China’s National Data Administration is moving embodied-intelligence data standards higher on its policy agenda, while leaving the technical rulebook for later. Director Liu Liehong chaired a September 10 symposium on embodied-intelligence data standards, and the administration followed on September 13 with an in-due-course signal: it said it would promote standards, guide local data bureaus, and support corporate data investment.
The statement marks a meaningful regulatory cue for robotics and embodied-AI developers, but not an implementation package. Pandaily’s coverage of the National Data Administration notice reflected the same limited but important message. The NDA did not publish a new draft, open a formal consultation, define data formats, set evaluation metrics, or provide a deadline.
A standards signal, not a rulebook
The September 13 notice places embodied-intelligence data within the NDA’s broader standards work without converting the September 10 symposium into a binding framework. A meeting chaired by the national data regulator’s director carries policy weight, especially in a field where training data can shape how robots perceive, move, and interact with physical environments. Still, the agency’s wording points to future standards promotion rather than a completed standard.
That distinction is central for companies building embodied-AI systems. A standards direction can influence investment planning, internal data governance, and local government coordination, but it does not yet give developers a checklist for compliance. The NDA has not yet specified new data formats or evaluation metrics tied to the symposium.
The administration’s pledge to guide local data bureaus also remains broad. The public notice does not define the specific roles those agencies will take as the standards agenda develops.
The NDA also said it would support corporate data investment, reinforcing the view that embodied-intelligence data is becoming a policy priority. For the private sector, the signal is strategic rather than procedural.
Embodied intelligence depends heavily on data drawn from physical-world tasks, including perception, motion, manipulation, and environment interaction. That makes standards especially consequential. Poorly structured or inconsistently labeled data can limit model transfer across robots and scenarios, while common data practices can make development more efficient. The NDA’s move suggests that China’s data authorities want a more organized foundation for this segment of AI, even though the operational details remain pending.
Participants show a practical policy focus
The symposium brought together institutions and companies connected to China’s AI and robotics ecosystem. Participants included CASIA, the Beijing Institute for General Artificial Intelligence, JD.com, Lightwheel Intelligent, Deep Wisdom, GalaXea AI, Zibianliang Robot, and the Beijing Humanoid Robot Innovation Center.
That mix gives the discussion a practical orientation. The participant list spans research bodies, platform and technology companies, robotics developers, and a named humanoid-robot institution. It suggests that the NDA is engaging organizations with exposure to data production, model development, robot training, and potential deployment settings. At the same time, the list should not be read as an assessment of any participant’s products, datasets, benchmarks, or commercial readiness. The administration did not publish such evaluations.
For related EastFrontier context on the regulator’s data agenda, see the National Data Administration’s earlier dataset plan. The September symposium is specifically focused on data for embodied intelligence.
The participant list includes the Beijing Humanoid Robot Innovation Center. For separate EastFrontier context on the city’s infrastructure, see Beijing’s embodied-AI robot-training center. The September notice did not announce a new mandate for the center.
For local governments and companies, the participant list may help identify the types of stakeholders the regulator considers relevant to embodied-intelligence data work. But it does not create a formal club, approval channel, or preferred-provider system. The public message remains focused on standards promotion, local guidance, and support for corporate investment.
Existing milestones frame the next phase
The NDA’s signal comes against a backdrop of existing standards milestones. The named milestones include YD/T 6770-2026, which has been in effect since June 1; YD/T 6771-2026, which takes effect November 1; and GB/Z 218.1-2026. These identifiers show that China’s standards environment for related technologies is already developing, even as the NDA’s embodied-intelligence data work remains at a directional stage.
Those milestones should not be conflated with a new NDA rule issued after the symposium. The September 13 statement did not publish a fresh embodied-intelligence data standard, attach a draft to the meeting, or provide a timetable for release. It also did not specify how existing standards would be used in future embodied-intelligence data governance.
The result is a two-layer landscape. One layer consists of standards milestones with identified designations and dates. The other is the NDA’s in-due-course commitment to promote embodied-intelligence data standards while guiding local data bureaus and supporting corporate data investment. The first layer gives the market concrete reference points. The second indicates administrative direction, but not yet an implementation route.
For robotics developers, that means the standards conversation is becoming harder to ignore, but direct technical requirements remain unpublished. For data teams, it signals that embodied-intelligence data will receive greater regulatory attention, while the detailed requirements remain unpublished. For local data bureaus, it means a national guidance role is being signaled before detailed responsibilities are made public.
The measured reading is therefore the strongest one. China’s National Data Administration has elevated embodied-intelligence data standards through a director-chaired symposium and a follow-up notice. It has named a policy direction and identified relevant participants. But it has not released a new draft, formal consultation, data format, evaluation metric, or deadline.