Rapid industry growth fuels calls for clearer AI rules.
Chinese industry players are urging faster rulemaking for embodied intelligence, as the technology moves from demonstrations into real-world deployment, while regulators warn of bubble-like risks from crowded investment and excess capacity.
The recommendations appear in the 2026 China Embodied Intelligence Industry Report, published Thursday by the China Mechatronics Technology Application Association and Shanghai International Exhibition Group. The report comes as more than 150 humanoid robot companies have flooded the market and robot-body manufacturing capacity already exceeds demand.
The association represents more than 1,100 members, including companies, research institutes and universities. The exhibition group, meanwhile, is China's first state-owned international trade show organizer in Shanghai, according to its website.
Embodied intelligence — AI systems that can sense, decide and act through physical machines — has become a priority for Chinese policymakers. While China holds an advantage in hardware supply chains, manufacturing capacity and industrial use cases, the report says the next phase of development will depend on standards, safety rules, data-governance requirements and liability frameworks that support reliable deployment rather than short-lived demonstrations.
Among these issues, safety, ethics and data governance are becoming central because embodied-intelligence robots operate in shared physical spaces and rely on cameras, microphones, lidar and other sensors. Large-scale deployment raises risks involving physical injury, privacy intrusion, trade-secret exposure, algorithmic bias and employment disruption, the report says.
Liability is another unresolved issue. The report calls for a liability framework that better reflects the autonomy of AI-enabled robots, noting that existing product-liability rules face major challenges when applied to autonomous decision-making systems because responsibility could fall on the manufacturer, software provider, operator or user.
National and local efforts
Rulemaking efforts are underway at both the national and local levels.
In February, China released the 2026 humanoid robot and embodied-intelligence standards system, following the December 2025 establishment of a sector-specific standardization technical committee under the Ministry of Industry and Information Technology.
Rather than a set of mandatory rules, the framework provides a roadmap for developing future standards across areas including basic terminology, brain-like computing, limbs and components, and safety and ethics. What remains missing are more detailed rules on scenario-specific safety requirements, common data and component standards, testing and certification, data handling, liability, and market entry and exit.
Locally, 31 provinces and cities had issued embodied-intelligence policies by June. Beijing is focusing on research, Shanghai on industrial deployment, Shenzhen on integration and exports, Hangzhou on legislation and Anhui on large-scale funding.
Hangzhou offers one local example of this regulatory shift. A municipal regulation on embodied-intelligence robots took effect on May 1, covering standards, data and personal-information protection, ethics and safety management, while calling for sandbox-style supervision.
The report warns, however, that competition among local governments could lead to duplicative investment and low-end capacity.
Fast and uneven growth
China’s embodied-intelligence market is forecast to reach 1.09 trillion yuan ($161 billion) this year, up from about 213.3 billion yuan in 2018. Chinese companies accounted for about 74 percent of global humanoid robot shipments in 2025, when global shipments rose to about 18,000 units from about 3,000 a year earlier, the report says.
But the National Development and Reform Commission has cautioned against bubble-like risks, including highly similar products rushing to market and pressure on core research and development.
The report says many companies remain focused on basic walking functions and demonstration products, while areas such as embodied large models, tactile sensors, simulation platforms, high-quality training data and reliability receive less attention. Companies lacking core technologies or clear use cases may be forced out in the next three to five years as part of industry consolidation.
More broadly, China still lags in embodied large-model algorithms, high-end sensors, training chips and long-term component stability. The report recommends guiding capital away from short-term hype and toward long-term industrial value, while improving market-entry and exit mechanisms to support industry consolidation.
Industrial deployment is expected to scale first, before commercial-service and home-use scenarios. To boost adoption, the report urges government agencies to open public-sector application scenarios, including public-facility inspection, elderly-care and disability-assistance services, and public-service halls.
Source: MLex
