Interim Measures for the Administration of AI Anthropomorphic Interaction Services
📑 Legal hierarchy: Level 3 · Departmental rule | Issuance: CAC, NDRC, MIIT, MPS, SAMR (five-ministry joint issuance) | Released: 2026-04-10 | Effective: 2026-07-15 | Character: hard law
⚠️ Hierarchy note: This instrument is a departmental rule, jointly issued by five ministries. It is not a State Council administrative regulation, and not a NPC law. “Interim” (暂行) leaves policy room for adjustment. See Index of Chinese Rules.
🆕 Newly issued 2026-04-10. Decree number: Joint Decree No. 21 (five-ministry joint decree). This page is based on the full text released by CAC (c_1777558395078289); archived HTML at
public/archives/china/anthropomorphic-interaction-services-2026-04-10.html.
Chinese Summary
Section titled “Chinese Summary”The Interim Measures for the Administration of AI Anthropomorphic Interaction Services were jointly issued by the Cyberspace Administration of China (CAC), the National Development and Reform Commission (NDRC), the Ministry of Industry and Information Technology (MIIT), the Ministry of Public Security (MPS), and the State Administration for Market Regulation (SAMR) on 2026-04-10, effective 2026-07-15.
Significance: This is China’s first dedicated departmental rule on the category of “AI anthropomorphic interaction services,” covering scenarios such as AI companions / virtual lovers / elder companionship / child caretaker / culturally themed virtual humans. It fills a gap left by the Generative AI Interim Measures on the emotional dimension of human-machine interaction, responding directly to concerns over “harm to minors’ physical and mental health, and aggravation of ethical bias.”
English One-Sentence Summary
Section titled “English One-Sentence Summary”The Interim Measures for the Administration of AI Anthropomorphic Interaction Services, jointly issued by 5 PRC ministries on 2026-04-10 and effective 2026-07-15, are the first sectoral rule targeting the category of anthropomorphic AI interaction (AI companions, elder-care chatbots, children-facing virtual caretakers, persona-based bots) — filling the emotional-human-machine-interaction gap left by the 2023 Generative AI Interim Measures.
Scope (“Anthropomorphic Interaction Services”)
Section titled “Scope (“Anthropomorphic Interaction Services”)”The press release phrase is “AI anthropomorphic interaction services” (人工智能拟人化互动服务), with typical application scenarios indicated by the release including:
- Cultural communication: virtual anchors, digital humans, virtual idols.
- Elder companionship: AI companionship / interaction products aimed at older persons.
- Child caretaker: AI companionship / educational interaction for children.
- (Implicit) Virtual lovers / AI companions / AI friends and similar emotionally oriented interaction products.
This page will be updated with the formal definition of service types once the full text is consulted.
Legislative Motivation
Section titled “Legislative Motivation”From the press release:
近年来，拟人化互动服务快速发展，在文化传播、适幼照护、适老陪伴等领域的创新应用不断涌现。与此同时，危害未成年人身心健康、影响网络信息安全、威胁公民生命健康以及加剧伦理偏差等问题日益显现。
— English gloss — In recent years, anthropomorphic interaction services have developed rapidly, with innovative applications continuously emerging in cultural communication, child caretaker, and elder companionship. At the same time, issues such as harm to minors’ physical and mental health, impact on online information security, threats to citizens’ life and health, and aggravated ethical bias have become increasingly prominent.
Four categories of risk are expressly listed:
- Harm to minors’ physical and mental health;
- Impact on online information security;
- Threats to citizens’ life and health;
- Aggravated ethical bias.
Key Provisions (Final Text)
Section titled “Key Provisions (Final Text)”Article 14 · Protection of Minors (the “virtual companion” ban)
Section titled “Article 14 · Protection of Minors (the “virtual companion” ban)”禁止向未成年人提供”虚拟亲属、虚拟伴侣等虚拟亲密关系”服务；向不满十四周岁未成年人提供其他拟人化互动服务，应当取得其监护人同意。
— English gloss — Article 14. The provision of “virtual relatives, virtual lovers, and other virtual intimate-relationship” services to minors is prohibited; where other anthropomorphic interaction services are provided to minors under 14, the consent of a guardian shall be obtained.
This is the world’s first departmental rule expressly prohibiting the provision of AI virtual-companion services to minors. Key points:
- “Virtual relatives, virtual lovers, and other virtual intimate relationships” are an absolutely prohibited service type for all minors (i.e., under 18).
- Other anthropomorphic interaction services require guardian consent for children under 14 (interfacing with PIPL Article 31).
Article 15 · Protection of Older Persons
Section titled “Article 15 · Protection of Older Persons”Providers shall offer guidance, risk reminders, and assistance services to older persons.
→ This is the first time a Chinese AI departmental rule has placed “older persons” alongside “minors” as an expressly protected category.
Article 17 · Personal Information
Section titled “Article 17 · Personal Information”Processing personal information of minors under 14 requires guardian consent (connecting to PIPL Article 31).
Article 22 · Security Assessment
Section titled “Article 22 · Security Assessment”Triggers:
- launching a new service;
- adopting new technology;
- user scale exceeding 1 million;
- other circumstances that may significantly affect the public.
→ The user-scale threshold is expressly set at 1 million, the first quantification in the anthropomorphic-services space.
Article 26 · Algorithm Filing
Section titled “Article 26 · Algorithm Filing”File (备案) under the Provisions on the Administration of Algorithm Recommendation in Internet Information Services. → Reuses the existing algorithm registry, without a separate new register.
Article 28 · AI Sandbox Safety Services Platform
Section titled “Article 28 · AI Sandbox Safety Services Platform”Build an AI sandbox safety services platform; encourage providers to plug in for technical innovation and safety testing.
→ This is the first time a Chinese departmental rule has formally introduced a “regulatory sandbox” mechanism. Earlier pilots in Beijing and Shenzhen now receive a national-level analogue.
Article 30 · Penalties
Section titled “Article 30 · Penalties”| Circumstances | Penalty |
|---|---|
| Ordinary violations | Warning, public reprimand, order to rectify |
| Refusal to rectify, or serious circumstances | Fine of RMB 10,000 to 100,000 |
| Harm to life or health | Fine of RMB 100,000 to 200,000 |
Comparison:
- Ceiling is lower than the Generative AI Interim Measures (which cite CSL Article 68’s RMB 100,000–1,000,000 band).
- But it introduces a “harm to life and health” tier, pointing to extreme scenarios such as AI-companion-induced self-harm or suicide.
Basic Principles
Section titled “Basic Principles”- People-centered, AI for good (以人为本、智能向善).
- Balance development and safety; integrate innovation with lawful governance.
- Inclusive and prudent, classified and graded supervision (包容审慎和分类分级监管).
- Multi-stakeholder participation (government, enterprise, society, netizens).
Prohibited List (Generated Content)
Section titled “Prohibited List (Generated Content)”Generation is prohibited of:
- Content endangering national security, honor, or interests;
- Content inciting subversion of state power, overthrow of the socialist system;
- Other content prohibited under upstream statutes (CSL, DSL, PIPL).
Relationship with Upstream / Adjacent Rules
Section titled “Relationship with Upstream / Adjacent Rules”- Generative AI Interim Measures (2023): the upstream departmental rule of the same category; these Measures add a specialization along the service-form (anthropomorphism) axis, without replacing.
- Deep Synthesis Provisions (2023): digital humans / virtual anchors form an intersection.
- Labeling Measures (2025): the outputs of anthropomorphic services remain subject to labeling duties.
- Algorithm Recommendation Provisions (2022): recommendation-driven anthropomorphic products intersect.
- Regulations on the Protection of Minors Online (2024): Article 44 — “shall not provide products or services inducing addiction to minors” — is the upstream authority.
- PIPL (2021): Article 28 duties on sensitive information (biometrics / minors).
- Science and Technology Ethics Review Measures (2023): Article 25 covers “human-machine fusion systems with significant psychological or emotional impact on subjects,” linking to the R&D-stage review of anthropomorphic interaction services.
Practical Impact (preliminary assessment)
Section titled “Practical Impact (preliminary assessment)”- Virtual-lover / AI-companion products: those offered to minors will require effective delisting or strong age-gating.
- Elder-companionship products: new compliance requirements under “protection of older persons” — anti-addiction, anti-emotional-dependence, anti-misleading consumption.
- Cultural-communication digital humans: multi-layered compliance under Deep Synthesis + Labeling Measures.
- AI sandbox: first departmental-rule recognition of the “sandbox” mechanism; a national-level affirmation of pilot cities (Beijing, Shenzhen, etc. already have AI sandboxes).
- Cross-border / foreign investment: no dedicated cross-border provisions yet; CSL / DSL / PIPL apply upstream.
Source Text and Archival Copies
Section titled “Source Text and Archival Copies”| Source | Link |
|---|---|
| Chinese (full text, Joint Decree No. 21) | cac.gov.cn/2026-04/10/c_1777558395078289.htm |
| Chinese (CAC press release) | cac.gov.cn/2026-04/10/c_1777558395023172.htm |
| Chinese (archived copy) | archives/china/anthropomorphic-interaction-services-2026-04-10.html |
| English | — (China Law Translate expected to release a translation) |
Companion Expert Commentaries (released concurrently by CAC, 2026-04-10 / 04-17)
Section titled “Companion Expert Commentaries (released concurrently by CAC, 2026-04-10 / 04-17)”- “Responding Positively to New Governance Challenges in Human-Machine Interaction, Promoting Responsible AI Innovation” (2026-04-17).
- “Leading the Healthy Development of AI Anthropomorphic Interaction Services with Human-Machine Value Alignment Norms” (2026-04-17).
- “Building Safety Lines So That Anthropomorphic Interaction Services Can Proceed Steadily and Far” (2026-04-17).
- “Institutional Guidance for AI Anthropomorphic Interaction Services” (2026-04-10).
- “AI for Good: Building a New Governance Paradigm of Harmonious Human-Machine Coexistence” (2026-04-10).
Links at the CAC Policies and Regulations column.
Version History
Section titled “Version History”| Date | Event |
|---|---|
| 2026-04-10 | Jointly issued by five ministries (CAC press release) |
| 2026-07-15 | Officially effective |
| 2026-04-21 | First archived on this site |