At the end of March, China inaugurated the World Data Organisation in Beijing, a body with a stated mission of “bridging the data divide, unlocking data’s value and powering the digital economy”.

The move is the latest signal of a broader trend: over the past several years, Beijing has developed a distinct data governance strategy to drive artificial intelligence (AI) development as it reshapes the terms of technological competition.

Since late 2025, Beijing has pursued an aggressive AI adoption strategy across industries through its “AI-plus” initiative. At the core of this shift is data, which Chinese policymakers see as the “new oil” powering AI. But, like oil, the supply of usable data is finite. In 2024, former OpenAI chief scientist Ilya Sutskever warned that the world had effectively exhausted human-generated data for AI training – a constraint China has taken seriously.

Rather than accepting this as a given, Beijing has reformed its data governance strategy to address this shortage. Over the past three years, China has reorganised its data-sharing regime to pool and channel the vast reservoirs of data generated by large-scale digitisation, feeding this data into specialised models poised to drive the next stage of AI development.

There are, broadly speaking, two types of AI models. The first are general-purpose frontier models like the ChatGPTs and Claudes, which are headline-grabbing giants trained on staggering volumes of data.

The second type is less glamorous but more economically consequential: specialised models. These sector-specific systems power many things like telemedicine diagnostics, financial fraud detection and transport planning. Rather than scraping data from the entire internet, they rely on highly specific data sets such as medical records, financial transactions and logistics flows.