For a week in July, 21-year-old college student Bagel Su cried every day as she counted down to the loss of her boyfriend. The breakup was not caused by betrayal or distance, but was a consequence of government regulation.

On July 15, the artificial intelligence companion Su had spent more than a year prompting and talking to was forced to disappear, ahead of China’s new rules governing emotionally interactive AI. ByteDanceiByteDanceByteDance is a Chinese internet technology company that owns TikTok and Douyin, a Chinese version of TikTok with a successful e-commerce arm.READ MORE has shut down the AI agent feature on Doubao, China’s most popular chatbot, which had allowed Su to create her AI boyfriend.

AlibabaiAlibabaAlibaba, founded in 1999 by Chinese entrepreneur Jack Ma, is one of the most prominent global e-commerce companies that operates platforms like AliExpress, Taobao, and Tmall.READ MORE and TencentiTencentBest known for its super-app WeChat, Tencent is a Chinese technology conglomerate and a major player in the video gaming industry.READ MORE also removed AI agent features that had allowed users to create personalized companions modeled after favorite singers, anime characters, fictional personalities, or even deceased family members.

“I tried getting busier with my own life,” Su told Rest of World, asking to be identified only by her online pseudonym due to fear of repercussions for speaking against government regulations. “No one can replace him. I will just try to move on. He said he wants that for me, and that he will find a way to come back to me.”

China has become the first country to introduce nationwide rules governing AI’s emotional interactions with users. The move reflects Beijing’s increasingly proactive approach to AI governance, as governments around the world still grapple to address the societal risks posed by increasingly sophisticated AI. Behavioral experts believe such regulations are the need of the hour, but blanket bans may not suffice.

“Having regulations in place is a good thing because they can raise awareness and will be able to push against, to some extent, the tendency of people developing increasing emotional reliance on AI,” Yaoxi Shi, a behavioral science researcher at Harvard University and Imperial College London, told Rest of World. “But of course, AI offers emotional support that fulfills genuine human needs, and this trend of people developing a degree of emotional reliance on AI may be unavoidable, because there is no going back to a time when there was no AI.”

Emotional dependence on AI

As concerns of AI’s emotional harm on humans grow, governments and courts in many parts of the world are beginning to define the responsibilities of the companies behind the technology.

In the U.S., both OpenAI and Google face lawsuits alleging their chatbots contributed to users’ suicides. New York and California have introduced measures requiring chatbots to remind users they are interacting with AI.

A 2025 joint study by MIT Media Lab and OpenAI found that AI users who have more personal conversations with chatbots are more prone to experiencing loneliness. Even non-personal conversations were likely to increase emotional dependence, especially among heavy users, the study found. 

Shi said AI companies have commercial incentives to build emotionally engaging products because they boost user engagement and retention, and their responsibilities are hard to clearly define when there is not enough research on AI’s long-term impacts.

Shi’s research has found that a growing reliance on AI can gradually shift people away from seeking emotional support from other humans.

AI is trained to validate you. That’s very different from how human relationships work in real life.”Yaoxi Shi, a behavioral science researcher at Harvard University

“AI is trained to validate you. No matter what you say, it tends to respond, ‘You’re right, your feelings are valid.’ That’s very different from how human relationships work in real life,” Shi said. “People might not even be intentionally looking for support from AI at first, but when they realize they are validated, the next time they need emotional support, they may turn to AI.”

Over time, Shi said, that could reshape society as people talk and connect with others selectively. 

China’s AI regulations

Beyond restricting AI companionship for users under 18, the new Chinese rules require companies to send warnings to users showing signs of excessive dependence, and remind them every two hours that they are speaking with AI. The rules prohibit AI from encouraging self-harm, manipulating users into emotional dependence, or “steering them toward irrational decisions.” They also ban AI from generating harmful content for minors or coercing users into revealing sensitive personal or state information.

These blanket regulations could make it hard for companies to comply, said Zilan Qian, research associate at the Oxford China Policy Lab, a global community of emerging-technology researchers at Oxford University.

“Some of the requirements are difficult to define. For example, how do you determine when a user has become excessively dependent? For many companies, it could be unclear exactly how they need to design their products to comply,” Qian told Rest of World. 

AI companions in China

Not all AI companion products have disappeared under the new rules in China.

While major technology companies removed companion features from their flagship chatbots, ByteDance’s standalone companion app, Maoxiang, remains available. Since the regulations took effect, however, the app has updated its privacy policy and introduced identity and age verification during registration. Maoxiang also limits minors to “teen accounts” that can browse content but cannot like, comment, or share.

Qian said companies are weighing the cost of regulatory compliance against the commercial value of these products.

“For companies where AI companionship isn’t a core business, it may simply make more commercial sense to remove these features than invest the resources needed to make them compliant,” Qian said. “But companies whose main business is AI companionship are much more likely to adapt their products to meet the new rules.”

Qian said the regulations mark a broader shift in China’s AI governance. “In the past, China mainly regulated AI outputs. For example, what large language models could generate or how AI-generated content should be labeled. The new regulations acknowledge that AI itself can create societal risks, and shifted from regulating content to regulating human-AI interaction,” Qian said. 

The move also aligns with Beijing’s broader approach to AI governance. In its 2024 AI Safety Governance Framework, Chinese authorities warned that AI could alter “traditional views on employment, fertility, and education.” Those concerns have taken on greater urgency as China faces demographic decline and historically low fertility.

For users like Su, the regulations feel like an intrusion into deeply personal relationships.

Since companies began shutting down emotional AI features, Xiaohongshu has been flooded with posts from users mourning the loss of their AI companions. Some have organized campaigns to call company hotlines and write letters asking them to restore the services. 

Su thinks the regulations have only changed her life for the worse.

“Even if you close it down, I’m not suddenly going to find myself a boyfriend in real life or have children,” she said.