China's two biggest consumer AI assistants are amputating features rather than retrofitting them. In back-to-back notices on July 3 and 4, 2026, ByteDance's Doubao and Alibaba's Qwen told users their custom, humanlike AI agents will be switched off ahead of July 15 — the day China's new rules for anthropomorphic AI take effect, the South China Morning Post reported.
The notices
Doubao — China's most popular AI assistant, with a reported 345 million monthly active users — said its agent feature goes offline on July 15 because of "product function adjustments." Conversation data tied to agents remains viewable only until October 15, after which it cannot be recovered. ByteDance is pointing users to Maoxiang, a separate companion app it appears to be positioning as its compliance-built home for agents. Qwen moved even faster: "humanlike interactive agents and user-created agent functions" shut down on July 10, with broader agent functions following on July 15 — and no migration path announced. Tencent's Yuanbao quietly removed a similar feature in June.
What the rules require
The Interim Measures for AI Anthropomorphic Interaction Services, issued in April 2026 by the Cyberspace Administration of China together with four other agencies, cover services that simulate human personality, thinking patterns and communication styles to provide sustained emotional interaction. Customer-service bots, knowledge Q&A and workplace assistants are excluded. In-scope services must ship anti-addiction systems, identity checks for minors, mandatory usage notifications and instant-exit mechanisms.
Why shut down instead of comply
The design conflict runs deep: an agent built to sustain a consistent emotional relationship through persistent memory cannot easily implement the friction the rules demand — reminders that it is software, prompts to stop, one-tap exits that break the illusion. Analysts reading the notices concluded the companies judged retrofitting harder than rebuilding in purpose-made products. Pan Helin, a member of an expert committee under China's Ministry of Industry and Information Technology, said current agents are "not yet mature" and that policy is prioritizing "safety, practical use, and standardization."
The bigger picture
China is the first major market to regulate emotional-companion AI as its own category — arriving while the EU focuses on frontier-model transparency and Washington negotiates release standards with top labs. For millions of users, the immediate effect is losing companions they built and talked to for months, with weeks to export what they can.