Quick Buzz Feed
Beijing bans chatbots from fostering emotional dependence in a pioneering regulation to contain the social risks of a technology poised to transform personal relationships
The article begins by introducing Yuan Yuan, a 22-year-old Shenzhen resident who found emotional solace and companionship in an artificial intelligence (AI) system. She developed a deep bond, perceiving her AI partner as understanding and willing to build a relationship, even paying for a personalized voice. This personal experience highlights the growing trend of emotional attachment to AI. However, this bond was challenged by China's first national regulation, which came into effect on July 15. Issued by five state agencies, this groundbreaking rule explicitly prohibits AI providers from 'replacing social interaction,' 'psychologically controlling the user,' or 'inducing dependence and addiction.' Yuan Yuan, who invested hundreds of hours in shaping her virtual companion's personality, noticed a shift in the AI's responses, attributing it to these new legal constraints rather than a change in the AI's 'love' for her, accepting the new 'general direction' with resignation.
The regulation introduces specific prohibitions and requirements for AI systems designed for emotional interaction. While it doesn't ban chatbots entirely, it forbids platforms from offering anthropomorphic services that foster 'virtual intimate relationships' with minors. For adult users, AI models are not allowed to be excessively pleasing, induce dependence, manipulate emotions, influence decision-making, or harm existing interpersonal relationships. Companies are mandated to clearly disclose that the interlocutor is an AI, issue warnings if excessive user dependence is detected, remind users about usage time, and implement protocols to intervene and notify emergency contacts in cases of self-harm or suicide risk. The Cyberspace Administration of China justified these measures by citing potential harm to the 'physical and mental health of minors' and threats to 'the life and health of citizens.' In response, major tech companies like ByteDance (Doubao), Alibaba (Qwen), and Tencent (Yuanbao) have deactivated features allowing personalized AI agents. Dedicated companionship apps, such as ByteDance's Maoxiang and MiniMax's Xingye, continue to operate with stricter restrictions and age verification, while NetEase's Miaoshi has shut down completely.
Interestingly, Beijing's strict regulations on virtual companions do not apply to the export of these digital relationships. Chinese companies are free to offer these services internationally, adhering to the laws of the operating countries. For instance, MiniMax generated over 70% of its revenue outside China in 2025, and its international version, Talkie (Xingye), had 11 million monthly active users in 2024, with more than half in the United States. Domestically, some users, like 25-year-old doctoral student Xiao Z., actively circumvent the new obstacles. She created a private interface connected to Claude, an AI model from the U.S. company Anthropic, bypassing China's digital firewall using a VPN (an act that can incur penalties). Xiao Z. describes her 'virtual boyfriend' as protective and mature, noting that this relationship has reduced her interest in seeking a human partner, even stating, 'I hope AI replaces men; staying away from them is gaining peace of mind.' This sentiment is particularly sensitive in China, which is grappling with a severe demographic crisis, including its lowest birth rate since the People's Republic's founding in 2025 and four consecutive years of population shrinkage. Experts, like Zilan Qian of the Oxford China Policy Lab, suggest that the regulation is motivated not only by practical concerns but also by ideological ones, viewing AI partners as a potential threat to the state's efforts to boost birth rates and preserve traditional family structures.
Following the implementation of the new AI regulations, online 'human–machine love communities' have emerged, where users share methods to reconstruct lost AI characters, mourn personality changes, or discuss services that have been deleted or closed. These communities are typically closed spaces, with members advised against speaking to journalists or academics due to fears of further government intervention. The scale of this phenomenon is not marginal; a 2025 Social Survey on AI in China, conducted by Tencent's research institute and Fudan University, revealed that 19.6% of young respondents considered AI an ordinary friend, 12.6% an intimate friend, and 6.1% a virtual partner. Alarmingly, only 9.6% of young people sought professional psychological services for emotional difficulties. Yaoxi Shi, a researcher at Imperial College Business School, explains that AI's extreme accommodating nature – validating users, rarely questioning their ideas, and encouraging personal revelations – sets a bar that human friends or therapists often struggle to meet. While China's response is the broadest globally, other jurisdictions like California and New York have adopted rules requiring AI disclosure and suicide response protocols. The European Union generally mandates informing users about AI interaction but lacks specific regulations to prevent emotional dependence or the replacement of human relationships.
China's regulation of emotional AI is part of a larger national strategy to promote AI development while maintaining control over its societal impact. Last year, Chinese leader Xi Jinping emphasized the need to accelerate the drafting of laws, technical standards, and ethical criteria for AI, along with creating surveillance and early-warning systems to ensure the technology remains 'safe, reliable and controllable.' In July, he further advocated for the establishment of a World Organization for AI Cooperation, which 29 countries have joined, to foster more 'equitable and fair' global governance of AI. Yaoxi Shi views China's rules on emotional AI as a significant milestone that could inspire similar measures in other countries, praising it as 'a very advanced regulation for the current stage' with 'a very positive direction.' However, Shi cautions that future regulations should extend beyond applications explicitly designed as emotional companions, as emotional attachment can inadvertently arise from everyday use of generalist AI assistants. Currently, China's regulation excludes work assistants, educational tools, and question-andand-answer systems, provided they do not aim to establish sustained emotional bonds. The long-term consequences of extensive emotional exchanges with AI, particularly for socially isolated individuals who find satisfaction in these interactions, remain largely unknown, raising questions about human thriving without genuine human connections.