Use of exclusive data for corporate research on machine learning and artificial intelligence: Implications for innovation and competition policy

Corporate research has been a primary driver of recent advances in Machine Learning and Artificial Intelligence (ML/AI). The present study contends that firms' prominent role in advancing the ML/AI research field is partly attributed to their use of exclusive data for ML/AI research. Using data on nearly 8000 preprints of ML/AI research papers archived in arXiv and the performance of their proposed algorithms, we found multifaceted evidence that corporate ML/AI research has exhibited a particularly significant citation impact compared to non-corporate research. Importantly, we showed that the significance of corporate research is more pronounced when it originates from the use of exclusive data. We argue that firms' use of exclusive data has been instrumental in not only encouraging their research on ML/AI, but also enhancing the research impact, which we call the “dual role” of the data in corporate research on ML/AI. In light of the policy concern regarding the potential anticompetitive implications of firms' utilization of data exclusivity in the evolving landscape of ML/AI, our conclusion calls for a comprehensive policy discourse on the consequences of firms' exclusive use of data for their ML/AI research within broader dimensions of societal welfare, including innovation and competition.