In November 2024 the National Data Administration issued the Trusted Data Space Development Action Plan (2024-2028), numbered Guo Shu Zi Yuan [2024] No. 119. The plan’s purpose is narrow and consequential: to make data circulate between organisations at industrial scale without each transaction turning into a fresh legal negotiation.
The problem a trusted data space solves
Data has an awkward legal shape. It is easy to copy, hard to own exclusively, and its value depends on combining it with data held by someone else. The result is that the most valuable uses — cross-company training, supply-chain optimisation, joint research — stall on liability questions rather than technical ones. A trusted data space is an attempt to standardise those questions: agreed rules on who may use what, for which purpose, with what logs and what remedies. The Action Plan builds on the earlier opinion on data circulation service institutions and on the data property rights registration guide, which together supply the institutional and registration layers.
What it connects to
The plan does not stand alone. Public data supply is being organised through the national public data registry platform and through the public data authorisation and operation rules. On the quality side, the industrial high-quality datasets action plan addresses the same problem from the input end: a data space is only useful if the datasets inside it are usable. The technology layer is picked up in the data technology innovation opinion.
Implications for AI companies
For AI companies, the practical question is whether to join an industry data space or to build bilateral data deals. Joining usually trades negotiating freedom for speed and legitimacy: participation signals to counterparties and regulators that the data handling is documented. It also imposes obligations, since membership conditions typically require logging, purpose limitation and audit cooperation, and those conditions have to survive contact with a company’s own model-training pipeline.
Before committing, a company should map which datasets it contributes and which it consumes, confirm it has a lawful basis for each, and check whether cross-border flows are involved — a trusted data space does not by itself resolve the export question.
A second point is governance. Data spaces only work if someone is accountable for the rules, and the plan expects service institutions to play that role. Companies should therefore read the participation agreement as carefully as they would a joint venture: liability allocation, exit rights and what happens to derived models after a member leaves are all questions that a framework document will not answer for them.
This article is based on the Trusted Data Space Development Action Plan (2024-2028) (Guo Shu Zi Yuan [2024] No. 119) and public information and is for reference only; it does not constitute legal advice.