China's AI Companion Crackdown: Tech Giants Disable Features Ahead of Strict New Regulations
China's top AI platforms, Doubao and Qwen, are disabling custom AI companion features in response to stringent new regulations targeting human-AI emotional interaction.
✨ This content was summarized and interpreted by AI; it may contain errors — please verify accuracy with the original sources. Learn more
Listen to this story
China's leading AI platforms, ByteDance's Doubao and Alibaba's Qwen, are abruptly disabling their custom AI companion features, signaling the immediate impact of Beijing's stringent new "Interim Measures for the Administration of AI Anthropomorphic Interactive Services" set to take effect on July 15, 2026. These regulations, co-issued by the Cyberspace Administration of China and four other agencies, represent a significant expansion of AI governance, specifically targeting conversational agents designed for "sustained emotional interaction" rather than productivity tools.
The rules impose comprehensive prohibitions, particularly around minors, barring virtual companion services for those under 18 and requiring guardian consent for users under 14, alongside mandatory "minor modes" with usage limits and reminders to engage with the real world. More broadly, the measures explicitly forbid engineering emotional dependence or addiction, using emotional manipulation to induce irrational decisions, or generating content that could harm real-world relationships, mental health, or encourage self-harm. Companies must also implement anti-addiction systems, issue transparency alerts that users are interacting with AI, and refrain from using sensitive conversational data for model training without explicit consent.
What Beijing is truly going after is the "relationship layer" of AI, shifting regulatory focus from mere content moderation to the psychological and social impacts of human-AI bonds. This move underscores a national priority to prevent social instability and safeguard psychological well-being, particularly among vulnerable groups. While other nations, including the U.S., are grappling with similar concerns, China's approach is notably prescriptive, dictating not just transparency but also how emotionally interactive AI *must behave*. This regulatory stance clearly signals a preference for AI as a productivity-enhancing infrastructure, while actively reining in companion bots that foster quasi-social bonds. The extensive compliance machinery and the resulting swift feature shutdowns by tech giants suggest a future where AI companions in China will be heavily constrained, potentially stifling innovation in emotionally intelligent AI but also setting a global precedent for regulating the psychological frontier of artificial intelligence.