the National Data Administration held the first national data work conference since it was established. The meeting set out work in eight areas — strengthening basic systems, releasing the potential of data as a production factor, accelerating transformation, promoting scientific and technological innovation, improving infrastructure, reinforcing security, advancing international cooperation and running pilot programmes — and stated that a data property-rights system would be established. The order of the eight tasks is itself a statement of method.
The order matters
Basic systems come first and pilots come last. The sequence suggests that market development will not proceed by deploying scenarios and patching rules afterwards; the framework for rights, circulation, benefit-sharing and security is to be settled first, then tested through pilots. The middle items form the connective layer between rules and industry.
International cooperation appears as a separate task, which places data governance in the context of external trade rather than domestic supervision alone.
What property rights must answer
Three questions sit at the core: who holds the data, who may use it, and how benefits are shared. In practice they are entangled. Platforms argue that they bear the cost of collection and curation; sources argue that they supplied the original information; public data is held by government departments.
A property-rights system moves these disputes from case-by-case negotiation into a rule framework. For AI companies, the arrangement directly affects whether training corpora can be obtained reliably and how ownership of model outputs is allocated in contracts.
Four parallel tracks
Property rights, circulation and trading, benefit-sharing and security governance run in parallel rather than in sequence. The common error is to focus on one: emphasising circulation while neglecting security, or asserting rights while leaving benefit-sharing to an unsigned appendix.
Three actions follow. The provenance and licence scope of training data must be explainable. Providing data products externally requires a check against circulation and outbound-transfer rules. Contracts with data suppliers should state use boundaries and ownership of derivative outputs.
What to watch after the meeting
The conference is deployment, not legislation. The useful signals are pilot programmes in specific regions, supporting rules from sector regulators, and how public-data authorisation actually operates. These determine how far the eight tasks become enforceable requirements.
There is no need to wait for every rule. Assembling provenance, licence chains and security measures into verifiable material is basic work under any regime.
For data provenance and licensing, see our data compliance service; for ownership and licensing of data products, see IP protection. Definitions are collected in the glossary.
This article is based on public information from the national data work conference held on 1–2 April 2024. It is for reference only and does not constitute legal advice.