AI Regulation Tracker  /  Regulation effective

China's New Rules for Human-Like AI Companions Take Effect July 15, 2026

China has issued what law-firm analysts describe as China's first national rulebook for human-like AI, covering virtual companions and emotionally interactive chatbots. Firms that ship companion AI to users in China, including US firms, must file, disclose the AI's non-human nature, and build in anti-addiction and crisis safeguards.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

China has finalized the first national rulebook aimed squarely at AI that behaves like a person. Under the "Interim Measures for the Administration of Anthropomorphic AI Interaction Services," issued by the Cyberspace Administration of China with four other agencies, providers of AI companions and emotionally interactive chatbots face a defined set of duties that come into force on July 15, 2026. According to the CAC and law-firm analyses of the text, the measures reach well beyond China's borders in practice, because they apply to companies that serve users inside China regardless of where those companies are based.

What the measures cover

The rules target what the CAC calls anthropomorphic AI interaction services. In plain terms, that means AI virtual companions, chatbots built to simulate a personality, emotional and relationship digital assistants, and AI-driven characters in games. These are products designed to feel like a person rather than a tool, and that human-like quality is exactly what the measures regulate.

Providers of these services must, according to the CAC, complete algorithm filing and a security assessment before operating. They must clearly disclose the AI, non-human nature of the service so users understand they are interacting with software. And they must build in anti-addiction measures, such as usage reminders after prolonged sessions, to limit the compulsive engagement that companion products can drive. Companion products are designed to hold attention, and the measures treat that design goal as a risk to be managed rather than a feature to be left alone.

Safety and minor-protection duties

Two of the heaviest obligations concern user welfare. First, providers must offer self-harm and crisis-intervention pathways, so a user in distress is routed toward help rather than left alone with a chatbot. Second, the measures set specific protections for minors, including restrictions on intimate or romantic interactions and parental consent requirements for younger children.

The text also limits how companies use what people say to these systems. Under measures issued by the CAC, providers may not use user interaction data to train models without explicit consent. That consent standard matters because emotional and companion products generate unusually intimate conversation logs, the kind of data that many jurisdictions now treat as sensitive. For a provider, it means the default of quietly recycling chat logs into the next model version is off the table for the China market.

What it does not do

The measures are not a ban on companion AI, and they are not a blanket content code for all generative AI. They are a targeted regime for services built to act human. Reporting on the exact article structure and numbered provisions should be read with care, because the primary text is in Chinese and specific article numbers have not been independently confirmed here. The load-bearing facts are the scope of the duties and the July 15, 2026 effective date, both of which the CAC and secondary legal analyses support. Readers should also note a separate, parallel CAC policy on intelligent agents that takes effect the same day; that is a companion measure and is not the subject of this piece.

Why a US professional should care

For US firms, the cross-border angle is direct. A US company shipping a companion or emotional-support chatbot into the China market must file and comply, the same as a domestic provider. Beyond market access, the measures function as a de-facto reference point for companion-AI safety worldwide, and they preview the direction US rules are already moving. The disclosure, anti-addiction, and crisis-intervention duties in the Chinese text closely track the mental-health-chatbot laws that states including Nevada, Utah, and Rhode Island have been building. A compliance officer or product lawyer reading the CAC measures is, in effect, reading a draft of obligations that are converging across several jurisdictions at once. The specific mechanisms differ, but the underlying expectations, tell the user it is a machine, do not let the product become compulsive, and route people in crisis to real help, are becoming common ground.

The practical response is unglamorous but clear. Inventory any product that presents an AI as a companion or emotional presence, confirm filing and assessment status for the China market, and treat disclosure, usage limits, crisis routing, and minor protection as product requirements rather than policy footnotes.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.