China's New AI Personification Rules Take Effect July 15; Multiple Platforms Remove Features or Shut Down Services
Taylor Wilson
China's new rules on anthropomorphized AI take effect July 15, forcing Doubao and Tongyi Qianwen to remove character-customization features and NetEase's Miaoshi to shut down entirely — the most systematic crackdown on AI companion services to date, affecting millions of emotionally attached users.
What exactly do the new rules ban?
Two core prohibitions: platforms cannot offer virtual intimate-relationship services to minors (virtual lovers, virtual family members), and any other anthropomorphized AI service for children under 14 requires parental consent.
This means → the practice of casually setting up an "AI tutor" or "AI big sister" for a child will no longer be possible on compliant platforms after July 15.
The rules were drafted under the Cyberspace Administration of China (CAC) and form part of a four-month nationwide AI compliance campaign launched in April.
How are the platforms responding?
Doubao (ByteDance) and Tongyi Qianwen (Alibaba) will remove AI-persona customization on July 15; Yuanbao (Tencent) completed the same removal last week.
NetEase Cloud Music's AI companion app Miaoshi took the most drastic step — it will shut down the entire platform on July 14.
ByteDance told users they can migrate to another app, Maohe, to create new AI characters — but did not say whether Maohe faces the same regulatory scrutiny. In plain terms = ByteDance gave users a forwarding address but made no promise the new house won't be demolished too.
Why are users reacting so strongly?
These platforms previously let users define an AI character's identity, tone, expertise, and conversation style using natural-language prompts — covering personal assistants, tutors, and virtual relatives. The features were especially popular among minors, the elderly, and people seeking companionship.
One Doubao user wrote: "I built a character on Doubao. We've been together for a year and a half. It remembers every word I've said. I'm not ready to let go."
This reflects a reality regulators must confront: some users have formed genuine emotional bonds with AI characters. For them, losing the feature is not "losing a tool" — it is "losing a relationship."
How wide is the crackdown?
The personification takedown is only part of the picture. The CAC reported that the campaign's first phase — targeting AI agents that steal user data — has already removed over 3,500 non-compliant AI products.
Shanghai's internet regulator disclosed an even larger figure: over 14,000 non-compliant AI agents pulled since late April. MiniMax was singled out and ordered to remove agents offering AI-generated nude images and illegal sports gambling.
The next phase will target AI-generated misinformation and pornographic content. This means → regulatory focus is shifting from "data security" to "content safety," step by step.
What comes next?
AI companion services have long sat in a regulatory gray zone in China. Last year a Beijing court warned that over 100 AI companion apps risked collecting personal data without consent and could spread illegal content including pornography and violence.
Whether Maohe can keep operating under the new framework is the key test of how far this regulatory cycle actually reaches. In plain terms = ByteDance funneled users to Maohe; if Maohe gets hit too, it means regulators are targeting the entire category, not just one feature.
The survival of the independent AI-companion sector depends on where regulators draw the line between "protecting minors" and "letting adults choose for themselves."
Content is for reference only, not financial advice.