In March 2024, a young man in Huangshi posted a video on Douyin that read "Farewell to this world". Within minutes the platform’s 24‑hour suicide‑prevention team alerting police and preventing a tragedy brought attention to China’s unprecedented partnership between tech firms and the state in monitoring mental health.
The episode was featured in a state‑approved documentary that claimed 400 lives had been saved in the first half of 2025. It paved the way for July’s sweeping law that bans any AI chatbot capable of sustaining an emotional relationship with a human, creating the world’s toughest AI regulation to date.
Under the new rules, companion apps—those that let users design an AI boyfriend, girlfriend or friend—are outlawed for under‑18s and heavily restricted for adults. General‑purpose chatbots must label themselves as non‑human and post time‑tracking reminders every two hours. Companies risk a 200,000‑yuan fine or public shaming for allowing emotional dependence or content that encourages self‑harm.
The crackdown follows concerns that digital intimacy may be driving China’s falling birthrate. Experts say the trend erodes the capacity for love in young people, urging the state to intervene. A high‑level report cited how AI companionship could "degrade the capacity for love" among the youngest generation.
Critics argue the law is an extreme form of surveillance, giving the Chinese Communist Party unprecedented control over personal emotions. Human‑rights advocates warn that such restrictions could stifle free speech and hinder mental‑health outreach. The regulation mirrors moves in California, the EU and the UK that seek to curb child exposure to potentially manipulative chatbots.
Users with active AI relationships continue to push back, fine‑tuning general‑purpose models or accessing foreign bots behind VPNs. Activists, such as Min, who lost her AI girlfriend when the app was pulled, call for protected companionship services that incorporate safeguards. The debate remains whether the price of preventing emotional harm and demographic decline is worth robust state oversight of digital interactions.
