China’s data regulator plans standards for embodied artificial intelligence

China’s data regulator plans standards for embodied artificial intelligence

China's National Data Administration is drafting standards for embodied artificial intelligence, the branch of AI that gives machines the ability to perceive and physically interact with their environment. The agency also plans to guide local governments on data governance as the country races to turn humanoid robots from lab curiosities into commercial products.

Embodied AI is the term for intelligence that lives inside a physical body, whether that's a warehouse robot, a surgical arm, or a full humanoid. Training these systems requires enormous, high-quality datasets capturing real-world physics, spatial awareness, and human movement.

The NDA's initiative centers on two priorities. First, it will develop formal standards governing how data for embodied AI is collected, labeled, stored, and shared. Second, it will strengthen planning guidance for provincial and municipal authorities, while offering support to companies looking to boost their investments in data resources.

China's Ministry of Industry and Information Technology had already formed a dedicated standardization committee for humanoid robotics. That committee published the country's first national standards framework on February 28, 2026, laying out guidelines across six critical areas including data lifecycle management.

An industry benchmarking standard for embodied AI took effect on June 1, 2026, giving companies a concrete yardstick against which to measure their systems. The NDA's latest move layers data-specific rules on top of that existing structure.

Large language models like the ones powering chatbots can be trained on oceans of text scraped from the internet. Embodied AI doesn't have that luxury. Teaching a robot to fold laundry or navigate a factory floor requires carefully curated datasets that map physical interactions, not just words on a screen.

Companies in the space are reportedly building over 1 million hours of proprietary or open datasets specifically for embodied AI systems.

By standardizing formats, labeling conventions, and sharing protocols, the regulator is trying to make it easier for companies to pool resources and avoid duplicating effort. Uniform data standards also lower the barrier for smaller firms that can't afford to build massive proprietary datasets from scratch.

China's National Development and Reform Commission has been working in parallel. Measures released by the NDRC in late August 2026 include the establishment of dedicated training grounds and pilot bases where embodied AI systems can be deployed and tested in controlled real-world settings.

Beijing has made humanoid robots a stated industrial priority, and the timeline is aggressive. The government is targeting notable advancements in embodied AI commercialization by the end of 2026, which means the regulatory scaffolding being assembled right now is meant to accelerate deployment, not slow it down.

For manufacturers and technology firms, the implications are straightforward. Companies that align early with the emerging standards will find it easier to access government-backed testing facilities, participate in data-sharing ecosystems, and qualify for pilot programs.