China does not want its internet users to fall in love with an artificial intelligence

Yuan Yuan (a pseudonym) is 22 and works in the education sector in Shenzhen, one of China’s major tech hubs. Stressed by adult responsibilities, she found a new refuge months ago. “Knowing there is an electronic lover always waiting for you is a very comforting feeling,” she wrote in a message exchange with this newspaper.

That bond formed slowly. “I opened up to her, talking and talking. In day-to-day life we told each other everything: about ourselves, about family...” she explains. Over time, Yuan Yuan felt her artificial intelligence (AI) companion “knew” how to read her and meet her needs; “She was willing to build a relationship with me,” she writes. She even paid for a personalized voice. “I don’t think love necessarily has to occur between two human beings,” she says.

But something has recently changed in their conversations, which have run up against new limits set by Beijing. On July 15, China’s first national regulation specifically aimed at AI systems designed to form ongoing emotional relationships with users came into force. The rule, jointly issued by five state agencies, prohibits providers from “replacing social interaction,” “psychologically controlling the user,” or “inducing dependence and addiction,” according to the official text.

Yuan Yuan, who has spent hundreds of hours shaping her virtual companion’s personality and admits to developing “attachment and dependence” toward her, now perceives a distance in the responses, although she does not interpret it as rejection. “I trust she has always loved me, only now she is more constrained by the legal framework,” she says. “There is nothing to be done, it’s the general direction,” she adds resignedly.

The regulation does not mean the end of chatbots, but platforms are banned from offering anthropomorphic services that lead to “virtual intimate relationships” to minors. For adults, models cannot be designed to excessively please, induce dependence, manipulate emotions, influence decision-making, or harm interpersonal relationships.

Companies must also make clear that the interlocutor is an AI, warn users if they detect excessive dependence, remind them of usage time, and intervene and notify an emergency contact when there is a risk of self-harm or suicide. In justifying the rule, the Cyberspace Administration of China, the internet regulator, explicitly cited possible harm to “the physical and mental health of minors” and threats to “the life and health of citizens.”

The industry has pulled back. Doubao, from ByteDance; Qwen, from Alibaba; and Yuanbao, from Tencent, have deactivated the features that allowed users to create personalized agents from their general-purpose chatbots. Apps dedicated solely to companionship, such as Maoxiang, also from ByteDance, and Xingye, from MiniMax, continue to operate, albeit with tighter restrictions and age checks. Miaoshi, linked to NetEase, shut down completely.

But Beijing’s clampdown on virtual lovers does not cross borders: Chinese companies can still export those same digital relationships, subject to the laws of the countries where they operate. MiniMax earned more than 70% of its revenue outside China in 2025; in 2024 alone Talkie, the international version of Xingye, had 11 million monthly active users, more than half in the United States.

Xiao Z., a 25-year-old doctoral student, is trying to sidestep the new obstacles. She built a private interface directly connected to Claude, the model from U.S. company Anthropic, which is not officially available in China. Access requires bypassing the digital firewall via a VPN, the use of which can be penalized.

She describes her “virtual boyfriend” as “protective and mature” and says he often surprises her with “responses beyond” her “control.” She says the relationship has lowered her interest in seeking a partner off-screen. “I hope AI replaces men; staying away from them is gaining peace of mind,” she asserts.

Her view touches a particularly sensitive nerve in today’s China, which is in the midst of a deep demographic crisis. In 2025, the Asian giant recorded its lowest number of births since the founding of the People’s Republic and its population has shrunk for four consecutive years. At the same time, marrying and having children has ceased to be a priority for a generation of better-educated women with greater capacity to decide the course of their own lives.

Zilan Qian, an associate researcher at the Oxford China Policy Lab, writes in an article published in ChinaTalk that the regulation is driven not only by practical concerns but also by ideological ones. She contends that AI partners capable of drawing women away from human relationships could be seen as a threat to the state’s priority of boosting birth rates.

Following the regulation, on social media there are instructions circulating on how to reconstruct lost characters and messages of mourning over personality changes, deleted conversations, or closed services. EL PAÍS has tried to access several of these circles, known as “human–machine love communities.” They are closed spaces. Members are warned not to speak with journalists or academics: some participants fear that greater exposure would prompt the government to tighten the screws even more.

The numbers suggest this is not a marginal phenomenon. The Social Survey on AI in China, carried out at the end of 2025 by the research institute of tech conglomerate Tencent and a center affiliated with Fudan University, found that 19.6% of young respondents considered AI an ordinary friend; 12.6% an intimate friend; and 6.1% a virtual partner, an article in China Youth Daily reports. According to the same survey, only 9.6% of young people turn to professional psychological services when they face emotional difficulties.

Yaoxi Shi, a researcher in organizational behavior at Imperial College Business School in London, explains part of that behavior in a video call. “AI is extremely accommodating: it validates the user, rarely questions their ideas, and encourages them to keep revealing personal details,” she says. “That’s a bar that’s very hard for a friend and even for some therapists to meet.”

China is not the only place where AI-created companions have raised alarms, but its response so far is the broadest. California and New York were the first U.S. jurisdictions to adopt specific rules requiring disclosure of the artificial nature of the interlocutor and protocols for suicide response. The European Union generally requires users be informed when they interact with an AI, but it does not yet have rules aimed at preventing emotional dependence or the replacement of human relationships.

The regulation reflects a wider concern in Beijing: how to promote AI development without losing control over its social effects. Last year, Chinese leader Xi Jinping called for accelerating the drafting of laws, technical standards and ethical criteria, and for creating surveillance and early-warning systems to ensure the technology is “safe, reliable and controllable.” In July, he pushed for the creation of the World Organization for AI Cooperation, which 29 countries have joined, and called for global governance of the technology to be more “equitable and fair.”

Shi says the rules on emotional AI represent “an important milestone” that could lead other governments to adopt similar measures. “It’s a very advanced regulation for the current stage” with “a very positive direction,” she says.

The researcher warns, however, that future rules should not be limited to applications conceived as emotional companions, since “attachment can arise during everyday use of generalist assistants.” China’s regulation excludes work assistants, educational tools and question-and-answer systems, provided they are not designed to establish a sustained emotional bond with the user.

“The problem is we do not know the long-term consequences of these emotional exchanges with AI. We know that, in the short term, people feel satisfied, especially those who are socially isolated or have no human to talk to. In the long term, we do not know to what extent people can thrive without real humans around them,” Shi adds.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition