China Publishes Embodied Intelligence Ethics Governance Report, Outlining Six Risk Categories and Five Risk Levels

leaderobot.com reported Oct. 17 that the National AI Standardization General Group published the Research Report on Ethical Governance of Embodied Intelligence on Oct. 13, and that CCTV’s Zhengdian Caijing covered it the following day. According to the reports, this is China’s first official report focused on the ethics of embodied intelligence. The report outlines six categories of ethical risk, five levels from R1 to R5, and six governance principles.

The release of an official report on embodied intelligence in China signals that responsibility boundaries and safety requirements are being raised as issues ahead of wider deployment. For robots operating in the same spaces as people, recordkeeping and the handling of incidents may become conditions for adoption alongside performance.

The new development in this report is that the safety debate around embodied intelligence has moved from general discussion to institutional design. By setting out six categories of risk and levels from R1 to R5, the report seeks to define what harms should be anticipated and how they should be managed before debating whether the technology is good or bad. Its treatment of mistaken identity, overestimation of capabilities and unclear responsibility as separate issues suggests recognition that privacy and bias, central to traditional AI ethics, are not enough. A related article by this newspaper, [leaderobot.com reports 26 quadruped robots operating in a chemical industrial park, with certification from the China Academy of Information and Communications Technology](/news/leaderobot-com-kakou-enku-deno-26-dai-yotsuashi-robotto-kado-23f9f0), showed that real-world industrial deployment is already underway. The ethical governance report can be read as an extension of that trend, setting rules with real-world applications such as industrial inspections and in-home assistance in mind. But compared with the certification and on-site operations covered in the earlier article, this report offers only upstream classifications and principles for safety standards; details on implementation are not yet clear. See the original source for details. For the industry, the implications may extend beyond competition over hardware performance: responsibility tracking, log management, handling of video and audio, and incident analysis could become competitive factors. Since the report calls for responsibility traceability, a key question will be where manufacturers, operators and oversight bodies keep records. Robots used in homes and public spaces are also more likely than factory-bound machines to face demands for explainability and emergency-stop design. As the boundaries between R1 and R5 have not been made public, it remains unclear which applications will fall under which level of oversight. The scope of application, the weight of obligations and the link to certification will be points to watch.

For Japanese companies, requirements such as emergency stops, log management, explainability and certification may become part of competitiveness, beyond the robots themselves. However, this report alone does not clarify specific certification systems or differences in how the rules apply in China and Japan; the original source should be consulted for details.