As the fifth scientific research paradigm, AI4S is an important tool to promote the development of the new quality productive forces in the field of scientific research, and one of its components is data, which is the basis for the training of AI4S big models, and it needs high-quality and multi-type data. At a time when the utilization of data for commercial big models is becoming more and more limited, it is especially important to pay attention to the supply of data elements for scientific research big models in time, so as to accelerate the release of the new quality productive forces of AI4S from the code. The supply of data elements for AI4S is a complex systematic project, and the value orientation should be data altruism rather than data egoism, so as to realize the public welfare purpose of scientific research and the empowerment goal of serving the society, the corresponding paths should be chosen according to different types of data, i.e., the supply paths of public data, enterprise data, personal data, scientific data, and artwork data, etc. that should be chosen respectively, such as the supply paths of conditional gratuitous, cost-compensated, voluntary agreement, mutual, sharing, and fair use. And at the same time, attention should be paid to preventing and controlling the potential risks of copyright infringement, privacy disclosure, data breaches, data leakage and value alignment.