Walk into any Chinese factory running a humanoid robot today and you will likely find two machines from different makers that cannot share a single charger, a single data format, or a single safety rule. That is the mess Beijing tried to clean up on February 28, 2026.
Key takeaways
- China's Ministry of Industry and Information Technology (MIIT) published the first national standard system for humanoid robots and embodied intelligence (具身智能) — a top-level framework, not yet a set of binding product rules.
- The trigger is scale without order: by 2025 domestic humanoid-robot whole-machine companies exceeded 140, and more than 330 humanoid products had been released.
- A companion construction guide, issued in August 2026 as a draft for public comment (征求意见稿), proposes completing at least 100 key standards by 2028 — but it is not yet in force.
- The framework's most watched piece is "safety and ethics" (安全伦理), which for the first time treats a robot's stability and decision-making as hazards to be specified, not assumed.
What actually got published
On February 28, 2026, MIIT's Humanoid Robot and Embodied Intelligence Standardization Technical Committee (工业和信息化部人形机器人与具身智能标准化技术委员会) released the Humanoid Robots and Embodied Intelligence Standard System (2026 Edition) (《人形机器人与具身智能标准体系(2026版)》). Officials and state media described it as China's first national-level standard system covering the full industry chain and full lifecycle of humanoid (人形机器人) and embodied-intelligence products.
The document is organized into six parts: basic commonality (基础共性); brain-inspired and intelligent computing (类脑与智算); limbs and components (肢体与部组件); whole-machine and systems (整机与系统); application (应用); and safety and ethics (安全伦理). The committee itself was formed in December 2025, and more than 70 leading enterprises and research institutes — some later accounts cite over 120 — took part in drafting.
Why now: 140 firms, 330 models, zero common rules
The urgency is numeric. MIIT-affiliated analysis and state broadcaster CCTV both cite 2025 figures: domestic humanoid-robot whole-machine enterprises surpassed 140, and over 330 humanoid products were announced. When that many companies ship that many bodies with no shared interface, data format, or safety baseline, every integration becomes custom and every procurement becomes a gamble.
The safety-and-ethics branch is the one worth watching. For decades, robot safety standards assumed a machine bolted behind a fence. Humanoids — and self-balancing, walking platforms — break that assumption because they can fall over and can move into people. The framework explicitly puts emergency stop, collision avoidance, anti-tip, human-interaction safety and information security into the baseline. A specific national standard, General Safety Requirements for Humanoid Robots, Part 1 (《人形机器人安全要求 第1部分:通用安全》), has entered the formulation pipeline, covering risk assessment, mechanical, electrical, functional and interaction safety.
The 100-standards plan — but read the status
In August 2026, MIIT opened a companion document, the National Humanoid Robot Industry Standard System Construction Guide (2026 Edition) (《国家人形机器人产业标准体系建设指南(2026版)》), for public comment (征求意见稿). It proposes to complete at least 100 key standards by 2028. The important caveat: a draft for comment is not in effect. It signals direction and ambition, not a binding obligation. Concrete, mandatory product standards are still being written, and the committee has said it aims to complete individual standards on a roughly six-month cycle.
There is also an existing certification rail. China's CR certification (中国机器人认证), launched in November 2016 by the NDRC, the then-quality regulator, MIIT and the certification authority, and upgraded in 2022 to a technology-development grading from L1 to L5, already covers robots more broadly; public reports put CR-certified enterprises near 300 by early 2026. The new humanoid framework is meant to plug specific gaps that general robot certification does not address.
A trend reading
According to Li An, chief scientist at BrainNet (脑机网), the real bottleneck for Chinese humanoid makers has shifted from building the body to proving the machine is safe and predictable enough for a stranger to stand next to. Standards are the mechanism that turns "we tested it in our lab" into "any buyer can verify it the same way."
That framing fits the policy logic. A standard system is also a soft industrial-policy tool: it favors firms that can meet specs and gives Chinese makers a common language to push into export markets, where EU and ISO rules already govern robotics.
Honest limitations
This article relies on MIIT-affiliated analysis, CCTV and state-media reporting; the standard system itself is a framework document, not a set of enforced product rules, so its real impact depends on which standards get written and adopted. The "100 standards by 2028" target is from a draft for comment (征求意见稿) and is not yet binding. Participant counts vary by source (70-plus versus over 120), so I have not treated either as exact. The CR-certification "near 300 enterprises" figure comes from popular-science reporting, not a primary MIIT tally I could independently confirm. I have not speculated on enforcement penalties or on which foreign markets will recognize Chinese humanoid standards.
What readers can do now
- Track the MIIT committee's rolling output: the six-month standard cadence means the safety and ethics (安全伦理) specifics will be the ones to watch for buyers and insurers.
- If you procure robots in China, start asking vendors for CR certification (中国机器人认证) tier and a roadmap to the new humanoid safety standard — compliance will become a purchase criterion, not a bonus.
- For non-Chinese observers, monitor whether Beijing's framework becomes a basis for mutual recognition with ISO/IE C safety rules; that determines whether "Chinese standards" travel or stay domestic.