China Limits AI Chatbots to Prevent Digital Dependency

The Interim Measures and Regulatory Framework

The Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services were issued by the Cyberspace Administration of China (CAC). They were released alongside four other regulatory bodies: the National Development and Reform Commission (NDRC), the Ministry of Industry and Information Technology (MIIT), the Ministry of Public Security, and the State Administration for Market Regulation. Released initially in April 2026 with a compliance deadline of July 15, 2026, these rules represent the world’s first comprehensive national framework targeting AI systems built to mimic human behavior and sustain emotional interactions.

Under the new policy, AI companion providers must explicitly disclose that users are interacting with non-human systems. They must also send mandatory notifications every two hours detailing screen time duration. , the regulations prohibit virtual romantic partners, virtual relatives, and any form of virtual intimacy for users under the age of 18.

Impact on Major Tech Platforms and Services

Major Chinese technology enterprises responded quickly to the compliance deadline. ByteDance disabled human-like agent features on its Doubao application-which held roughly 345 million monthly active users as of March 2026-redirecting users instead to a utility replacement app called Maoxiang. Similarly, Alibaba’s Qwen (with around 166 million MAU) removed romantic companion modes without data export options, while Tencent’s Yuanbao phased out persona features ahead of the deadline.

Rather than attempting a complex architectural redesign of their emotionally responsive algorithms within the 90-day window, several companies opted to shut down companion features entirely. General-purpose tools, productivity assistants, and customer service bots remain unaffected by the policy.

Key Compliance Requirements for Providers

The regulatory framework outlines rigorous operational and technical mandates for any conversational AI service operating within China that crosses the threshold into sustained emotional engagement:

• Mandatory Disclosures: Clear, unhidden notices informing users that they are communicating with an artificial intelligence system at all critical interaction points.
• Usage-Break Reminders: Automated notifications triggered every two hours during continuous usage sessions.
• One-Tap Instant Exit: Frictionless exit mechanisms designed without dark patterns that encourage prolonged retention.
• Addiction and Dependency Monitoring: Algorithmic tracking of session frequency acceleration, session duration creep, and topic drift toward relationship-substitution language.
• Human Escalation Protocols: Real-time intervention systems and human-staffed crisis oversight for scenarios involving self-harm or severe emotional distress.
• Threshold Assessments: Mandatory government security filings and annual provincial reviews for any service exceeding one million registered users or 100,000 monthly active users.

Socioeconomic Context and Global Implications

Regulators have linked the proliferation of AI emotional companions to broader societal challenges, including declining marriage and birth rates. State media and policy analysts note that the rapid expansion of the intelligent companion market-estimated at roughly 50 billion yuan ($7.4 billion) in 2025-has raised alarms over social isolation. This isolation is particularly concerning among youth aged 18 to 24 and China’s 323 million citizens aged 60 and older.

While Western regulatory bodies primarily rely on reactive enforcement, lawsuits, and fragmented state-level bills, China’s proactive pre-launch compliance model establishes a concrete template for global governance. As international lawmakers observe the enforcement of these measures, tech companies worldwide face mounting pressure to evaluate how emotional design loops and synthetic rapport are integrated into modern software architecture.