On 11 January 2024 the Ministry of Finance issued the Guidance on Strengthening Data Asset Management (Cai Zi [2023] No. 141). The document contains eighteen articles in three parts and places whole-process management, ownership, benefit distribution and security compliance in a single framework. It connects with the provisional accounting rules for data resources that took effect on 1 January 2024: those rules govern whether data can enter the financial statements, while the guidance governs how it is managed before and after recognition.

For AI companies the change is not a new ledger account. It reframes “what is our data worth” as “why is the data ours, how is the cost calculated and who bears the risk”. Recognition is the accounting outcome; compliance is the precondition.

Recognition rests on rights and measurement

Accounting recognition requires a company to show that it holds rights in the data resource, that cost can be measured reliably and that future economic benefits are expected. In practice this means three sets of documents: evidence that the source is lawful, internal and external arrangements on ownership, and complete vouchers for cost aggregation. If any set is missing, recognition tends to become a bookkeeping exercise that does not survive audit.

Sources are usually mixed: public collection, customer input, third-party purchase and business-generated data. Each carries a different licensing basis, and one dataset may contain several rights statuses at once. The work is to record those statuses separately rather than cover everything with a general authorisation. For training corpora, whether the licence covers training, fine-tuning and external service must be checked item by item.

Cost aggregation and capitalisation

The provisional rules capitalise qualifying data resources held for internal use as intangible assets or inventory. Capitalisation depends on whether expenditure is attributable to the development or acquisition of the resource and whether the recognition conditions are met. Purchase price, cleaning and labelling, system development and security spending must be allocated between cost and expense on a reviewable basis.

Two errors recur. One is capitalising all data-related spending, leaving later amortisation and impairment without support. The other is keeping only purchase invoices and ignoring the labour and compute invested in self-built data. Auditors look less at the amount than at consistency of method and whether the evidence closes.

Limits of authorised public data

The guidance brings authorised operation of public data within data asset management. Authorisations usually carry limits on purpose, duration, territory and onward provision. An authorisation is not a transfer of ownership, nor permission to resell or to use the data directly in external model services. Using authorised data for training, fine-tuning or external output requires checking the scope and whether additional approval or filing is triggered.

Pricing, benefit distribution and usage records of public data are also within the document’s scope. For operators, a separate project ledger is more effective than an explanation after the event.

Security, audit and continuing obligations

Once recognised, security compliance shifts from a project requirement to an asset management requirement. Classification, access control, processing records, incident response and audit trails must correspond to the carrying value. State-owned enterprises and large platforms also face internal rules, allocation of responsibility and periodic assessment.

Recognition does not create rights. It expresses existing rights and costs in accounting language. Data with defective title or unclear provenance does not become clean because it sits in an asset account.

For a provenance and licensing review, see our data compliance service; for ownership, licensing and registration, see IP protection; definitions are collected in the glossary.


This article is based on the Guidance on Strengthening Data Asset Management (Cai Zi [2023] No. 141) issued by the Ministry of Finance on 11 January 2024, and on public information. It is for reference only and does not constitute legal advice.

Geason Law Offices AI Industry Research Group | Reviewed by Ho Qianggao