Key Takeaways:
China's National Data Administration issued its first systematic national framework for data-powered AI development, targeting 2028 as the deadline for industry-wide high-quality dataset deployment.
China's data regulator on June 3 published its first systematic national framework for using data to power artificial intelligence, mandating the construction of high-quality datasets across more than 20 industries by 2028 as the government seeks to accelerate AI adoption beyond the consumer internet.
"The plan addresses the full chain of dataset construction, from collection and annotation to application and value realization," Hu Jianbo, president of the National Data Development Research Institute, said in a state media interview. "It targets key and emerging fields including scientific research, industrial manufacturing, low-altitude economy and embodied intelligence."
The action plan, released under document number Guo Shu Ke Ji [2026] No. 25, outlines six special campaigns covering foundation expansion, data annotation, quality improvement, application enablement, management services and value release. It calls for multi-modal datasets spanning text, code, images, audio, video, point cloud and time-series data, with specific mandates for embodied intelligence, autonomous driving and world model development. The plan also designates data annotation as a strategic capability, building on seven pilot cities already designated for annotation innovation and planning additional test zones.
The policy marks a turning point for China's AI data economy. By establishing a token-based value system — the National Data Administration in March 2026 formally classified tokens as "ciyuan" (word elements), elevating them from a technical concept to a national strategic asset — the framework creates the infrastructure for data pricing, trading and settlement. The government is directing financial institutions, patient capital and industry funds to increase investment in dataset construction, while encouraging government agencies and state-owned enterprises to include data procurement in budget planning.
Data Labeling and the Token Economy Take Center Stage
A central pillar of the plan is the transformation of data annotation from a labor-intensive process to an AI-augmented one. The framework mandates a shift toward human-machine collaboration with expert participation, promoting model-assisted pre-labeling and automated quality verification. It calls for an expert certification mechanism to ensure domain-specific knowledge is embedded into training data for instruction fine-tuning and reinforcement learning.
On the value-capture side, the plan explicitly endorses token-based transactions as a new trading model, aiming to build a quantifiable and priceable data value system with tokens as the base unit. This aligns with parallel industry developments: on June 8, Shanghai-listed PATEO (02889.HK) announced a tripartite partnership with Xunce Technology and Saimo Technology to build TokenOS, an edge-side token infrastructure for in-vehicle AI scenarios, and a physical AI world model — the first full-stack value closed-loop alliance within the Nvidia ecosystem.
Data Rights Framework Opens Path to Assetization
The plan advances China's three-rights separation system for data — holding rights, usage rights and operational rights — and calls for clearer rules on using copyrighted content for model training. It encourages pilot programs in data asset registration, valuation, and innovative financing models including data collateral loans, equity contributions, asset-backed securities, data trusts and data insurance.
The framework also mandates a national dataset management service system with a physically distributed, logically centralized architecture, using privacy-preserving computation and blockchain to ensure traceability. A unified quality evaluation standard is planned, with a one-evaluation, nationally-recognized certification mechanism to reduce redundant testing.
This article is for informational purposes only and does not constitute investment advice.
