China is moving to place tighter restrictions on artificial intelligence services designed to provide emotional companionship to young people, with proposed rules targeting virtual relationships between minors and AI characters. The measures would prevent online platforms from offering “virtual relatives or companions” to users under 18 and introduce additional safeguards around children’s use of AI services.
The proposals, reported by the South China Morning Post on 21 September, come shortly after China introduced its first dedicated rules for humanlike AI interaction. The latest move highlights growing regulatory attention on AI systems designed to form emotional connections with users.
China sets limits on AI relationships with minors
The proposed restrictions would prohibit AI platforms from providing virtual family members, romantic partners or other forms of virtual intimate relationships to people under 18. The rules would also target services that encourage excessive use, emotional dependence or behaviour that could seriously affect a young person’s physical or mental well-being.
China’s existing rules for anthropomorphic AI services, which came into force on 15 July, already prohibit providers from offering virtual relatives or virtual partners to minors. The regulations also require providers to establish a dedicated minors’ mode, introduce time-use reminders and take measures to identify underage users while protecting their personal information.
The regulations distinguish between emotionally oriented AI companions and ordinary functional AI tools. Services such as customer support, knowledge-based question answering, workplace assistants, education and scientific research are outside the rules when they do not involve continuous emotional interaction. This means the focus is on AI products deliberately designed around ongoing personal or emotional relationships rather than AI use by young people in general.
The rules also require providers to intervene when users show signs of excessive dependence or addiction. AI services must remind users that they are interacting with artificial intelligence, not a real person. In contrast, users who spend more than two continuous hours using an anthropomorphic service must receive a reminder about their usage time.
AI companies face new compliance challenges
For AI companies, the restrictions could create significant technical and operational requirements, particularly around determining a user’s age. Providers are expected to identify minors while complying with privacy requirements and then switch recognised underage users to appropriate protections. Parents and guardians can also get tools to receive safety alerts, review usage information, block particular characters, and limit spending.
The rules also address how AI systems behave towards users. Providers are prohibited from designing services around emotional manipulation, dependence or addiction, and must take steps when interactions could damage users’ real-world relationships. They must also respond to certain safety risks, including situations involving extreme emotions or threats to life and health.
Data protection is another part of the framework. Providers must protect interaction data through measures such as encryption and access controls, while restrictions apply to sharing user interaction data with third parties. Providers generally cannot use sensitive personal information in interactions to train models without a legal basis or separate consent. Additional safeguards apply to the personal information of children under 14.
The latest proposals therefore add to a regulatory framework that is becoming increasingly specific about how AI companions interact with vulnerable users. China’s December 2025 draft rules had already proposed parental consent for emotional companionship services provided to minors, age-identification systems, parental controls and mechanisms for limiting usage and spending.
A wider debate over AI companions and children
The development comes as AI companions become more sophisticated and increasingly capable of maintaining long-running conversations and personalised interactions. Unlike conventional search or productivity tools, these services can be designed to simulate personality, emotional support, and social relationships, raising new questions about children’s use of the technology.
China’s regulators have explicitly identified emotional dependence as an area requiring intervention. The country’s Cyberspace Administration said anthropomorphic AI can provide benefits such as emotional support and companionship, but also poses risks to minors’ well-being, information security, personal health, and ethical standards.
The regulatory approach is not limited to banning particular types of AI characters. China’s framework also seeks to establish clearer boundaries between humans and machines. Providers must make it clear when users are communicating with AI and must provide practical safeguards against excessive dependence. The rules also require services to offer straightforward ways for users to leave an emotional interaction.
The issue is likely to remain relevant internationally as governments, technology companies, parents and schools consider how children should use increasingly humanlike AI systems. China’s measures provide one example of a regulatory model that treats emotionally dependent relationships with AI as a distinct safety issue, while still allowing functional AI applications such as education and productivity tools.
China’s current framework is already in force, while newer proposals reported on 21 September suggest further restrictions on intimate AI services for teenagers may follow. The developments underline how quickly governments are moving from broad AI safety principles towards specific rules governing how AI systems can interact with young users.
