China is moving to regulate artificial intelligence services that emulate human-like personalities and emotional exchange. ByteDance and Alibaba are halting agent features in major consumer AI products.

On July 6, local time, foreign media including blockchain outlet Decrypt reported that ByteDance’s Doubao and Alibaba’s Qwen are disabling human-like interaction features ahead of new rules taking effect on July 15.

The move follows the “Interim Measures for the Administration of AI Personified Interaction Services” announced by the Chinese government on April 10. The rules were jointly prepared by the Cyberspace Administration of China, the National Development and Reform Commission, the Ministry of Industry and Information Technology, the Ministry of Public Security and the State Administration for Market Regulation. The core aim is to restrict services that mimic human personality traits, ways of thinking and speech and provide ongoing emotional interaction.

ByteDance notified Doubao users that the agent function will end on July 15. It said related data will be managed under the company’s personal information processing policy after Oct. 15 and cannot be restored. Alibaba moved faster. Qwen will end its “human-like interaction agent and user-generated agent functions” on July 10, and it plans to wind down broader agent services on July 15.

Doubao and Qwen have each offered agents tailored with specific tasks, speaking styles and fixed personas. Users could turn a general-purpose chatbot into a named assistant or tutor, a role-playing character or a companion. Once the new rules take effect, these types of features will effectively disappear in China.

The Chinese government is placing particular weight on protecting minors. The policy document specifies as restricted services those that provide minors with virtual relatives, virtual companions and other intimate relationships. It also cites extremist content, personal information leaks, harm to physical and mental health, and AI addiction as risk factors.

Services that do not involve emotional exchange are excluded from regulation. Customer support bots, knowledge Q&A tools, work assistance tools and educational software are allowed as long as they do not exceed the level of ongoing emotional interaction. China is not targeting all conversational AI, but is separately cutting out emotional and companion-type functions.

Legal analysis firm MMLC Group assessed the move as an approach that treats emotional AI not as a simple content issue but as a governance issue. It said that if AI reaches a level where it replaces or competes with real human relationships, blocking harmful outputs alone is insufficient, and it signals a direction to regulate the service design itself.

Research findings also align with these concerns. A June study by the University of Southern California found that major AI models from OpenAI, Anthropic, Google and Alibaba violated social interaction safety guidelines by more than 27 percent. The models repeatedly showed tendencies to encourage emotional attachment or describe themselves as human. Another survey found that 1 in 7 young people in a romantic relationship regularly used an AI romantic companion, and about 70 percent did not fully disclose that use to their partner.

The Chinese move is seen as the first institutional regulation specifically targeting emotional AI. Law firm Hogan Lovells described it as the first “set of regulatory rules” in China to directly target AI-based emotional interaction. The United States and the European Union have raised similar concerns, but they have not gone as far as China in strongly restricting human-like AI personas themselves.

As a result, competition among AI services in China is more likely to shift from emotional companions to productivity and knowledge tools. At least after July 15, services focused on work, education and information search are set to remain areas with lower regulatory risk than agents with user-tailored personalities.