How did the Chinese labs catch up so fast if most of the early training work was in the US?

Chinese AI labs like DeepSeek rapidly closed the gap with US counterparts such as OpenAI and Google, despite the latter's significant head start in training. This quick advancement is attributed to factors like open-source papers and computational resources. However, the origin of their extensive datasets remains a key question, with speculation including US data labs selling information and other undisclosed sources.

This discussion uniquely focuses on the data acquisition strategies of Chinese AI labs, unlike other analyses that primarily credit open-source papers and compute for their rapid progress.