Chinese regulator drafts rules to shield teenagers from "intimate" AI companions

China’s top internet regulator has proposed new rules aimed at protecting minors from potentially harmful interactions with artificial intelligence, including a ban on platforms offering “virtual relatives or companions” to users under 18.

The draft regulations, published by the Cyberspace Administration of China on September 18, would prohibit online platforms from providing “virtual intimacy services” to minors or services that could encourage addiction or otherwise seriously affect their physical and mental health, South China Morning Post reports.

The proposed rules come as AI-powered companions and humanlike chatbots become increasingly popular, raising concerns about emotional dependence among younger users.

The regulations, titled “Ensuring minors’ safe and healthy use of the internet”, would also restrict social networking between strangers. An exception would apply to people aged over 16 who earn their own wages.

Platforms operating online games, social networks and AI services that “could affect the cognition of” users under 16 would be required to activate a dedicated minors’ mode.

Operators would also have to improve their ability to identify underage users and prevent minors from being targeted by algorithms that “induce emotional dependence, addiction or excessive spending”.

Companies that violate the rules could face fines of up to 10 times the income earned by the platform if that income exceeded 1 million yuan ($149,000). Authorities could also confiscate all of the income involved.

The proposals are open for public comment until October 17.

Children and teenagers aged six to 19 accounted for 18 per cent of China’s internet users at the end of 2025, representing about 112.5 million people, according to the state-run China Internet Network Information Centre.

The proposed restrictions also raise questions about how authorities and technology companies will determine what constitutes a prohibited virtual intimate relationship.

“The key question in defining a ‘virtual intimate relationship’ is whether a service is designed to simulate or substitute for real family or partner relationships and build sustained emotional bonds through character settings and emotional guidance,” said He Qianggao, a lawyer and director at Geason Law Offices.

He said ordinary AI conversations and functional role-playing would generally not fall under the rules. However, services that encouraged emotional dependence or effectively replaced real-world relationships could face scrutiny even if they were not explicitly marketed as AI companions.

For AI companies, the compliance challenge will therefore extend beyond verifying users’ ages to determining which services minors can access once they have been identified.

Platforms could consequently face higher costs related to product design, technical infrastructure and compliance reviews, He said.

The latest proposals come two months after China introduced its first dedicated rules governing humanlike AI interaction. Those rules, which took effect in July, targeted AI systems that excessively cater to users or encourage emotional dependence in ways that could damage real-world relationships.

To comply, ByteDance’s Doubao and Alibaba’s Qwen disabled customised agent functions that had allowed users to create AI characters with fixed names, personalities and speaking styles.

By Aghakazim Guliyev