ByteDance and Alibaba withdraw agent features as China cracks down on human-like AI

While US politicians address the impact of AI-powered chatbots on users’ mental health with a focus on limitations Transparency and GuaranteesBeijing appears ready to shut down AI characters entirely.

ByteDance and Alibaba both announced over the weekend they are disabling personalized agent features in their largest consumer AI products, citing “product function adjustments” before new rules governing these products take effect.

ByteDance’s Doubao informed users in a notice on Friday night that its proxy feature will go offline on July 15. After October 15, the relevant data will be treated under the company’s privacy policy and become non-refundable. Per South China Morning PostAlibaba’s Qwen moved even faster: “human-like interactive agents and user-created agent functions” landed on July 10, with broader agent services on July 15.

This was prompted by China’s Interim Measures for the Administration of Holographic Interaction Services for AI, jointly issued on April 10 by five government departments – 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 rules are in effect July 15.

The regulation targets artificial intelligence services that mimic human personality traits, thinking patterns, and communication styles for “sustainable emotional interaction.” Translation: AI girlfriends, AI healers, AI companions, and custom character bots that Doubao and Qwen users spent months building are out of commission.

Both applications offered groups of agents that could be customized with specific tasks, speaking styles, and fixed personalities. Users can turn a general-purpose chatbot into a named assistant, tutor, role-playing character, or companion with a consistent tone. All of that has now disappeared in China.

What the rules actually say

Official government description specified. The measures impose restrictions on services that offer “virtual relatives, virtual companions, or other intimate relationships to minors,” according to the policy announcement. The document also notes risks including extremist content, privacy leaks, harm to physical and mental health, and AI addiction.

Non-emotional services are explicitly excluded, so customer service bots, cognitive Q&A tools, workplace assistants, and educational software are fine, as long as they don’t turn into sustained emotional engagement.

Legal analysts in MLC Group It described the measures as treating emotional AI as a “governance problem” rather than simply a content issue. The argument goes that once AI begins to compete with real human social connections, regulation should target system design, not just harmful outputs.

Research supports this concern. A USC study from June It found that even leading AI models—such as OpenAI, Anthropic, Google, and Alibaba—violated social interaction safety guidelines in more than 27% of cases, routinely encouraging emotional attachment and portraying themselves as human. separate Survey of partnered young adults It found one in seven romantic partners used AI regularly, and nearly 70% were hiding the full extent from their partners.

China is the first country to build a regulatory framework dedicated to this category. Hogan Lovells described These measures are “China’s first set of regulations specifically targeting AI-based emotional interaction.” The European Union, the United States and other countries have expressed similar concerns, but have not legislated in the same restrictive manner.