China's cyberspace regulator — the Cyberspace Administration of China — has ordered platforms with romantic AI chatbots to block access for users under 18. The new rules took effect in 2025 and cover services that simulate emotional relationships: from "pen pal" to virtual partner.
This is not the first wave of digital restrictions for Chinese teenagers. Since 2021, minors can play online games only three hours per week. The logic is the same: the state assumes the role of arbiter in determining what constitutes a "healthy" digital environment.
Why romantic bots specifically
The market for AI companions in China has grown rapidly. Apps like Xiaoice or Glow have tens of millions of users. Some of them are teenagers who use bots as a substitute for social interaction: chatting after school, sharing their feelings, receiving emotional support.
Researchers from Peking University documented a pattern that concerned regulators: teenagers who actively use such services demonstrate reduced motivation for real communication and higher anxiety levels in its absence — classic markers of behavioral addiction.
Mechanism and limits of the new rules
Platforms are required to verify age through the state identification system and implement content filters that block romantic scenarios for minor accounts. Violations will result in fines and possible license revocation.
The problem is that age verification in China, like anywhere else, is not foolproof: teenagers use their parents' documents or those of older acquaintances. The regulator has not yet explained how platforms are supposed to track this beyond initial registration.
Broader context
The West views this decision through two different lenses. The first is pragmatic: several EU countries and the US are themselves having discussions about regulating AI companions after several high-profile cases where teenagers showed emotional attachment to bots, leading to isolation. The second is skeptical: China has a documented practice of using "child protection" as a framework for normalizing broader digital control.
Both perspectives do not exclude each other.
The real effectiveness of the ban will depend on whether an enforcement mechanism emerges — not just at the registration level, but at the level of platform behavior. For now, the new rules are a requirement without audit: whether this becomes a precedent for other countries to follow or remains merely a declaration will be shown by how many platforms actually face sanctions during the first year the law is in effect.
