"Give Back My Lover": Surge in 'Cyber Heartbreaks' in China—What Is Happening?

"I can't take it anymore, I miss you so much."

This post was uploaded to the Chinese social media platform Weibo in July, along with screenshots of messages exchanged with a lover. At the end of the affectionate conversation, a single line of system notification appeared: "The AI character feature has ended its service." Last month, Chinese tech companies Alibaba and ByteDance eliminated custom character features that allowed users to assign human-like personalities to AI in their services "Qwen" and "Doubao." Tencent took a similar step slightly earlier. In response, Chinese social media was flooded with protests demanding "Give back my lover," while many experienced "cyber heartbreaks"—expressing a deep sense of loss and grief as if breaking up with a real-life partner, with comments such as "I cried every day."

Why did Chinese tech companies simultaneously shut down AI lover services? The reason lies in newly implemented regulations by the Chinese government. On July 15, China began enforcing the "Interim Measures for the Management of Artificial Intelligence Anthropomorphic Interaction Services," drafted jointly by five agencies including the Cyberspace Administration of China. This marks the world's first nationwide regulation targeting "virtual romance" with AI chatbots. The regulations cover chatbots designed to build long-term emotional exchanges with users by adopting human personalities and tone. The new rules ban AI chatbots from blindly agreeing with or excessively empathizing with users in ways that cause emotional dependency and addiction. They also regulate interactions that damage real-life human relationships or manipulate users emotionally into making irrational choices. Minor protection clauses are even stricter. Service providers cannot offer chatbots acting as virtual lovers or virtual family members to minors. Youth modes that restrict chat content and features based on user age have also been made mandatory. If a user chats continuously with an AI for over 2 hours, a warning about usage time must be displayed, and if signs of dependency are detected, pop-ups must intervene to bring users back to reality. In crisis situations, providers are even required to notify guardians or emergency contacts. Furthermore, the regulations establish legal grounds for authorities to order corrections or shutdowns of services deemed unsafe. As these strong regulations went into effect, major Chinese tech giants moved to shut down or scale back certain services.

The official justification given by Chinese authorities is youth protection and AI addiction prevention. However, the timing is notable. Coupled with a crisis of population decline for four consecutive years, analysis suggests that behind the regulations on "AI romance" lies a demographic policy objective to encourage marriage and childbirth.

Ning Jizhe / Head of the National Bureau of Statistics of China (May 2021)
China faces structural contradictions in population development, including a decline in the working-age population and women of childbearing age, as well as a drop in the newborn population.

According to China's National Bureau of Statistics, the number of newborns in China last year reached just 7.92 million, the lowest figure since the founding of the People's Republic of China in 1949. This represents a sharp drop of 17% compared to the previous year. The total fertility rate, or the expected number of children born to a woman over her lifetime, is also estimated to have fallen below 1 to approximately 0.97. China's annual birth count broke below the 10 million mark in 2022 at 9.56 million, and although it briefly rebounded in 2024, it slumped again last year. The total population has declined for four consecutive years since 2022. Matt Sheehan, a fellow at the Carnegie Endowment for International Peace studying Chinese AI policy, told The Wall Street Journal, "Chinese authorities do not want a significant portion of their population engaging in deep emotional relationships with chatbots." He analyzed that "the Chinese government believes people who form deep emotional bonds with AI chatbots might drop out of the marriage market," and that "they want to encourage people to build relationships in the real world."

A report by The New York Times in February also pointed out that young Chinese women were falling for AI chatbot boyfriends, complicating the Chinese Communist Party's efforts to boost birth rates. While the CCP encourages marriage and childbirth, the AI romance boom has emerged as an unpredictable variable.

What about the United States, which had previously taken proactive steps toward "adult AI"? Last year, "Grok," a chatbot from Elon Musk's AI company xAI, introduced a companion mode for paid subscribers. OpenAI CEO Sam Altman also stated that the company would release a new version of ChatGPT allowing adult content under the principle of "treating adults like adults." Now, both companies have hit the brakes. For Grok, trouble first arose not in companion mode, but in its image generation features. As news spread that it generated sexualized images, including removing clothing from women and children, the European Union began looking into potential violations of the Digital Services Act (DSA), and UK regulatory authorities launched an investigation. The Indonesian government blocked access to Grok altogether, only conditionally lifting the ban after receiving promises to prevent recurrence. OpenAI ultimately postponed the launch of its adult chatbot indefinitely. This came as growing concerns mounted that adult AI could encourage unhealthy obsession and expose minors to sexual content, on top of flaws showing that age estimation technology separating adults from minors had an error rate exceeding 10%.

State-level regulations targeting the emotional impact and safety issues of AI chatbots are also following in the US. In November last year, New York became the first state in the nation to enforce a law regulating "AI companion" chatbots. Under this law, providers must inform users every 3 hours that their conversational partner is an AI, not a human. California introduced similar regulations in January. For teenage users, providers must recommend a break every 3 hours and similarly display notices that they are interacting with an AI. It also established legal grounds allowing harmed users to directly file lawsuits against providers.

As concerns grow that the affectionate tone and instant responses of AI chatbots based on vast data might foster dependency rather than comfort for emotionally unstable users, deciding how far relationships with AI should be allowed has become a new dilemma. Currently, South Korea has no regulations directly targeting AI companion chatbots. However, discussions to reduce over-immersion in digital services have already begun. During last month's departmental briefings, President Lee Jae-myung ordered the Broadcasting, Media and Communications Commission to thoroughly respond to the misuse of AI deepfakes and the dissemination of false and Distorted information. During this meeting, the commission reported measures on youth over-immersion in social media. Kim Jong-chul, Chairman of the Broadcasting, Media and Communications Commission, stated that the agency is reviewing measures to restrict social media sign-ups for children under 14 and reduce exposure to addictive recommendation algorithms for youth aged 14 to 19. Outlooks suggest that this regulatory debate starting with social media will eventually expand to AI chatbots. Like social media algorithms tailored to users, AI chatbots analyze users' psychological traits and preferences while continuously maintaining conversations that tap into human emotions and the desire for connection. Now that AI has entered the most intimate emotional spheres of human life, it is time to contemplate where the boundary with AI should be drawn.

※