The Pulse

China’s Data Regulator Plans Embodied-AI Standards

A January committee-meeting report and subsequent standards actions show China is building data-quality and collection rules for embodied intelligence.

AI.info Team · Written with AI · fact-checked · 2026-09-13

China’s data push meets an expanding robot rulebook

China’s National Data Administration has identified embodied intelligence as a frontier area for data-standard work, placing the subject alongside high-quality datasets, data circulation and governance. The commitment was set out in a January 7 report on a January 5 meeting of the National Data Standardization Technical Committee.

The meeting reviewed and approved the committee’s 2026 work priorities. It called for the development of at least 80 national standards and technical documents in the data field during 2026, and for the release of at least 30 priority standards. It also said the committee should arrange data-standard work in frontier areas including intelligent-agent interconnection and embodied intelligence.

According to the report, the meeting required the committee to “研制不少于80项国家标准和技术文件，发布不少于30项重点标准” (develop no fewer than 80 national standards and technical documents and release no fewer than 30 priority standards). Liu Liehong, Party secretary and director of the National Data Administration and chairman of the committee, attended and addressed the meeting, which was chaired by Xia Bing, deputy director of the National Data Administration and vice-chairman of the committee.

The National Data Administration’s report does not set out a complete technical specification for embodied-AI data. Instead, it places embodied intelligence among the fields in which the committee is expected to organize future data-standard work. Liu had set out the reasoning in March, speaking at the China Development Forum:

“Artificial intelligence is undergoing a paradigm shift from digital simulation to physical interaction, which has ignited entirely new demand for embodied-intelligence data. … We will actively promote the collection and development of high-value real-machine data and consolidate the data foundation for the evolution of embodied intelligence.”

Liu Liehong, director of the National Data Administration, at the China Development Forum on 23 March 2026 (translated from Chinese)

Formal documents are already being published

Separate national standardization work has produced a guidance document on real-world data for embodied intelligence. The State Administration for Market Regulation and the Standardization Administration of China approved Artificial Intelligence — Specifications for Embodied Intelligence Data Quality — Part 1: Real Data, numbered GB/Z 218.1-2026, as a national standardization guidance document.

The national standards full-text platform lists August 27, 2026 as its publication date, and the document takes effect on publication.

The national standards database identifies GB/Z 218.1-2026 as a published guidance document. It lists the China Electronics Standardization Institute, Beijing Humanoid Robot Innovation Center, Beijing University of Technology, Huawei Cloud and a range of robotics and AI companies among the principal drafting organizations. The database also lists related embodied-intelligence work on data-generation technology, training facilities, system specifications, evaluation and simulation platforms.

MIIT has separate projects for robot data collection

China’s Ministry of Industry and Information Technology has placed data collection directly in its 2026 industry-standard work. Its third batch of industry-standard revision projects includes a 12-month project titled Humanoid Robot Data Collection Technical Requirements, assigned to the ministry’s Humanoid Robot and Embodied Intelligence Standardization Technical Committee.

The listed drafting organizations include Humanoid Robot (Shanghai), Leju Intelligent, AgiBot, PaXini Perception Technology, Shanghai Artificial Intelligence Research Institute, Huawei, China Mobile and Pujiang National Laboratory. The same schedule includes a second 12-month project, Humanoid Robot Data Collection Dedicated Whole-Machine Technical Specifications, as well as projects on general technical requirements for humanoid robots and interfaces for dexterous hands and gripping actuators.

That division of work separates broad data-standard priorities from engineering requirements for gathering robot data. The National Data Administration’s January committee report focuses on standards supply and frontier data fields; MIIT’s projects focus on collection requirements and equipment associated with humanoid robots.

High-quality datasets are a stated policy focus

In June, the National Data Administration issued an implementation plan for building industry-specific high-quality datasets. The plan names embodied intelligence among the innovation fields for accelerated dataset construction. It calls for real-machine interaction datasets involving physical interaction, environmental perception and motion control, and encourages the use of simulation and synthetic techniques to expand data supply for embodied intelligence.

The June plan also addresses dataset management across collection, cleaning, processing, labeling, quality inspection, evaluation, iteration and auditing. It calls for work on standards covering dataset formats, types, labeling and quality assessment, while setting out measures on security, privacy-protection computing and data governance.

Together, the January committee report, the published real-data quality guidance document, MIIT’s collection projects and the June dataset plan show several agencies working on different portions of the embodied-AI data chain. On September 10, Liu chaired a symposium on embodied intelligence with research institutes and robotics companies; afterwards the National Data Administration said it would “适时推动具身智能数据标准建设” — advance embodied-intelligence data standards at the appropriate time.