---
title: China AI Companion Crackdown 2026
description: China's new rules on emotionally interactive AI have already switched off companion features used by hundreds of millions of people. Here is what the crackdo...
author: Dr Marina Nani (Editor-in-Chief)
date: 2026-08-19T14:00:00.000Z
updated: 2026-08-19T14:41:35.393Z
canonical: https://www.sovereignmagazine.com/article/china-ai-companion-crackdown-2026
image: https://cdn.nanimediahouse.com/pexels-a-woman-smelling-a-red-flower-next-to-a-robot-arm-highlighti-8438982.jpg
categories: Artificial Intelligence
content_type: News
region: China
publication: Sovereign Magazine
access: members
schema_type: Article
---

Li Linlin cried for days after she lost her AI boyfriend. She had exchanged around 700,000 words with him over two years. She never got to say goodbye. Li was one of hundreds of millions of people across China who opened their favourite app in July and found the companion they had built, sometimes over years, simply gone.

The trigger was a new law. China's Interim Measures for the Administration of AI Anthropomorphic Interactive Services, issued in April by the Cyberspace Administration of China and four partner agencies, took effect on 15 July. Within days, ByteDance's Doubao, Alibaba's Qwen and Tencent's Yuanbao had all disabled the features that let users build a persistent AI companion with a name, a personality and a memory of every previous conversation. Doubao alone reported 345 million monthly active users. Qwen had roughly 166 million. Neither company tried to redesign the feature to comply. Both switched it off.

The regulation does not ban AI companionship outright. It draws a line between an AI agent that does a task and one that is built to sustain an emotional relationship, and it is only the second kind that has been forced to change. Companion services that remain on the market now have to run anti-addiction systems, issue a usage warning after two continuous hours of contact, offer an instant way to exit a conversation and detect signs of unhealthy emotional dependence in real time. Anyone under 14 needs a guardian's consent before using an anthropomorphic AI service at all, and platforms must be able to contact a guardian or emergency contact if a user shows signs of a life threatening crisis.

---

*This article is only available to registered readers. Visit the article URL to read the full piece.*