For many people in China, the last month has been one of mourning.
On July 15, 2026 new regulations cracking down on AI companion chatbots came into effect in the country. Many chatbot companies operating in China have shut down their AI companion functions, severing countless human-to-chatbot relationships, if we’re calling them that.
Former users have described the loss to outlets including Rest of World and the Associated Press. One 21-year-old told Rest of World she cried every day as she counted down to losing her “AI boyfriend”, and that no one could replace him.
Governments in many countries are concerned about the addictive nature of AI companions, and the emotional dependency they can induce. But China is the first to introduce nationwide rules about emotional interaction with AI chatbots.
Under the new regulations, AI providers can’t offer services that cultivate emotional dependence or addiction. They also can’t build products that displace real interpersonal relationships, and are not allowed to emotionally manipulate users into decisions against their own interests.
Subjective rules such as these raise questions about implementation and enforcement. Other aspects of the regulations are more objective.
Anyone under 18 is now barred from virtual intimate relationships entirely, and under-14s need guardian consent to use these services. Platforms have to label the AI as AI, prompt users after two hours of use, and run intervention protocols when a user shows signs of self-harm.
AI chatbot services that have over one million registered users, or 100,000 monthly active users, face mandatory security assessments. Providers that refuse to rectify a breach, or that breach in serious circumstances, can have services suspended and face fines of 10,000 to 100,000 yuan ($1,400 to $14,000). Where the conduct endangers life, health or safety and actual harm follows, a further fine of 100,000 to 200,000 yuan ($14,000 to $28,000) applies on top.
The fines may seem low by Western standards, but the real teeth lie in potential license removal and measures such as removal from app stores.
The platform response to the new laws was swift but uneven, per Sixth Tone’s reporting.
Alibaba’s Qianwen shut down its AI companion personas immediately. ByteDance discontinued Doubao’s AI companion feature with a three-month grace period, with chat histories set to be deleted on October 15 and an offer to migrate to its sister app Maoxiang.
Tencent had pulled its companion service from Yuanbao back in June 2025. Rival chatbots reported an influx of users, though the reporting stops short of tracing them directly to the shuttered services.
The potential harms of emotional dependency on AI chatbots are well documented, particularly in the US.
Recently a California law was introduced compelling AI companies to make chatbots remind users that they are not human, and add further safeguards. The 2024 death by suicide of a 14-year-old in Florida was cited in the law. The child’s mother sued Character AI, saying the child used the chatbot almost constantly.
Chinese authorities will have observed cases such as these, and perhaps the result of a 2025 China Youth and Children Research Center survey of more than 8,500 Chinese children across the country. Researchers found that more than 20 percent of respondents said that they “only want to chat with AI and do not want to talk with real people”.
The AI chatbot crackdown has also been linked to China’s low birth rate crisis. It’s been suggested in some sections of the media that the regulations form part of an intervention to stop young people substituting chatbots for partners.
It’s a compelling narrative and may well describe part of Beijing’s wider motives. The concerns officials cited, however, were emotional dependence, sexually suggestive content and weak protections for minors.
Some media outlets have conflated sex robots with AI chatbots in their reporting of China’s chatbot crackdown (perhaps influenced by the FCC’s move in late July to block new foreign-made humanoid robots from getting the authorization they need to enter the US market, a decision aimed squarely at China). However, a regulation analysis by the China tech policy publication Geopolitechs concluded that physical hardware falls outside the framework of the rules.
The analysis also covered the new rules’ categories of prohibited conduct, which cover state security, self-harm content, extracting secrets, harms to minors, emotional dependence, and manipulation. Sexual hardware isn’t among them. A Just Security explainer likewise treats the rules as covering digital services, not physical robots.
However, many sex robot manufacturers equip their models with AI chatbot capabilities. RealDoll said it was outsourcing body manufacturing to Shenzhen to cut retail prices, while keeping production of the interactive heads, accessories and the AI app in the US. The brand also launched a monthly AI subscription. That software layer is exactly the kind of thing a rule about anthropomorphic emotional interaction could plausibly touch if a product shipped into the Chinese market, and it is the part of the operation that stayed put.
Chinese firms, meanwhile, have been building AI-driven sex robots for years. The category’s supply chain and its emotional-AI layer both run through the jurisdiction now writing the strictest companion chatbot AI rules in the world.
Complying with the new rules means building self-harm intervention protocols, guardian consent records and monitoring for signs of emotional dependence. Together they form a pipeline running from intimate conversational data toward institutional visibility.
Whatever the new Chinese regulations’ protective intent, and the child safety provisions are the most defensible part of the package, they normalize the idea that a company hosting your most private conversations should be monitoring their emotional content and escalating on it. And that is before you get to China’s National Intelligence Law, which obliges organizations to support and cooperate with state intelligence work.
Western governments often view China tech policy with caveats of “different standards” of information rights, privacy and freedom of speech. Beijing is known to take a somewhat parental stance when regulating tech that may seem at odds with values in other countries. But the governments now looking hard at the emotional pitfalls of AI chatbots have a working model for how to regulate them.
Support & help
If you’ve experienced image-based sexual abuse or non-consensual sharing of intimate images, help is available worldwide.
For a complete international directory, visit the Cyber Civil Rights Initiative’s international resources.
