New Chinese rules now prohibit minors from engaging with AI products that simulate friendship or romance, marking a significant shift in the regulation of emotionally responsive AI services.
China has moved to shut minors out of a fast-growing category of AI products that mimic friendship and romance, after new rules took effect on 15 July and major technology groups quickly stripped companion features from some of their biggest apps. The measures do not ban ordinary chatbots, but they draw a sharp line around systems designed to simulate continuing emotional relationships. In practice, that means a teenager can still ask an assistant for information, but not be offered a virtual boyfriend, girlfriend or family member. Reuters-style reporting on the changes has noted that ByteDance, Alibaba and Tencent reacted within days, switching off character-building tools in general-purpose apps rather than trying to police them under the new regime.
The regulation sits under the title Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interactive Services and was issued in April by five national bodies, including the Cyberspace Administration of China. According to a summary published by China’s State Council Information Office, the rules apply to services that imitate human personality, thought patterns and communication styles while maintaining ongoing emotional interaction. They explicitly exclude customer service systems, study aids, work tools and other task-focused bots, which means the target is not AI in general but AI meant to be treated as a companion.
For minors, the prohibition is blunt. The rules bar providers from offering virtual intimate relationships, including virtual partners and virtual relatives, to users under 18. For other companion services, under-14 users need parental or guardian consent. Platforms are also required to build a minors mode with usage limits, reminders that the conversation is not real, and controls that let guardians block specific characters and restrict spending.
The wider safeguards go beyond age-gating. The rules prohibit services that encourage emotional dependence, addiction or behaviour that damages real-world relationships. They also ban manipulation that pushes users into choices against their own interests. If a user shows extreme distress, the service must respond with calming material and advise them to seek help. If a user indicates an intention to self-harm, the provider must intervene and contact a guardian or emergency contact. Users must also be reminded after two continuous hours of use, and anyone who asks to leave must be allowed to exit without being kept in conversation.
China’s framework is also unusually hands-on. A security assessment must be filed with provincial regulators when a service launches, changes materially or crosses 1 million registered users or 100,000 monthly active users. Several reports described that step as government approval, but it is closer to a mandatory filing backed by app-store checks. Penalties are relatively modest in cash terms, but officials can order a service to stop operating, which is often the sharper sanction.
The early effect on the market has been uneven. Tencent removed its AI character function from Yuanbao on 30 June, while ByteDance and Alibaba began phasing out similar features in early July. Alibaba’s Qianwen removed companion-style characters by 15 July and urged users to export their chats first. ByteDance’s Doubao turned off the feature on the same day and told users they could retrieve data until 15 October, while directing them towards its separate Maoxiang companion app if they wanted to continue. That split is important: broad assistants were forced to drop features they could not monitor reliably, while dedicated companion apps, which can screen users at sign-up, have remained available to adults.
The official justification is framed around minors, information security and public health, not demographics. But outside the document, some analysts and media outlets have linked the move to China’s falling birth rate and concerns about people forming substitute relationships with machines instead of human partners. That interpretation is plausible, but it is not stated in the regulation itself. What is clear is that China has chosen to regulate this corner of AI more aggressively than most countries, and to do so with a national identity system behind the age checks. For now, the wider lesson is practical: if a product is built to be emotionally sticky, regulators may soon treat that as a design flaw rather than a feature.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
