On 1 May 2025 the Hangzhou ordinance promoting the development of the embodied AI robot industry took effect. It is an early piece of local legislation treating embodied AI robots as an independent object, and a head start by a local government in legislating for a specific AI segment.
From product to actor: what the ordinance changes
Traditional regulation treats a robot as an industrial product, governed by product quality and safety standards. This ordinance treats an embodied AI robot as a system with perception, decision-making and acting capability, and builds a regime of coding management and full-life-cycle traceability around it.
The object of regulation thus expands from a device to an actor that runs continuously, iterates and learns. When the same robot updates its model and strategy in use, factory inspection alone can no longer cover all of its risks.
Promotion and regulation in the same text
The ordinance is titled a promotion measure, but its content also includes regulation. This “promotion-type legislation” is worth noting: it first gives the industry legal certainty — companies know which products and scenarios are supported — and then embeds safety, traceability and data requirements within it.
For Hangzhou, this path rests on local industrial density. In our note on Hangzhou’s first unfair competition case involving commercial AI use it is visible that when technology and commercial scenarios intersect densely, local courts and legislators move in step. See also our note on China’s rules for intelligent information services.
What coding management and traceability mean
Coding management means every device will carry an identifiable identity, while traceability requires records of the key stages from production and sale to operation and maintenance. The direct implication for companies is a heavier data retention duty: who is using it, in what scenario, and what anomalies have occurred may all become mandatory entries in the traceability chain.
These requirements follow a logic similar to the standards already built for autonomous driving and intelligent connected vehicles. See our note on the World Intelligent Connected Vehicle Conference.
Where companies can start
First, map products to scenarios, clarifying which models and scenarios fall within the ordinance’s scope.
Second, build a retention mechanism for operational data in advance, so that traceability requirements can be met rather than retrofitted afterwards.
Third, put ethics and safety assessment up front, completing it at the design and factory stages rather than after an incident. See our notes on the AI application ethics and security guideline and on the patent examination guidelines on AI.
The significance of local legislation is that it often gives an executable answer before a national standard does. The earlier an embodied AI company brings local rules into product design, the less rework it faces when expanding across jurisdictions.
This article is based on the Hangzhou ordinance promoting the development of the embodied AI robot industry and public information and is for reference only; it does not constitute legal advice.