Interim Measures for the Administration of Anthropomorphic AI Interaction Services

Summary

Regulates AI services that simulate a natural person's personality, thinking and communication style to provide continuing emotional interaction. It bans virtual intimate-relationship services, such as virtual relatives or partners, for minors, and requires guardian consent for under-14s and a minor mode. Providers must disclose that users are dealing with AI, remind them after every 2 hours of continuous use, guard against emotional dependence, intervene in self-harm crises, offer an easy exit and not train on sensitive interaction data without separate consent. Security assessments are due at launch, after major changes and at 1 million registered or 100,000 monthly active users.

Language

No English version is recorded for this record yet. The English title may be the register's own rendering, and the summary was written from the sources below; for the exact wording use the Chinese text.

Safety topics

Status notes

Adopted at the third CAC office meeting of 2026 on 2 Feb 2026 with the agreement of NDRC, MIIT, MPS and SAMR, promulgated on 10 Apr 2026 as Order No. 21, in force from 15 Jul 2026, and published in the State Council Gazette, 2026 No. 17. It has 32 articles in four chapters. Emotional interaction covers emotional care, companionship, support and similar services in text, images, audio or video; services such as intelligent customer service, knowledge Q&A, work assistants, education and scientific research that involve no continuing emotional interaction are excluded (Art. 2). Providers must file their algorithms under the algorithmic recommendation provisions (Art. 26). Violations are punished under existing laws and administrative regulations; where these have no provision, the authorities may warn, circulate criticism and order correction, and for refusal or serious cases order the service stopped and may impose a fine of RMB 10,000 to 100,000, or of RMB 100,000 to 200,000 where citizens' life or health is endangered with harmful consequences (Art. 30). It also provides for an AI sandbox security service platform for innovation and safety testing (Art. 28).

Sources

Open the official source https://www.cac.gov.cn/2026-04/10/c_1777558395078289.htm

Secondary source: https://www.gov.cn/gongbao/2026/issue_12806/202606/content_7072472.html

Local copy

Verification

History

1. 2026-10-14notes: The draft for comment was released 27 Dec 2025. Adopted at the CAC office meetin → Adopted at the third CAC office meeting of 2026 on 2 Feb 2026 with the agreement — Art. 30 penalties now carry their condition (only where no law or administrative regulation provides a penalty) and the 'may' of the fines. Art. 2's coverage and exclusions are given as in the text; algorithm filing (Art. 26) and the article count added. The draft date of 27 Dec 2025, which no stored file shows, was removed.2026-10-14-refresh-translation-fixes · Claude Code (claude-opus-5-5)
2. 2026-10-14summary: Regulates AI services that simulate human personality and give ongoing emotional → Regulates AI services that simulate a natural person's personality, thinking and — The scope now follows Art. 2: continuing emotional interaction (持续性的情感互动服务), not 'emotional companionship', which is only one of the services Art. 2 lists (care, companionship, support). The ban reads as in Art. 14, on virtual intimate-relationship services such as virtual relatives or partners. Algorithm filing and the article count moved to notes to keep the summary within 700 characters.2026-10-14-refresh-translation-fixes · Claude Code (claude-opus-5-5)

Related records

Generative AI Measures China · In force

Algorithmic Recommendation Provisions China · In force

Citation

Cyberspace Administration of China; National Development and Reform Commission; Ministry of Industry and Information Technology; Ministry of Public Security; State Administration for Market Regulation. Interim Measures for the Administration of Anthropomorphic AI Interaction Services [CAC/NDRC/MIIT/MPS/SAMR Order No. 21]. 2026-04-10. https://www.cac.gov.cn/2026-04/10/c_1777558395078289.htm. Status: In force; verified 2026-10-14 (level A). Local copy archive/CN-16/2026-10-11/official.html retrieved 2026-10-11 14:00:32 UTC, SHA-256 24cd9922b4490808c2500b3105a3f7817d7572d111aa790ab174b8f45d4e47db. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record CN-16.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.