Global data empires: Analysing artificial intelligence data annotation in China and the USA

As the two leading countries in the development of artificial intelligence (AI) systems, China and the United States largely rely on separate AI infrastructure and data annotation ecosystems. Studies have focussed almost exclusively on data annotation associated with American and European companies, limiting our understanding of how this contrasts with the Chinese development of AI. This article provides a comparative analysis of the political economy of the Chinese and American/European AI data annotation ecosystems, focusing on the role of the state and the practice of outsourcing to data annotation institutions. It finds that while the US state plays a protectionist role concerning AI infrastructure such as semiconductors and data centres, it adopts a laissez-faire approach to data annotation. The Chinese state, however, understands it has a comparative advantage in data and invests heavily in its own data ecosystem while maintaining stringent regulations for Chinese tech companies. Secondly, many US companies outsource data annotation work to business process outsourcing centres and digital platforms, whereas Chinese companies maintain these activities in-house or, through a process of ‘inland-sourcing’, send this work to ‘third-tier’ cities in Chinese provinces to data labelling bases, often jointly managed by local government and private companies.