China to establish standards for embodied AI data
Regulator responds to industry demand with new data guidelines for AI training.
1Regulatory response to industry demand
China's National Data Administration announced plans to develop standards for embodied AI data just ten days after seven companies formally requested such regulations. The agency will guide local authorities and support companies investing in data resources. This swift response highlights China's proactive approach to AI governance compared to Europe's more reactive regulatory framework.
2Critical need for training data
The China Academy of Information and Communications Technology estimates embodied AI foundation models require approximately 10 million hours of real training data. Currently, only 100,000 to 1 million hours of high-quality data exist globally. This massive gap explains why China is prioritizing data infrastructure development, with robotics advancement now constrained by data availability rather than model capabilities.
3Expanding training infrastructure
China already operates over 70 embodied AI training grounds across more than half its provincial regions, with 46 additional facilities planned or under construction. Notably, 86% target industrial manufacturing applications. Three major clusters have emerged in the Yangtze River Delta, Beijing-Tianjin-Hebei region, and Pearl River Delta, creating concentrated hubs for AI development.
4European regulatory contrast
Europe's Data Act, effective since September 2025, focuses on governing access to machine-generated data without establishing data production programs. The promised data labs from the November 2025 Data Union Strategy remain unimplemented after ten months. European companies like NEURA Robotics are building private training facilities, with five operational by year-end across three continents.
5Upcoming EU design obligations
Starting September 2026, new EU regulations will prohibit connected products from using generated data to build competing products. These design obligations apply to all connected products entering the EU market. While creating market protections, these rules don't address Europe's lack of coordinated public data infrastructure comparable to China's training ground network.
6Industry participation in policymaking
The National Data Administration's September 10 meeting included research institutes, tech companies, and robotics organizations, reflecting China's collaborative approach to AI governance. This follows earlier industry consultations where companies specifically requested public data infrastructure and common standards for embodied AI development.
7Market readiness challenges
Despite rapid infrastructure development, only 23% of Chinese enterprises express satisfaction with currently available robots. This indicates that while China is building data collection capacity, converting these resources into commercially viable products remains challenging. Training grounds currently prioritize data accumulation over immediate commercial applications.
8Global AI development implications
China's standardized approach to embodied AI data could give it significant advantages in robotics development. By systematically addressing the data bottleneck through coordinated public-private infrastructure, China positions itself to lead in creating AI systems that effectively interact with physical environments, potentially reshaping global tech competition dynamics.
Original article: The Next Web — China to establish standards for embodied AI data