Данные как топливо: Китай создает легальный рынок для обучения нейросетей

The era when AI developers simply vacuumed the internet for texts and images is coming to an end. Everyone has hit a "data wall": there's less quality content in the public domain, and more lawsuits from rights holders. In this situation, China decided not to wait for market mercy and engaged the state apparatus.

The National Data Bureau, together with the Ministry of Industry and other agencies, released a document that effectively creates a new industry in the country — a legal market for information trading. At the center of the new strategy lie three types of organizations that should turn raw bytes into a valuable asset. The first are data exchanges, official platforms where transactions take place under regulatory supervision.

The second are service platforms that provide the technical side of things, from storage to computing. And perhaps the most interesting are "data dealers" or data providers. These are companies that will professionally engage in searching, cleaning, and structuring information for specific customer needs.

It sounds like a profession from a cyberpunk novel, but for China this is now official economic reality. Why is this needed right now? The answer lies in the national "AI Plus" program.

Beijing understands that technological leadership in LLM and robotics is impossible without gigantic volumes of specific data: medical, industrial, logistical. Most of these treasures are currently locked in local databases of ministries and corporations. Creating legal intermediaries is a way to "open the archives" and give model developers legal access to them without (in theory) violating security laws.

After last year's creation of the National Data Bureau, many wondered what this initiative would lead to. Now the answer is clear: China is building an infrastructure where data becomes as much a commodity as lithium or oil. If previously training a neural network on closed government data was a grey zone, now it's becoming a state-encouraged business.

This gives Chinese companies a colossal advantage — the ability to train models on real data from the government sector and industry that companies like OpenAI will never get access to. Of course, the question of implementation remains. A data market is a capricious substance.

How do you value a patient database or the logs of a power plant's operation? How do you guarantee that data won't leak to the black market? Chinese authorities plan to solve this through strict certification of those very "data dealers."

Effectively, the state is creating a closed club of trusted suppliers that will become a bridge between information owners and AI creators. This is an attempt to organize chaos and turn the "AI race" into orderly industry building. The bottom line: Beijing is betting on data legalization and commercialization.

While the rest of the world argues about scraping ethics, China is building information processing plants. Whether state regulation can create a flexible enough market for AI needs — that's the main question for the coming year.