China is moving to establish stronger governance around agentic AI, proposing new draft regulations that would require humans to retain final decision-making authority over autonomous AI systems. The policy, published by China’s Cyberspace Administration, signals Beijing’s growing focus on balancing AI innovation with security, oversight, and ethical controls.
The proposed framework highlights China’s ambition to accelerate AI agent development while introducing stricter guardrails for how autonomous systems operate. The draft encourages the creation of specialised datasets, security standards, and ethical frameworks designed to support the safe deployment of AI agents across industries. It also calls for mandatory behavioural standards in sensitive sectors such as healthcare, transportation, media, and public safety.
A key principle within the draft is maintaining humans “in the loop.” Developers would need to define clear boundaries for agentic decision-making, including which actions require direct user approval and which can operate autonomously. The policy explicitly states that users must retain the right to know how decisions are made and maintain final authority over actions that AI systems take.
The regulations also outline a broad vision for how China could use AI agents across its economy. Proposed applications include analysing medical images, evaluating employee performance, supporting disaster relief operations, managing bidding and procurement processes, and even assisting with educational tasks such as homework grading.
ソース：
