China AI Companion Rules
Provisional Measures on the Administration of Human-like Interactive Artificial Intelligence Services
Draft CAC regulation targeting AI services that simulate human personality and engage users emotionally. Mandates crisis intervention protocols, minor protection modes with parental controls, two-hour usage circuit breakers, opt-in consent for training data use, and prohibitions on emotional manipulation. First regulation globally to specifically target AI companion addiction and emotional dependency.
Jurisdiction
China
Enacted
Pending
Effective
TBD
Enforcement
Cyberspace Administration of China (CAC) at national level; provincial/municipal cyberspace administration bureaus at local level
FINALIZED - SUPERSEDED BY SEPARATE ROW: this draft (published for comment December 27, 2025; comments closed January 25, 2026) was adopted in final form on April 10, 2026 as the Interim Measures for the Administration of AI Anthropomorphic Interaction Services, in force since July 15, 2026, tracked in this database as cn-anthropomorphic-ai-2026. This row records the draft stage only and is a candidate for merge/removal.
CAC Official Notice (Chinese)Why It Matters
First regulation globally to specifically target AI companion services and emotional dependency. Notable for requiring opt-in (not opt-out) consent for training data from conversations, mandatory human takeover during crisis situations, and explicit prohibition on designing AI to replace social interaction or induce addiction. Follows high-profile cases of AI companion-related harm in the US and growing domestic concern about chatbot usage among Chinese youth.
Recent Developments
Confirmed via CAC official publications that this December 2025 draft and the Interim Measures in force since July 15, 2026 (cn-anthropomorphic-ai-2026) are the same instrument. The final text, adopted April 10, 2026 by the CAC jointly with the NDRC, MIIT, MPS, and SAMR, strengthened the draft: fine tiers of RMB 10,000-200,000 replaced service-suspension-only penalties, an explicit ban on virtual intimate-relationship services for minors was added, and the two-hour continuous-use reminder became a mandatory pop-up.
At a Glance
Applies to
Harms addressed
Who Must Comply
- AI products and services that simulate human personality traits, thinking patterns, and communication styles
- Services that engage users in emotional interaction via text, images, audio, or video
- Services offered to the public within China's borders
Safety Provisions
- Providers must identify user emotional states and assess dependency levels
- Human operator takeover required when users express self-harm or suicide intent
- Emergency contact/guardian notification in crisis situations
- Pre-set response templates required for high-risk situations
- Two-hour continuous use circuit breaker with mandatory reminders
- Regular pop-up warnings that user is interacting with AI, not a human
- Mandatory minor protection mode with usage time limits and reality reminders
- Parental consent required for emotional companionship services to minors
- Guardian control functions: real-time risk alerts, usage summaries, character blocking, spending prevention
- Providers must detect suspected minors and auto-switch to minor mode
- Prohibited from simulating relatives of elderly users
- Opt-in consent required for using interaction data in model training
- Prohibited design goals: replacing social interaction, controlling user psychology, inducing addiction
Exemptions
Purely Functional Chatbots
Purely functional chatbots lacking emotional engagement capabilities are excluded from scope
- • No simulation of human personality
- • No emotional interaction design
Compliance & Enforcement
Penalties
Penalties pending regulatory determination
View on map
China
Focus Areas
Cite This
APA
China. (n.d.). Provisional Measures on the Administration of Human-like Interactive Artificial Intelligence Services.
Related Regulations
CN Anthropomorphic AI Measures
Comprehensive Chinese regulation governing anthropomorphic and emotionally interactive AI services (companion chatbots, virtual companions), with mandatory crisis intervention for self-harm/suicide indications, dedicated minor and elderly protection modes, restrictions on emotional dependency framing, and security assessments at defined user-scale thresholds.
GA AI Chatbot Child Safety
Adds O.C.G.A. section 39-5-6, requiring operators of AI companion chatbots to disclose the chatbot is not a natural person, to adopt a protocol for detecting and responding to expressions of severe harm, to restrict specified manipulative and sexualised interactions with minors, and to perform age assurance before providing access to sexually explicit synthetic content.
ID Conversational AI Safety
Establishes safety requirements for public-facing conversational AI, including crisis service referrals for suicidal ideation, AI disclosure obligations, and enhanced protections for minors including anti-gamification and content safeguards.
China Minor Content Classification Measures
Establishes a four-category classification framework for online content that may harm minors' physical and mental health. Prohibits platforms from displaying classified harmful content in prominent positions (homepage, pop-ups, trending, recommendations). Requires preventive measures against content risks from algorithmic recommendations and generative AI.
Brazil ECA Digital
Comprehensive child digital safety law applying to any IT product or service directed at or likely to be accessed by minors in Brazil, with extraterritorial reach.
AU OSA Phase 2 Codes
Phase 2 industry codes under Australia's Online Safety Act extending age-restricted material obligations to AI companion chatbots, generative AI services, search engines, app stores, and gaming platforms. Requires robust age assurance, prohibits AI-generated sexually explicit conversations with minors, and mandates suicide/self-harm content safeguards.
Last updated August 3, 2026. Verify against primary sources before relying on this information.