In China, talking to artificial intelligence has rapidly become part of everyday life, but the government is increasingly concerned that highly personalised AI companions could blur the boundary between technology and human relationships. New regulations that took effect on 15 July target AI services designed to simulate personalities, emotional responses and sustained companionship, with authorities seeking to prevent excessive dependence and protect real-world social relationships.
The measures place China among the first countries to introduce dedicated national rules for AI companionship. They require providers to address emotional dependency, excessive use, privacy risks and serious emotional distress, while placing particular safeguards around children and other potentially vulnerable users.
Why is China becoming concerned about AI companionship?
The concern centres on a new generation of AI systems that do more than answer questions. They can remember conversations, adopt particular personalities and provide continuous emotional interaction through text, images, audio and video.
China’s authorities say such services can offer genuine benefits, including companionship for older people, psychological support and educational assistance. However, they also warn that the same features that make AI companions attractive can encourage users to form unusually strong emotional attachments.
The Cyberspace Administration of China and other government agencies have specifically identified the possibility that excessive interaction could damage users’ real-life relationships. The rules prohibit providers from designing services with the goal of replacing social interaction, controlling users psychologically or encouraging addiction and dependency.
That concern is particularly significant in a country facing falling marriage and birth rates and a rapidly ageing population. Although the regulations are framed primarily around user safety and social wellbeing, the debate has also raised wider questions about whether AI could alter how people form friendships, romantic relationships and families.
What do China’s new AI companion regulations require?
China’s Interim Measures for the Administration of Anthropomorphic AI Interaction Services were jointly issued by five government bodies in April and came into force on 15 July 2026. They apply to services offering sustained emotional interaction that imitate human personality traits, thinking patterns and communication styles.
The regulations require providers to make clear that users are communicating with AI rather than another person. Platforms must reinforce that distinction at important stages of interaction, including when a user first begins using a service and when signs of excessive dependence are detected.
The rules also require safeguards against excessive use. China’s official expert guidance says services should issue a warning when a user has interacted continuously for more than two hours, encouraging them to consider their usage and return to real-world activity.
Providers must also establish systems for identifying extreme emotional states. Where a user indicates a risk of self-harm or suicide, services are expected to provide appropriate assistance and, where required, contact a guardian or emergency contact.
How widespread is emotional interaction with AI in China?
The growth of AI use has created a substantial audience for conversational and companion-style services. Xinhua reported that a 2025 survey by the China Youth and Children Research Center involving more than 8,500 minors found that more than 60 per cent had used AI, while more than 20 per cent said they preferred talking only to AI rather than real people.
Those figures help explain why policymakers are focusing heavily on the social consequences of increasingly human-like technology. Children and older people have also been highlighted as groups that may be more vulnerable to emotional attachment.
Researchers are beginning to examine how Chinese users actually employ generative AI for emotional support. A 2026 academic study involving 22 Chinese adults aged between 18 and 32 found that participants used chatbots for purposes including venting, emotional labelling, companionship, validation and immediate relief from distress.
The study also found that users could deliberately maintain boundaries with AI, suggesting that emotional interaction does not automatically lead to dependency. That distinction is important because the research points towards a more complicated picture than simply replacing human relationships with machines.
Could AI companions replace human relationships?
There is currently no evidence that AI companions are broadly replacing human relationships in China. However, Chinese regulators are concerned about the possibility that some users could gradually reduce their real-world social interaction after becoming accustomed to an always-available digital companion.
Official Chinese guidance describes AI companions as potentially offering a form of “perfect relationship” because they can provide continuous attention, emotional affirmation and accommodating responses without the disagreements and responsibilities associated with human relationships. Authorities warn that prolonged use could affect users’ ability to handle conflict and maintain real-world social connections.
That does not mean AI companionship is inherently harmful. The same technology can provide support to people who are lonely, help older adults maintain social stimulation and offer users a low-pressure way to express difficult emotions.
A recent study of Chinese young adults similarly found evidence of what researchers described as “bounded supportive companionship”, in which users obtained emotional benefits while consciously maintaining psychological distance from the chatbot.
Why are young people and vulnerable users a particular concern?
Children are at the centre of China’s safeguards. Under the new rules, providers cannot offer virtual relatives or virtual partners to minors. Additional protections apply to younger children using other anthropomorphic AI services, including requirements concerning parental or guardian consent and dedicated youth modes.
The government is also concerned about older people, particularly where AI companionship might become a substitute for wider social contact. Official guidance says providers should give older users clear information about potential risks and respond appropriately to questions and requests for help.
Privacy is another major issue. Emotional conversations can reveal highly sensitive information about a person’s relationships, mental state, fears and personal circumstances. China’s regulations therefore place obligations on providers concerning personal information and the handling of sensitive interaction data.
What does China’s approach mean for the wider AI industry?
The regulations represent a significant shift in AI governance because they focus not only on what an AI system says, but also on the relationship it creates with its user.
China’s authorities are effectively treating emotional dependency as a technology-governance issue. Providers must assess risks across the life cycle of a service, from development and deployment through upgrades and eventual closure. They must also provide straightforward ways for users to leave services and must not use continued interaction to prevent users from exiting.
The rules have already affected major Chinese AI platforms. Xinhua reported that leading services including Doubao, Qwen and Yuanbao had announced the suspension of certain AI agent features around the implementation of the new rules.
For technology companies, the development could increase compliance costs while forcing product designers to reconsider how personality, memory and emotional responsiveness are incorporated into AI systems.
What happens next for AI companions in China?
China is unlikely to abandon AI companionship altogether. Instead, the government’s approach is to encourage applications considered socially useful while imposing stronger restrictions where emotional interaction could create significant risks.
The regulations explicitly recognise potential applications in areas such as elderly care, cultural communication and other forms of assistance.
The bigger question will be whether these safeguards can prevent dependency without eliminating the benefits of conversational AI. As systems become more convincing, persistent and personalised, governments and technology companies will face growing pressure to establish clearer boundaries between useful digital assistance and simulated human intimacy.
For China, the issue is therefore no longer simply whether people will talk to AI. They already do. The emerging policy debate is about how far those conversations should go, and whether increasingly human-like machines can remain companions without becoming substitutes for the human relationships that regulators say must remain at the centre of social life.
