Digital News

Recently, the National Data Administration issued a notice on the Implementation Plan for Advancing High-Quality Dataset Development Across Industries (hereinafter referred to as the Implementation Plan). The plan focuses on key links of high-quality industry datasets, including supply, circulation, and application, and sets out six major actions: strengthening foundations and expanding capacity, tackling annotation bottlenecks, improving quality and efficiency, enabling industrial applications, enhancing management services, and unlocking data value. It also seeks to build a “data flywheel” in which scenarios drive data, data power models, models enable applications, and applications, in turn, generate value. In doing so, it aims to accelerate the development of a symbiotic ecosystem where data elements and artificial intelligence evolve in tandem.

The Implementation Plan sets the target that, by the end of 2028, a range of high-quality industry datasets will have been built and validated through real-world applications across key sectors. It also aims to develop a series of benchmark application scenarios demonstrating data-driven innovation in AI, cultivate a group of leading innovative data enterprises and skilled professionals, and establish a suite of tools and standards for the development of high-quality industry datasets. By that time, a virtuous cycle from data supply to value realization will be largely in place. The role of data in enabling AI-driven innovation will be more fully realized, with deeper integration between the data industry and AI, continuously generating new growth drivers for the intelligent economy.

Action 1: Strengthening Foundations and Expanding Capacity.This action responds to the accelerating penetration of AI across industries, as it shifts from dialogue-based systems toward multimodal generation, decision-making and execution, embodied AI, and physical interaction. It aims to broaden data supply channels, diversify the types of data supply, and accelerate the development of high-quality industry datasets, providing the necessary “fuel” for AI development and applications. It focuses on advancing high-quality dataset development across industries, consolidating foundational pathways for dataset development, diversifying high-quality dataset formats in line with AI application needs, and strengthening organic linkage with the development of data infrastructure.

Action 2: Tackling Data Annotation Bottlenecks.Data annotation is the process of embedding knowledge and expertise into training data and is a critical component in the development of high-quality industry datasets. This action promotes a shift from a predominantly human-dominated approach to a multi-level model involving human-machine collaboration and in-depth expert participation, advancing data annotation toward greater specialization and intelligence. It aims to drive the transformation and upgrading of data annotation, carry out pilot trials on an ongoing basis, and increase the supply of data annotation talents.

Action 3: Improving Quality and Efficiency.This action promotes the development of AI-ready, high-quality datasets that meet standards for structural integrity, content diversity, annotation accuracy, and model adaptability. It aims to reduce training and inference costs while improving model performance. It seeks to enhance the quality and efficiency of high-quality industry dataset development, advance the establishment and implementation of relevant standards, and strengthen quality assessment and mutual recognition of assessment results for such datasets.

Action 4: Enabling Industrial Applications.This action promotes the deep integration of high-quality industry dataset development with real-world applications, fostering a virtuous cycle in which models generate demand for data and data, in turn, enhances model performance. It encourages coordinated progress between high-quality dataset development and initiatives such as “Data Elements ×” and “AI Plus,” thereby comprehensively enabling digital and intelligent transformation across industries. It seeks to build a closed-loop “data flywheel” for applications, develop industry application benchmarks and model cases, and foster a vibrant ecosystem for collaborative dataset development.

Action 5: Strengthening Governance and Service Support.This action aims to enhance dataset management, improve data ethics and governance mechanisms, and advance the implementation of policies related to data rights and interests, thereby fostering a more standardized and well-organized framework for dataset development. It seeks to establish a full lifecycle management system for datasets, explore data-related systems tailored for AI development, and adhere to the principles of prioritizing ethics, fairness, and universal accessibility.

Action 6: Unlocking Data Value.This action seeks to unlock the application value of datasets and leverage high-quality industry datasets to advance the development of artificial intelligence. It aims to unlock the value of data as a production factor, promote the commercialization and assetization of datasets, and foster market consensus on paying for data. It also explores a token-based value framework. It further seeks to enhance the application value of high-quality industry datasets, foster innovative business models, and explore new pathways for dataset assetization, while building market consensus on paying for high-quality data.

Source: Official WeChat Account of People’s Posts and Telecommunications News(RENMIN YOUDIAN)

Reporter: Su Deyue