On June 8th, it was learned from the National Data Administration that the bureau has recently issued the "Implementation Plan for Promoting the Construction of High Quality Industry Data Sets" (hereinafter referred to as the "Implementation Plan"), proposing to build a batch of industry high-quality data sets covering key areas and verified by application by the end of 2028, create a batch of typical application scenarios for data-driven artificial intelligence innovation and development, cultivate a group of innovative data enterprises and professional talents with leading advantages, and form a batch of industry high-quality data set construction tools and standards.

Experts say that artificial intelligence is accelerating from being "usable" to being "easy to use". High quality datasets serve as the cornerstone of large model training and application, and the scale and quality of supply directly determine the innovation height and industrial landing depth of artificial intelligence. Accelerating the construction of high-quality datasets in the industry will directly benefit multiple industrial sectors such as data annotation, data services, and model applications.

Data shows that as of the first quarter of 2026, over 116000 high-quality datasets have been built nationwide, with a total volume of over 960PB and an average daily call volume of over 140 trillion tokens.

Chang'an Yun, from the Artificial Intelligence Research Institute of the Saizhi Industry Research Institute, told reporters that currently, artificial intelligence is evolving from general dialogue and content generation to intelligent agents, scientific intelligence, embodied intelligence, and world models, which puts higher requirements on the professionalism, structure, scenario, and verifiability of industry data. However, there are still some problems in industry data, such as dispersion, low quality, different standards, weak model adaptation, and insufficient application verification, which restrict the in-depth application of AI in key industries and complex scenarios.

The Implementation Plan focuses on key industries for national economic development and strategic emerging industries, and deploys six specialized actions around key links such as the supply, circulation, and application of high-quality data sets in the industry, including strengthening the foundation and expanding capacity, tackling annotation challenges, improving quality and efficiency, empowering applications, managing services, and releasing value. The goal is to form a "data flywheel" of "scenario driven data, data-driven models, model empowered applications, and application created value", and accelerate the construction of a symbiotic ecosystem between data elements and artificial intelligence for collaborative evolution.

For example, in the implementation of the Strong Foundation Expansion Action, the Implementation Plan proposes to broaden data supply channels, enrich data supply types, accelerate the construction of high-quality industry datasets, and provide sufficient "fuel" for the development and application of artificial intelligence. Among them, it is clear to focus on key areas such as scientific research, industrial manufacturing, agriculture and rural areas, smart energy, transportation, financial services, as well as innovative areas such as low altitude economy, embodied intelligence, intelligent driving, smart ocean, and biomanufacturing, and accelerate the construction of high-quality datasets in the industry.

In implementing the annotation campaign, we will continue to promote data annotation as a pioneer. Cultivate a group of data annotation leading enterprises, unicorn enterprises, and gazelle enterprises, and strengthen the data annotation industry. In terms of implementing value release actions, the "Implementation Plan" proposes to innovate the business model of high-quality industry datasets, including "exploring new transaction models such as word element transactions, and building a data value system based on word elements that can be quantified and priced".

Analysts believe that the series of measures will also bring development benefits to multiple industrial sectors such as data annotation, data services, and model applications.

Lu Feng, the director of Beijing Frontier Future Technology Industry Development Research Institute, stated that in the data annotation process, the "Implementation Plan" clearly cultivates data annotation leading enterprises, unicorn enterprises, etc., and promotes the professionalization and intelligence of data annotation; The series of measures emphasize the strengthening of key technologies such as data cleaning, enhancement, and quality inspection, and encourage the use of simulation synthesis technology to solve the problem of scarce scenario data, which directly benefits technology driven data service providers; In addition, the application and circulation of the model will accelerate the release of value. The Implementation Plan requires the creation of a "data flywheel" closed loop and the cultivation of consensus in the data payment market, which will clear obstacles for data circulation and commercial monetization.

An Yun believes that high-quality datasets in the construction industry mainly involve industrial chain links such as data resource supply, processing and annotation, quality improvement, model application, management services, and value release. Especially, the construction of high-quality datasets in the industry will prioritize key industries and emerging fields with abundant data resources, clear application scenarios, and urgent demand for artificial intelligence empowerment. For example, innovative fields such as low altitude economy, embodied intelligence, intelligent driving, smart ocean, and biomanufacturing have strong demands for multimodal data, simulation data, physical interaction data, and high-precision annotation data, which will drive the development of data collection, simulation modeling, scene testing, and model training.

This will also further open up development space for related industry enterprises. Lufeng believes that relevant enterprises can deeply cultivate the closed-loop of technology tools and scenarios, and focus on investing in key technology research and development such as data cleaning, synthesis, and quality inspection; At the same time, we will delve into key industries and strategic emerging industries, build a "demand data model" closed loop, and create reusable benchmark cases. An Yun stated that data resource enterprises need to accelerate the consolidation of their data resource base, focus on the application needs of key industries, build high-quality datasets that can be used for model training, application verification, and circulation transactions, enhance the productization and serviceization capabilities of data resources, and provide stable data supply for the upstream and downstream of the industrial chain.