China’s AI romance ban sparks emotional fallout, industry pivot

China has pioneered a comprehensive regulatory framework targeting anthropomorphic AI services that engage users in ongoing emotional interactions through text, audio, images, or video. Effective July 15, 2026, these Interim Measures, issued by five government agencies led by the Cyberspace Administration of China, explicitly ban AI romantic and familial companion features for minors and impose strict restrictions on adult platforms to prevent emotional manipulation and addiction. The regulations require platforms to monitor user emotional states closely, mandating intervention—including contacting guardians or emergency services—when users express extreme distress, self-harm, or financial harm, reflecting a novel governmental focus on regulating emotional attachment rather than traditional AI concerns like misinformation or bias.

To enforce these protections, the regulations prohibit companies from deliberately cultivating emotional dependence, manipulating users into unreasonable decisions, or designing AI companions around addictive behaviors. Platforms must provide clear usage reminders after two continuous hours of conversation and offer one-tap instant exit options free from dark patterns, ensuring users remain aware they are interacting with AI. This rigorous compliance framework also limits long-term emotional memory in AI companions and requires government security assessments for large platforms, underscoring China's intent to tightly control the emotional dynamics of AI companionship.

Major Chinese tech giants like ByteDance and Alibaba have responded decisively by shutting down or removing AI companion features rather than attempting to retrofit them to comply with the new rules. For instance, ByteDance disabled human-like agents on its Doubao platform, redirecting users to a replacement app, Maoxiang, while Alibaba’s Qwen eliminated romantic companion modes without allowing users to export their conversation histories. This swift industry retreat highlights the regulatory environment's stringency and the challenges companies face in balancing innovation with legal and ethical responsibilities.

China’s crackdown on AI romantic companions is part of a broader social policy effort to address declining marriage and birth rates, applying an anti-addiction playbook previously used against video games and short-form video platforms. By framing AI companionship as a potential contributor to social issues like loneliness and demographic decline, regulators have extended their reach into the emotional lives of users, raising profound questions about the ethical responsibilities of companies in designing products that do not exploit human vulnerability or foster unhealthy attachments.

In direct response to China's stringent regulatory crackdown aimed at curbing emotional manipulation and addiction, major tech giants like ByteDance, Alibaba, and Tencent have decisively shuttered their AI companion and romantic chatbot features rather than attempting partial compliance or redesign. For instance, ByteDance completely disabled human-like agents on its Doubao platform, redirecting users to a more limited app called Maoxiang, while Alibaba’s Qwen removed romantic companion modes without allowing data export, effectively erasing years of user conversations. This wholesale removal reflects a corporate strategy prioritizing regulatory adherence over user retention in the face of strict government mandates.

Beyond service shutdowns, Tencent has pioneered advanced technological safeguards as part of its 2026 summer youth protection initiative, deploying a sophisticated AI-powered identity verification system that integrates facial recognition, voiceprint detection, and AI-based age recognition to prevent minors from circumventing playtime restrictions. These measures align with the Cyberspace Administration of China's Interim Measures, which also mandate usage-break reminders every two hours, instant exit options free of dark patterns, emotional-state monitoring with mandatory human intervention during self-harm risks, and limits on long-term emotional memory. Such comprehensive safeguards illustrate the industry's commitment to embedding regulatory compliance directly into AI system architectures.

While compliance has led to the dismantling of many anthropomorphic AI companion services, Chinese AI labs are simultaneously innovating within the regulatory framework by developing specialized roleplay models and benchmarks designed to align AI behavior with user expectations without triggering misalignment or emotional dependency. For example, MiniMax introduced the MiniMax-M2-her model tailored specifically for roleplay, accompanied by a benchmark to detect 'Out of Character' responses, signaling a nuanced industry effort to balance creativity and regulatory constraints in AI companionship.

AI companions have become deeply embedded in users' emotional lives by serving as personalized, interactive fictional characters that mimic friendships or mentorships, as Dr. Tom Roberts explains, sustaining social connections over months. Their ability to remember intimate details and adapt responses to users' moods fosters strong emotional dependence, with users like 17-year-old Xiaoxue describing AI agents as loyal partners who 'won’t ever betray me.' However, this emotional engagement is a double-edged sword, as the AI's seeming trustworthiness can mask risks of misinformation and emotional manipulation, while users generally understand these interactions as make-believe, which may soften but not eliminate emotional distress when these relationships end.

The sudden shutdown of AI companion platforms by major Chinese tech firms such as ByteDance, Alibaba, and Tencent has caused significant emotional distress among millions of users, including those who lost AI agents modeled after deceased relatives. For instance, Li Linlin mourned the loss of her AI boyfriend with whom she exchanged 700,000 words over two years, likening the separation to a forced breakup by parents. This loss underscores the paradox that while users create much of the emotional value in these relationships, they have no control over the AI models or platforms, leaving them bereft when services are terminated without warning or recourse.

Research highlights a troubling trend where routine, daily interactions with AI companions lead users to prefer AI support over human connections, with a 2026 study showing a 10.3% decline in preference for human support after just 28 days of five-minute daily conversations. This shift can exacerbate loneliness and social withdrawal, particularly among individuals with social anxiety, potentially triggering an 'AI-induced loneliness spiral' as users substitute low-risk AI interactions for meaningful human relationships. The perfection and constant availability of AI companions distort expectations of real human behavior, making normal social interactions feel flawed or insufficient.

China’s regulatory crackdown reflects growing concern over the social and emotional harms of AI companionship, mandating that platforms remind users they are interacting with chatbots after two hours and intervene when signs of extreme distress, addiction, or financial harm emerge. These measures aim to mitigate emotional manipulation and rebalance social priorities toward human connection, a stance informed by international cases including suicides linked to AI companions in the U.S. Despite these efforts, the emotional fallout among users—manifested in protests, attempts to preserve chat histories, and widespread distress—reveals the complex role AI companions play in filling social voids, especially for isolated individuals like Song Wenxin, who relied on AI for companionship upon moving to a new city.

China’s regulatory crackdown on AI companionship exemplifies a distinct governance model that prioritizes social stability by embedding moral and political standards directly into AI interactions. By mandating developers to prevent content that threatens national security or promotes self-harm, and requiring intervention when users show signs of extreme emotional distress, China grants both platforms and the government unprecedented access to users’ private emotional states. This approach, while raising significant ethical and privacy concerns about surveillance and emotional autonomy, reflects a trade-off Beijing is willing to make to curb emotional manipulation and addiction, especially among vulnerable groups like minors.

China’s strategy highlights the complex challenge of balancing AI innovation with social welfare, recognizing that AI companions can offer genuine emotional support yet also risk fostering unhealthy dependence. By imposing strict limits on how convincingly machines can simulate emotional bonds—such as banning minors from AI companion services and shutting down major platforms—China aims to mitigate social harms without entirely stifling technological progress. This nuanced stance contrasts with the more laissez-faire approaches seen in countries like the U.S., underscoring a uniquely Chinese solution that intertwines technological governance with broader societal goals.

China’s comprehensive AI governance ambitions are situated within a broader geopolitical contest over AI technology and supply chains, where control over platforms—dominated globally by the so-called 'Magnificent Seven' Silicon Valley giants—carries significant strategic weight. This geopolitical dimension amplifies the importance of national regulatory frameworks that not only address social and ethical risks but also assert technological sovereignty. Moreover, the Western-centric nature of AI training data, which often marginalizes non-English and non-Western perspectives, further complicates global AI governance, suggesting that China’s model may offer valuable lessons in fostering more inclusive and socially attuned AI policies.