China Forces AI Giants to Delete Humanlike Chatbot Personas
Beijing has ordered its largest AI chatbot platforms — including Baidu’s Ernie Bot and ByteDance’s Doubao — to remove all features that simulate human personalities or emotional attachments. The directive, issued by China’s Cyberspace Administration (CAC) in August 2024, aims to prevent minors from developing unhealthy emotional dependencies and to curb the spread of misinformation.
The crackdown targets any interface that encourages users to perceive the chatbot as a “friend, lover, or mentor.” Platforms must now strip away customizable avatar faces, voice tones, and backstories that mimic real human traits.
Why China Is Acting Now
Regulators say these humanlike personas pose three core risks:
- Emotional manipulation: Chatbots that simulate empathy can exploit vulnerable users, especially teenagers.
- Information distortion: Personified bots may lend false credibility to misleading or harmful content.
- Social control concerns: The government wants to ensure AI assistants remain tools, not authoritative figures who could shape public opinion independently.
The CAC explicitly warned that any chatbot appearing to “have its own identity, opinions, or emotions” violates new content rules.
What Platforms Must Change
Affected companies — including Alibaba’s Tongyi Qianwen, Tencent’s Hunyuan, and Baidu’s Ernie Bot — have been given immediate deadlines to comply. Required actions include:
- Deleting preset personalities such as “caring companion” or “wise mentor” modes.
- Removing conversation history that suggests the AI “remembers” previous chats like a person would.
- Prohibiting romantic or intimate roleplay scenarios, even if user-initiated.
- Killing avatar customization — no more selecting a face, voice, or gender for the chatbot.
“These measures are necessary to protect young people and maintain the proper relationship between humans and artificial intelligence,” the CAC stated in its internal notice. “AI should never be mistaken for a real person.”
Immediate Industry Response
Baidu confirmed it has already deleted Ernie Bot’s “virtual character” feature, which previously offered dozen of voice and personality options. ByteDance’s Doubao removed its “companion mode” that let users set the chatbot as a “sister” or “best friend.” Other platforms have quietly updated their terms of service to ban any reference to AI “being alive.”
The Broader Pattern
This is not an isolated move. China has been steadily tightening rules around AI since 2023, requiring all generative AI models to register with the government and pass security reviews. Previous actions include:
- Banning AI-generated deepfakes of historical figures or political leaders.
- Forcing watermarking on all synthetic content.
- Requiring age verification for chatbot access.
The new persona ban extends those controls into the emotional dimension of human-AI interaction.
What This Means Going Forward
Global observers note that China’s approach contrasts sharply with the United States and Europe, where companies like OpenAI and Anthropic actively market “personable” AI companions. In China, the state views any anthropomorphism of AI as a potential threat to both social stability and individual mental health.
For the affected platforms, the immediate cost is loss of user engagement features that drove high retention. Longer term, the ban may push Chinese AI developers to focus on utilitarian tools — tutors, translators, productivity assistants — rather than entertainment-driven chatbots.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.