Under the guidance of the China Cyberspace Security Association, Frost & Sullivan and LeadLeo jointly released the group standard 'Industry Research Data Governance Classification and Grading Guide'.

Data has become a core production factor in the digital economy era. The "Data Security Law" specifies that the state establishes a data classification and grading protection system. Based on the importance of data in economic and social development, as well as the potential risks when data is altered, damaged, leaked, or illegally obtained or used, data is protected through classification and grading.

The "Opinions of the CPC Central Committee and the State Council on Establishing Data Basic Systems to Better Leverage Data Elements" proposes building data basic systems around data ownership, circulation and trading, income distribution, and security governance. The National Data Bureau's "Development Action Plan for Trustworthy Data Spaces (2024–2028)" further states that standards for data classification and grading should be developed quickly. By 2028, more than 100 trustworthy data spaces will be established, basically forming a network of trustworthy data spaces with extensive interconnection, resource aggregation, ecological prosperity, value creation, and orderly governance.

Industry research is a knowledge-intensive scenario that connects industrial policies, capital markets, and corporate decisions. As an important asset, whether the source of research data is clear, its ownership is clear, and its processing is traceable directly affects the credibility of research results, as well as the compliant circulation and value realization of data elements in the field of industry research.

The formulation of this standard precisely applies the top-level requirements of a "trustworthy data ecosystem" to the data governance foundation of industry research.

Artificial intelligence is reconstructing the way information is produced and research conclusions are disseminated. Algorithms, computing power, and data are key elements, but issues such as AI hallucinations, text pollution, and data poisoning also make it uncertain whether "research conclusions are correctly understood".

Industry research has long had problems such as opaque research processes, difficulty in reproducing results, and insufficient traceability of data sources. When large models are widely involved in information organization and content generation, low-quality texts will reverse-taint research conclusions, and enterprises may also lose accurate expression during misreading and simplification.

For research institutions, the core issue is no longer just "whether it is seen", but "whether it is correctly recognized". And the premise of correct recognition is that each piece of data can answer three questions:

• Where does the data come from;

• What kind of processing has been done;

• Can the conclusion be traced, verified, and reproduced?

Trust cannot remain at the level of expression, but must enter the data governance process.

"Guidelines for Industry Research Data Governance and Classification and Grading" combines the characteristics of industry research business to establish a scientific, standardized, and operable data classification and grading system, focusing on achieving three goals:

First, standardize data asset management to solve problems such as unclear data inventory, unknown sources, and inconsistent metrics;

Second, identify data security risks and provide differentiated control criteria for core data, important data, and general data;

Third, promote standardized use and orderly circulation of data while ensuring security.

The main contents of the standard include:

1. Three-level classification structure. Using a hybrid classification approach, a classification system of "industry field—business process—specific attributes" is built. The first-level classification aligns with GB/T 4754 "National Economic Industry Classification", enhancing the standardization of cross-institutional data exchange; the second-level classification covers the entire process of collection and acquisition, research and production, application of results, and circulation and trading, including "process layer" data such as research minutes, analysis sheets, research drafts, and model codes into the scope of governance.

2. Three-level classification rules. Following the "Data Security Law" and GB/T 43697 "Data Security Technology: Data Classification and Grading Rules", a classification framework for core data, important data, and general data is established, and practical identification and judgment methods are provided from dimensions such as field, group, region, accuracy, scale, coverage, depth, and importance.

3. Dynamic management mechanism. Considering the strong timeliness of industry research data, a grading mechanism is established: expert interviews processed with masking can be reasonably downgraded, and when fragmented data is aggregated to form a panoramic picture of the industrial chain, higher-level control should be applied, so that the data level can be dynamically adjusted according to business scenarios.

4. Trustworthy traceability principle. The standard clarifies that classification and grading should serve the credibility and transparency of the entire research process, fully considering the authority of data sources, the legality of collection, and the traceability of processing, ensuring complete, true, and tamper-proof data flow records.

5. Full-process implementation guidelines. From preparation, asset sorting, implementation of classification and grading, to identification control, review and supervision, and continuous improvement, it provides an operational path covering the entire life cycle for research institutions, consulting companies, think tanks, and enterprise research departments.

*The full text of the standard can be viewed on the National Group Standard Information Platform (https://www.ttbz.org.cn/).