China’s latest online-minor-protection proposal is not a blanket prohibition on artificial intelligence. It is a broader draft governing how network services interact with young users, with a specific restriction that places virtual intimate relationships inside a wider framework for identification, platform design, algorithms, and enforcement.
The Cyberspace Administration of China published the consultation draft on September 18 and opened it for public comments through October 17. Its Chinese title is 《国务院关于保障未成年人健康安全使用网络的规定（征求意见稿）》. A cautious working English rendering is “Regulations on Safeguarding Minors’ Healthy and Safe Use of the Internet (Draft for Soliciting Comments).” It is a draft for comment, not an enacted law, and the proposed effective date remains blank.
South China Morning Post reported that the proposal would bar platforms from offering minors virtual intimacy services. The official text is more specific: it would prohibit providers from offering virtual relatives, virtual partners, and other virtual intimate-relationship services to minors. That language gives the proposal a much more targeted scope than a headline suggesting a total ban on all AI companionship.
The Draft Treats Virtual Intimacy as One Part of Online Child Protection
Article 4 of the draft would also restrict stranger-focused online social networking for minors and services that induce addiction or harm, or may seriously affect, minors’ physical or mental health. It would prohibit livestream publishing services for users under 16.
The proposal includes an important exception within the stranger-social-networking provision. Minors aged 16 or older who rely primarily on their own labor income for their livelihood would not fall under that particular restriction. The exception is a reminder that the draft differentiates among age, circumstances, and types of service rather than treating every young internet user identically.
For AI companies, the intimacy clause is especially notable because it focuses on a relationship format, not merely on whether a system uses generative technology. The official wording names virtual relatives and partners, then adds other virtual intimate-relationship services. The text names service categories and does not identify a particular company, chatbot, or model.
No named platform, company, or model appears in the verified draft materials. It would be inaccurate to suggest that the text targets a specific domestic or foreign provider. The rules describe categories of service and provider obligations.
The measure is distinct from the separate AI human-like-interaction rules that took effect in July. EastFrontier covered how China’s AI companion rules took effect as users posted emotional goodbyes online. The September proposal should not be treated as a relaunch of those earlier measures. Its focus is the safe and healthy use of the internet by minors, and its legal status is still consultative.
Minor Mode Would Reach Some AI and Social Services
Article 5 would require specified services for users under 16 to operate through a “minor mode.” The text covers social-networking services and AI services that may affect minors’ cognition. The draft defines minor mode as an arrangement intended to identify harmful or potentially harmful information, protect personal-information rights, prevent internet addiction, and help guardians perform their responsibilities.
This is a design and governance requirement as much as a content rule. A provider would have to build mechanisms capable of recognizing that a user is a minor and applying the correct product settings. Article 6 says providers should use lawful methods to improve the accuracy of minor-user identification while protecting personal information. The text lists options including identity documents, public network-identity authentication, non-stored one-time facial verification, and behavioral-feature recognition.
Those examples should not be read as a universal instruction to use each method for every user. The draft emphasizes comprehensive and lawful approaches to identification. Its practical impact would depend on final wording, implementing guidance, platform architecture, and the balance regulators strike between accuracy and privacy.
The proposal also reaches the algorithms used after a user has been identified. Article 8 would bar providers from using algorithmic models designed to induce minors’ emotional dependence, internet addiction, or excessive consumption.
The Consultation Stage Leaves Questions Open
The draft proposes duties beyond service providers themselves. Operators of smart-terminal products and app-distribution platforms would have responsibilities related to minor mode, age suitability, review, and enforcement. Providers of minor mode would need to file settings and functions with provincial cyberspace authorities, while very large or highly influential platforms would file through provincial authorities with the national CAC.
The draft also contains proposed penalties. For violations of several core provisions, it sets out correction orders, warnings, confiscation of illegal income, and provider fines ranging from RMB100,000 to RMB1 million in specified circumstances. Responsible personnel could face fines from RMB10,000 to RMB100,000. These are proposed enforcement ranges, not penalties that are already being imposed under a final regulation.
The consultation permits public comments through October 17 before a final text is issued.
The new proposal is distinct from the separate AI human-like-interaction measure that took effect in July, and the two regimes should not be blended. Readers seeking historical context can compare the separate minor companion restriction with EastFrontier’s account of when the July companion rules took effect.
For now, the defensible conclusion is straightforward. China has proposed new minor-online-protection rules that would restrict virtual intimate-relationship services and impose broader obligations on platforms. The proposal is detailed, but it is not final. Its eventual force will depend on the text that emerges after October 17.