ByteDance’s Doubao and Alibaba’s Qwen are disabling their custom AI agent features ahead of China’s Interim Measures on Anthropomorphic AI Interaction Services, which take effect July 15. The regulations are the world’s first rules specifically targeting AI-driven emotional interaction, and they draw a hard line between companion bots and productivity tools.
Doubao notified users on Friday that its agent feature would go offline on July 15, citing “product function adjustments,” according to the South China Morning Post. Related data will stop being viewable or recoverable inside the app after October 15. Qwen followed on Saturday, announcing that humanlike interactive agents and user-created agent functions would be disabled on July 10, with broader agent functions shutting down by July 15.
Both apps had let users build named assistants, tutors, role-playing characters, and companions with fixed personas and speaking styles. Tencent pulled a similar feature from its Yuanbao assistant in June, and Chinese state media confirmed the shutdowns are about regulatory compliance.
What the Rules Cover
The Interim Measures were issued on April 10, 2026 by the Cyberspace Administration of China and four other agencies. According to Hogan Lovells’ legal analysis, they cover services that “utilize AI technologies to simulate human personality traits, thinking patterns, and communication styles for the purpose of providing continuous emotional interaction.” The determination rests on three elements: simulation of human-like personality traits, continuity of interaction, and the presence of emotional engagement.
Customer service bots, knowledge Q&A systems, workplace assistants, and education and research tools are explicitly excluded, provided they avoid sustained emotional interaction. The rules cite risks spanning extremist content, privacy leaks, harm to mental health, and addiction. They require anti-addiction systems and identity verification for minors.
Services that reach more than 1 million registered users or 100,000 monthly active users must complete a security assessment and submit it to regulators, per the Hogan Lovells analysis.
Companions Out, Workers In
Beijing is targeting companion bots specifically, while actively promoting agents as enterprise infrastructure. In May, three government agencies issued implementation guidelines to promote AI agent development, and in June, China released national cybersecurity standards covering agent identity, discovery, interaction, tool use, and lifecycle security. The TC260 standards require agents to undergo security assessments before deployment, enforce least-privilege access, maintain comprehensive audit logs, and ensure secure data erasure at decommissioning.
“Current agents are not yet mature,” Pan Helin, a member of an expert committee at China’s Ministry of Industry and Information Technology, told the SCMP. He said the policy prioritizes safety, practical use, and standardization.
The Global Times reported that Shanghai’s Internet Information Office removed more than 14,000 non-compliant AI agents from platforms in late June, including agents facilitating sports gambling on MiniMax. Liu Dingding, a veteran internet analyst, told the Global Times the shutdowns also reflect platforms reducing investment in businesses with limited commercial viability.
The Regulatory Split
The pattern is deliberate: squeeze companion bots that form quasi-social bonds with users while accelerating agents as enterprise infrastructure. China’s enterprise AI agent market is expected to surpass 48 billion yuan ($7.07 billion), according to figures cited by Global Times from China Development and Reform Press.
Users mourned the shutdowns openly on Weibo. One poster called their agents long-standing emotional support and lamented the lack of any easy way to export conversation histories, according to TNW.
For agent builders outside China, the signal is worth watching. No other jurisdiction has drawn this specific a line between emotional AI interaction and productivity AI interaction. The EU AI Act classifies emotion-recognition AI as high-risk but does not ban companion features outright. Beijing chose to regulate first and sort out the product implications later.