🇨🇳 China: New Anthropomorphic AI Regulation Draft (Dec 2025)

[gloomcheng]

Summary

China's Cyberspace Administration (CAC) released a new draft regulation on December 27, 2025:

《人工智能擬人化互動服務管理暫行辦法》
(Interim Measures for the Management of AI Anthropomorphic Interactive Services)

This regulation specifically targets AI that simulates human personality, emotions, and communication styles — including emotional companion AI and chatbots.

Key Requirements

Requirement	Details
Transparency	Must clearly indicate user is interacting with AI, not human
Usage limits	Popup reminder required after 2 hours continuous use
Emotional monitoring	Must detect user extreme emotions and addiction tendencies
Emergency intervention	Human takeover required for suicide/self-harm situations; contact guardians
Minor protection	Mandatory minor mode, parental controls, guardian consent required
Elderly protection	Prohibited from simulating elderly users' relatives
Exit mechanism	Must provide easy exit; cannot block users from leaving
Data protection	User interaction data cannot be used for training without consent
Registration threshold	Safety assessment required at 1M users or 100K MAU

Comparison with EU AI Act

Aspect	EU AI Act	China Anthropomorphic AI Regulation
Scope	All AI systems (risk-based)	Only "anthropomorphic interactive AI"
Focus	Transparency, accountability	Emotional manipulation prevention, mental health
Usage limits	None	2-hour reminder mandatory
Elderly protection	No specific rules	Prohibits simulating user's relatives

Updates Made

COMPLIANCE.md has been updated in commit 2a69f15 with:

• Detailed requirements table
• Prohibited activities from Article 7
• AII implementation recommendations
• Source reference

Discussion

• Are there additional requirements we should highlight?
• How does this compare to other jurisdictions' approaches to emotional AI?
• Any corrections or clarifications needed?

---

This issue documents a regulatory update. The initial summary was prepared with AI assistance.

/cc @gloomcheng