China plans safety rules for AI shopping assistants, by how risky they are

China wants more AI helpers in stores and deliveries, and it plans to regulate them by how risky they are. On 19 October 2026, China's Ministry of Commerce posted an 18-step plan signed by nine government agencies. The aim is to bring more technology into commerce.

What the plan says

One step asks for new businesses built around commercial AI agents. These are programs that act on a task for you. The plan names three examples: a shopping guide that advises customers, a tool that runs store operations, and one that organizes the movement of goods. The list is not closed.

Another step covers safety. It says these agents should be governed in tiers, depending on how they are used. In May, China's Cyberspace Administration set up the same tiered approach for all agents. Sensitive sectors face registration, testing and recall of defective products. Low-risk areas can rely on self-assessment, meaning the company checks itself.

What is still unknown

The plan does not say which tier shopping, operations or logistics agents fall under. It only says the boundaries should be clarified. For comparison, the European Union's AI Act does not treat a shopping guide agent as high-risk.

Companies that sell or run AI agents in Chinese commerce now know the direction of the rules. They do not yet know the cost. The same ministry also proposed export controls on intelligent agents on 16 October.