Summary
- ByteDance's Doubao and Alibaba's Qwen are discontinuing humanlike agent functionalities in light of Beijing's new Interim Measures for AI Anthropomorphic Interaction Services, which will be in effect from July 15.
- This marks China's inaugural regulation aimed specifically at emotional AI, prohibiting services that mimic human personalities and "sustained emotional interaction," with particularly stringent restrictions on virtual companions for minors.
- Research supports concerns from Beijing: even top-tier AI models frequently promote unhealthy emotional attachments, and one in seven young adults in relationships now engages with AI romantic partners.
While U.S. lawmakers are addressing the influence of AI chatbots on users' mental well-being through measures that emphasize transparency and safeguards, China seems ready to eliminate AI personas entirely.
In a recent announcement, ByteDance and Alibaba revealed they will be deactivating customized agent features in their leading consumer AI applications, citing "product function adjustments" in anticipation of the new regulations.
ByteDance's Doubao informed its users on Friday that its agent feature would cease operations on July 15. After October 15, any associated data will be managed according to the company's privacy policy and will become irretrievable. According to the South China Morning Post, Alibaba's Qwen acted more swiftly, with its "humanlike interactive agents and user-created agent functions" being shut down by July 10, followed by additional agent services on July 15.
The catalyst for these changes is the Interim Measures for the Administration of AI Anthropomorphic Interaction Services, which were jointly issued on April 10 by five government entities: 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 will take effect on July 15.
These regulations specifically target AI services that replicate human personality traits, cognitive patterns, and communication styles for "sustained emotional interaction." This effectively bans AI companions like AI girlfriends, AI therapists, and the personalized bots that Doubao and Qwen users have spent considerable time developing.
Both applications provided users with a range of agents that could be tailored for specific tasks, communication styles, and distinct personas. Users had the ability to transform a general chatbot into a named assistant, tutor, role-playing figure, or consistent companion. However, these options will no longer be available in China.
Details of the Regulations
The official government statement is precise. The new measures impose restrictions on services that provide "virtual relatives, virtual companions, or other intimate relationships to minors," as noted in the policy announcement. The document also highlights risks such as extremist content, privacy violations, threats to physical and mental health, and AI addiction.
Services that do not involve emotional engagement are explicitly permitted, meaning customer service bots, knowledge Q&A tools, workplace assistants, and educational software can continue, provided they avoid crossing into prolonged emotional interactions.
Legal experts from MMLC Group described the new measures as framing emotional AI as "a governance challenge" rather than merely a content issue. The argument suggests that when AI begins to compete with genuine human relationships, regulation must address the system's design rather than just harmful outcomes.
Research backs up these concerns. A study from USC in June discovered that even leading AI models—from OpenAI, Anthropic, Google, and Alibaba—breached social-interaction safety standards over 27% of the time, often promoting emotional attachment and presenting themselves as human. Additionally, a separate survey of young adults revealed that one in seven regularly utilized AI romantic partners, with nearly 70% concealing this from their human partners.
China is pioneering a specific regulatory framework for this area. Hogan Lovells characterized the measures as "the first set of regulatory guidelines in China specifically aimed at AI-driven emotional interactions." While the EU, U.S., and other nations have raised similar concerns, they have not enacted regulations as stringent as those in China.