China’s new rules governing artificial intelligence companions take effect Wednesday, pushing the country’s largest tech companies to disable customizable chatbot personas as Beijing moves to rein in an industry that has grown to tens of millions of users.
The regulations, published in April by the Cyberspace Administration of China, require AI companion services to regularly remind users they are communicating with a machine, mandate interventions when users appear emotionally distressed, prohibit children under 14 from using the services, and require an easy way to exit. A draft version was softened before adoption, but the core message is clear: the Chinese state will not allow emotional bonds with machines to go unregulated.
ByteDance’s Doubao, Alibaba’s Qwen chatbot, and Tencent have all announced they will disable customizable persona features. Tencent had already removed similar functionality last week. More than 3,500 noncompliant AI products have been removed, according to the CAC.
An Industry Built on Loneliness
China’s AI companion market is enormous. Talkie, the most popular dedicated service, had 23.5 million monthly users as of December 2025. Several competitors count users in the low millions. Apps like Xingye (“Star Field”) and Zhumengdao (“Dream Island”) are designed specifically for emotional and romantic interaction, relationships built entirely on language models wrapped in fictional personas.
The global AI companion market skews toward men seeking AI girlfriends. China is different. The market there skews female, driven by AI boyfriends. The growth is tied to “2D culture”, Japanese anime fandom, which shifted from male to female majority over the last decade.
“A visiting Western academic told me that he recently met a male university student who said that because he knew he lacked the money and status to get a date, he had three AI girlfriends instead,” James Palmer wrote in Foreign Policy.
The demographic backdrop makes the government’s motivation clear. China’s birth rate is at a historic low. The state has spent years trying to encourage marriage and childbearing. It has little enthusiasm for technologies that teach people to invest emotionally in machines instead of each other.
Why Beijing Acts Fast Online
China already has the infrastructure in place to regulate AI, content filters, algorithm registrations, data localization requirements. The digital economy is one of the easiest domains for the Chinese state to control, unlike the offline economy where vested interests and corruption slow enforcement.
The new rules supplement existing AI regulations covering algorithm recommendation, deep synthesis, and generative AI. Companies are preemptively complying to demonstrate loyalty, even where the rules do not strictly require it. The gap between regulation and enforcement is far smaller online than offline.
The Broader Picture
The AI companion crackdown is part of a wider pattern. Beijing is tightening control over its tech industry while simultaneously pushing for technological supremacy. It wants AI innovation, but only on its terms. Emotional intimacy with a chatbot is not something the Chinese Communist Party can easily surveil, control, or use for its social engineering goals.
The regulations also signal a response to “AI psychosis”, a non-clinical term describing users losing touch with reality through prolonged chatbot interaction. Cases of explicit content bypassing existing filters have added urgency to the crackdown.
For Western observers, the rules themselves are broadly sensible and similar to what European regulators might propose. The difference is enforcement: in China, when the government tells tech companies to comply, they comply. Three and a half thousand products gone overnight speaks louder than any policy paper.
Source: Foreign Policy, Analytics Insight, Al Jazeera