China Issues First-Ever Rules for AI Companions to Curb Addiction
China has introduced its first dedicated regulatory framework for AI-powered companion services, imposing strict anti-addiction measures and mandatory protections for minors as the fast-growing industry faces its most significant compliance challenge to date. The Interim Measures for the Management of AI Human-Like Interaction Services, jointly issued by five government agencies including the Cyberspace Administration of China (CAC), took effect on July 15, 2026.
The regulations target products that use artificial intelligence to simulate human personality traits and provide ongoing emotional interaction. According to Caixin Global, the rules mandate screen-time limits, crisis interventions, and a prohibition on virtual intimate relationships for minors, putting guardrails on a sector that has drawn hundreds of millions of users.
What the Rules Require
The new measures impose sweeping requirements on AI companion service providers. Services must issue usage reminders every two hours and deploy pop-up alerts when signs of addiction are detected. For minors, the rules are even stricter: an absolute ban on virtual relatives or romantic partners, mandatory parental consent for children under 14, and a dedicated minor mode with usage time limits and parental oversight.
Safety assessments are required before launching services, upon reaching one million registered users or 100,000 monthly active users, or when significant changes occur. Data protection provisions mandate encryption, prohibit third-party sharing without consent, and allow users to copy or delete chat history. Sensitive personal information cannot be used for training without separate consent.
Violations can result in fines ranging from RMB 10,000 to 200,000 (approximately $1,400 to $28,000), suspension of services, and regulatory interviews.
Industry Response: Major Platforms Act Quickly
The regulations have already prompted swift action from China’s largest tech firms. On July 15, both Alibaba’s Qianwen assistant and ByteDance’s Doubao — which had 166 million and 345 million monthly active users respectively — disabled their persona-based chatbot features. As Caixin reported, the companies notified users on July 4 that they would discontinue services allowing users to build and chat with customizable AI characters.
Huo Hongyi, Director of the Asia-Pacific Social Innovation Research Institute, told Securities Times that the past “low-barrier, high-risk approaches are basically blocked off,” leaving only enterprises with safety thresholds, content review capabilities, and technical barriers. He noted that “split operations” — where persona features are moved to separate apps — may become industry standard.
Compliance Challenges for AI Toy Market
While major tech platforms moved quickly, the broader AI toy market appears less prepared. A Securities Times investigation of physical stores in Shanghai found that most AI toys targeting children lack anti-addiction controls and parental notification features. In one store, a “companion robot” marketed to users aged 15 and above had no usage limits or guardian monitoring functions whatsoever.
Niu Tengchang, founder of Qingxin Yichuang, told reporters that “compliance thresholds will become a competitive dividing line,” with companies that build safety and ethics into their underlying architecture gaining structural advantages. His company recently released Dino OS, an embodied character robotics platform designed with built-in safety guardrails including identity verification and content filtering.
Broader Regulatory Context
The AI companion rules are the latest addition to China’s increasingly comprehensive AI governance framework, which includes the Algorithm Recommendation Regulation (2022), Deep Synthesis Regulation (2023), Generative AI Regulation (2023), and the AI Ethical Review Measures (2026). The new measures supplement rather than replace existing regulations.
The urgency of the rules is underscored by a 2025 survey from the China Youth and Children Research Center, which found that over 20% of children showed a tendency to rely on AI instead of thinking independently, and over 20% preferred chatting with AI over real people.
Calls for Clearer Standards
Despite the comprehensive scope of the regulations, industry insiders are calling for more detailed implementation standards. Li Qingdu, founder of Zhuoyide, noted that determining what constitutes “induced addiction” remains technically challenging. The industry is exploring standardized test data set scoring mechanisms, but some AI models have already become sophisticated enough to evade detection.
Huo Hongyi suggested that the industry should establish a third-party certification mechanism for “emotional health assessment,” allowing professional institutions to evaluate whether product designs exhibit tendencies toward addiction inducement. He also called for leading companies to initiate industry self-regulation agreements, which could reduce regulatory trial-and-error costs and provide compliance benchmarks for smaller enterprises.
What’s Next
As the regulations take effect, the focus shifts to enforcement and implementation. Key questions remain: How will the CAC monitor compliance across thousands of products? Will overseas AI companion services targeting Chinese users face enforcement? And can automated systems reliably detect emotional manipulation and induced addiction?
The rules represent a significant step in China’s approach to AI governance, but their ultimate impact will depend on the detailed implementation standards that industry and regulators are now working to develop. The coming months will reveal whether the new framework strikes the right balance between innovation and protection.