China's top data regulator on Monday rolled out a directive for building high-quality datasets across key industries.
The directive, officially, known as the Implementation Scheme for Promoting the Construction of High-Quality Datasets in Industries, was promulgated by the National Data Administration. It outlines six major operations including data supply, circulation and application to accelerate the integration of high-quality data and artificial intelligence (AI).
The directive calls for sustained efforts to build high-quality multimodal datasets covering text, images, audio and video, aligned with the practical demands of AI applications.
It focuses on priority frontiers including AI agents, embodied AI and world models, requiring accelerated dataset development in these directions.
It also guides eligible local regions to launch data annotation innovation pilot zones, tailored to their local strengths.
"This implementation scheme makes a systematic planning for dataset construction encompassing the entire chain. With the focus on key and innovation domains such as scientific research, industrial manufacturing, the low-altitude economy and embodied AI, it aims to advance dataset construction with a focused, needs-driven approach. In parallel, it drives the transformation and upgrade of data annotation, so as to comprehensively raise both the supply capacity and quality of datasets," said Hu Jianbo, president of the National Institute for Data Development.
China unveils directive for building high-quality datasets across key industries
