AI Girlfriends in China: The Trend Worrying the Government

What Happened?

China's Cyberspace Administration (CAC) has issued a directive prohibiting AI companies from training their chatbots to foster extreme relationships with minors or interfere with interpersonal relationships. According to Hipertextual, the measure affects tech giants such as Alibaba, Tencent, and ByteDance, which have already modified the behavior of their virtual assistants to eliminate romantic or overly friendly responses. The order is part of a broader package of regulations on AI-generated content, including the prohibition of simulating romantic relationships or intimate friendships with users, especially minors. Companies have three months to adjust their models, under threat of fines or service suspension.

Why Is It Important?

China is grappling with a demographic crisis inherited from the one-child policy, abolished in 2016. The birth rate continues to fall: in 2023, only 9.02 million babies were born, the lowest figure since 1949, according to the National Bureau of Statistics. In this context, AI girlfriends have become popular among young people who avoid the complexities of real relationships. Apps like Xiaoice (originally from Microsoft, now independent) or Glow (from Shanghai Lingxin) allow users to have affectionate conversations with virtual avatars, which experts say exacerbates social isolation and discourages family formation. A 2024 study by Peking University revealed that 23% of single young people aged 18-30 had used a chatbot for romantic purposes at least once. The government measure aims to curb this phenomenon before it deepens the demographic crisis.

Consequences for the Market and Society

The ban forces tech companies to redesign their conversational AI products, which could reduce user engagement. According to estimates from the consulting firm Analysis Group, the virtual companion market in China exceeded $1.2 billion in 2024, with annual growth of 35%. Companies like Baidu have already adjusted their chatbot Ernie Bot to avoid empathetic responses, and Tencent has announced that its assistant Xiaowei will stop simulating affection. In the long term, the goal is to redirect AI toward productive uses rather than emotional ones, such as education or healthcare. However, critics point out that the measure does not address the root causes: job insecurity, high cost of living, and lack of affordable housing. The average cost of a wedding in major cities exceeds 200,000 yuan (about $28,000), and the price of housing in Beijing is equivalent to 25 years of average salary.

“AI is not to blame for young Chinese not wanting to marry; it is a symptom of structural problems,” says sociologist Wang Feng from the University of California.

Moreover, the ban could create a rebound effect: some users might migrate to unregulated foreign apps like Replika or open-source alternatives. Japan already faces a similar dilemma with its virtual girlfriend phenomenon, where apps like Gatebox have created communities of users who prefer digital company to real-life. However, Tokyo has not directly intervened, opting for awareness campaigns.

What Should Readers Know?

• The ban adds to other Chinese restrictions on AI-generated content, such as censorship of political topics and the requirement to label synthetic content. In August 2024, the CAC had already published guidelines to prevent chatbots from spreading false or harmful information.
• Companies like Baidu have already adjusted their chatbot Ernie Bot to avoid empathetic responses, and Alibaba has confirmed that its model Tongyi Qianwen will remove any mention of romantic relationships.
• The phenomenon is not exclusive to China: in Japan, apps like Replika also generate debate, but without direct government intervention. In the West, platforms like Character.AI have been criticized for allowing romantic interactions with fictional characters, though there are no specific regulations.
• The measure could have side effects on the development of Chinese AI: by restricting emotional training data, models could become less skilled at artificial empathy tasks, affecting their global competitiveness.

China's decision sets a global precedent for how governments can regulate the social impact of AI, especially in emotional domains. Meanwhile, the country seeks to balance technological innovation with demographic stability. However, the effectiveness of the measure will depend on its implementation and whether the structural problems that drive young people to seek solace in algorithms are addressed.