On October 28, the National Data Administration convened a symposium on the work of chain leader enterprises in building high-quality industry datasets. The meeting emphasized the need to develop a clear understanding of data requirements for artificial intelligence applications. Chain leader entities should ground their efforts in their own needs, proactively expand external demand across the upstream and downstream of the industrial chain, and adopt a demand-oriented approach to precisely advance dataset construction. They should explore profit distribution and commercial return mechanisms on a pilot basis, map out industry data landscapes, promote the distributed aggregation of industry data, and guide industry datasets to be registered in the national dataset management service system. They should also establish data feedback and iterative optimization mechanisms based on actual application scenarios, continuously collect user feedback, promote the dynamic updating of datasets, and enhance the compatibility between data and application scenarios. Finally, they should give full play to the role of "gathering the chain" by linking upstream and downstream entities in the industrial chain, expanding the scope of cooperation, and promoting the development of the industrial ecosystem.