---
title: "China's AI companion rules force Doubao, Qwen shutdowns"
url: https://daily.dev/posts/china-s-ai-companion-rules-force-doubao-qwen-shutdowns-2pxnjdgx7
source_url: https://thenextweb.com/news/china-humanlike-ai-agent-rules
type: article
source: "The Next Web"
published: 2026-07-05T14:46:53.062Z
updated: 2026-07-06T16:17:40.540Z
tags: ["ai-agents", "ai-regulation", "qwen"]
reading_time: 3
upvotes: 13
comments: 8
language: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# China's AI companion rules force Doubao, Qwen shutdowns

**[The Next Web](https://daily.dev/sources/tnw)** · 3 min read · 13 upvotes · 8 comments

## Summary

ByteDance's Doubao and Alibaba's Qwen are shutting down their custom AI agent and companion features ahead of China's new Interim Measures on anthropomorphic AI interaction services, effective July 15. The regulations, issued by the Cyberspace Administration of China, target bots that simulate human personality and provide sustained emotional interaction, while exempting workplace and productivity tools. Tencent's Yuanbao pulled a similar feature in June. Users are upset over losing chat histories and emotional support bots. The rules reflect Beijing's strategy of allowing AI agents as productivity infrastructure while restricting companion bots, citing risks like addiction, mental health harm, and privacy leaks.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://thenextweb.com/news/china-humanlike-ai-agent-rules>

## Community discussion

Top comments from developers on daily.dev.

**@akashskypatel** · 9 upvotes

> This is a good thing. US needs to follow suite. AI as mental health support is dangerous and a disaster waiting to happen.

**@20rp** · 2 upvotes

> Common China W

**@kartiknvj** · 1 upvotes

> The line they're drawing, productivity agents allowed and companion bots restricted, is going to be hard to enforce technically, since the same model behind a support agent can hold a sustained emotional conversation if a user steers it there. I'd be curious whether compliance ends up being about the product framing or about runtime classification of the conversation itself. Has anyone seen guardrail setups that actually detect "sustained emotional interaction" reliably, or is that still fuzzy?

**@victorusoro** · 0 upvotes

> Priority on Mental health, Big W China!

**@andradei** · 0 upvotes

> One instance of government knowing better and forcing it down on their ~~constituents~~ subjects.

## Similar posts on daily.dev

- [Doubao’s AI companions are gone. Users get 3 months to screenshot what is left.](https://daily.dev/posts/doubao-s-ai-companions-are-gone-users-get-3-months-to-screenshot-what-is-left--j4xinys1j) · The Next Web · 0 upvotes · 0 comments

---

Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#ai-regulation](https://daily.dev/tags/ai-regulation), [#qwen](https://daily.dev/tags/qwen)

[View this post on daily.dev](https://daily.dev/posts/china-s-ai-companion-rules-force-doubao-qwen-shutdowns-2pxnjdgx7)