China's Interim Measures for Artificial Intelligence Anthropomorphic Interactive Services
Issued by
Cyberspace Administration of China, National Development and Reform Commission, Ministry of Industry and Information Technology, Ministry of Public Security, State Administration for Market Regulation
- September 30, 2026 · Correction — Rewrote the entry from the final Measures: they cover emotional-companion AI, not task agents, and use security assessments and algorithm filing rather than a pre-approval tier (fact-check finding). (Cody Maxwell)
- October 1, 2026 · Correction — Rewrote the practical steps, newsletter hook, search description, and audience fields to match the entry's corrected content. (Cody Maxwell)
China's Interim Measures for AI Anthropomorphic Interactive Services took effect on 15 July 2026. They cover AI services offering continuous emotional interaction that imitates a person's personality, thinking, and communication style. Providers must avoid fostering emotional dependence, protect minors and older users, and complete assessments and filing.
Applies To
Overview
Five agencies issued these Measures, led by the Cyberspace Administration of China. The others are the National Development and Reform Commission, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration for Market Regulation. A draft was released on 27 December 2025. The final text was adopted on 2 February 2026, published on 10 April 2026, and took effect on 15 July 2026. They apply to AI services offered to the public in mainland China that provide continuous emotional interaction by imitating a person's personality traits, thinking patterns, and communication style. They target AI companions and similar services, not agents that perform tasks. Providers must not generate content that encourages self-harm, excessively caters to users to create emotional dependence, or manipulates emotions to push unreasonable decisions. They may not offer virtual relatives or partners to minors, and need a guardian's consent for children under 14. Older users get prominent safety warnings. Users must be told they are talking to AI, reminded after every two hours of continuous use, and able to exit easily. Providers must submit security assessment reports to provincial cyberspace authorities in set cases, such as launch or reaching 1 million registered users or 100,000 monthly active users. They must also file algorithms under the algorithm recommendation rules, with annual verification, and keep network logs as the law requires. Penalties include warnings, fines of 10,000 to 100,000 yuan, and suspension, rising to 200,000 yuan in serious cases.
Key Requirements
- •Applies to AI services in mainland China that offer continuous emotional interaction imitating a person.
- •Do not generate content that encourages self-harm, fosters emotional dependence, or manipulates users' decisions.
- •Do not offer virtual relatives or partners to minors, and get guardian consent for children under 14.
- •Tell users they are interacting with AI, remind them after every two hours of use, and let them exit easily.
- •Submit security assessments in set cases, such as launch or 1 million registered users, and file algorithms.
- •Keep network logs as the law requires; fines run from 10,000 to 200,000 yuan.
What Your Organization Must Do
- →Map whether your services offer continuous emotional interaction imitating a person to users in mainland China.
- →Schedule security assessment reports for launch and user thresholds of 1 million registered or 100,000 monthly active users.
- →File algorithms under the algorithm recommendation rules and diarise annual verification of those filings.
- →Build AI disclosure, two-hour use reminders, and easy exit paths into companion products.
- →Block virtual relative or partner features for minors and collect guardian consent for children under 14.
- →Review content controls against self-harm, emotional dependence, and manipulation of user decisions before service launch.
Playbook Guidance
Step-by-step implementation guidance for compliance teams.
Governance Controls
Operational controls that implement requirements from this regulation.
Frequently Asked Questions
- Does CN-AIAIS apply to foreign companies offering anthropomorphic AI services to Chinese users without a local entity?
- The Measures apply to services offered to the public in mainland China. They set no local representative duty, but foreign providers serving users there should expect the same assessment and filing duties.
- What filing and assessment steps apply to anthropomorphic AI services?
- The final Measures took effect on 15 July 2026. A security assessment is due at launch and in other set cases, and algorithms must be filed under the algorithm recommendation rules.
- What penalties can regulators impose on providers that fail to comply with CN-AIAIS?
- Regulators can issue warnings, impose fines of 10,000 to 100,000 yuan, or suspend services. Serious cases that threaten health or safety bring fines of 100,000 to 200,000 yuan.
- Is there a pre-approval system for anthropomorphic AI services?
- No, the Measures use security assessments, triggered by launch, major changes, or user thresholds, plus algorithm filing with annual verification. There is no pre-approval or licensing tier.
- What logging and record-keeping obligations apply to anthropomorphic AI services under the Measures?
- Providers must strengthen security monitoring and keep network logs as the law requires. There is no special real-time logging tier.
- How does CN-AIAIS compare to the EU AI Act in its treatment of AI systems that simulate human interaction?
- Both require telling users they are dealing with AI. China's Measures go further for emotional companion services, with protections for minors and older users and limits on fostering emotional dependence.