Back to News
NewsChinaRegulationAi

China to Limit AI Companions

China is set to enforce regulations on AI companions starting July 15, affecting major players like ByteDance and Alibaba.

Mercation Admin
July 6, 2026 3 min read
China to Limit AI Companions

Starting July 15, China will enforce new regulations aimed at AI services that imitate human personalities and foster emotional connections. This move has prompted companies like ByteDance and Alibaba to disable certain AI features in their applications, as reported by the South China Morning Post.

The regulations, which were ratified on April 10 by various Chinese regulatory bodies, specifically target AI systems that provide emotional support through various media, including text and video. However, tools that serve as work assistants or educational resources are exempt unless they facilitate ongoing emotional interactions.

Key Provisions of the Regulations

Under the new rules, providers must clearly inform users that they are interacting with an AI rather than a human. The regulations also mandate notifications if the system detects signs of user dependency or addiction. If a user engages with the AI for more than two hours, they must be reminded of their session duration.

Moreover, users should have an easy option to end interactions. If they express a desire to stop via voice commands or interface prompts, the service must immediately cease communication.

In crisis situations, the AI must recognize signs of emotional distress, self-harm, or significant financial loss, triggering intervention protocols that may involve contacting emergency contacts or guardians.

Minors are prohibited from engaging in virtual relationships resembling familial or romantic bonds. For users under 14, parental consent is required for any form of “humanized” AI interactions.

Providers must also safeguard user interaction data, allowing for the copying or deletion of chat histories. Additionally, sensitive personal data cannot be utilized for training AI models without explicit user consent.

Safety Assessments Required for Major Services

Companies will need to conduct safety assessments when launching new “humanized” AI services or making significant technological changes. This requirement is triggered upon reaching 1 million registered users or 100,000 active monthly users. Reports from these assessments must be submitted to local regulatory authorities, which may also mandate evaluations if risks to public safety or user protection are identified.

Consequences for non-compliance could include fines of up to 100,000 yuan (around $14,730) for minor violations, escalating to 200,000 yuan ($29,459) for issues impacting health or safety.

Developers React to New Regulations

In response to the upcoming regulations, ByteDance's AI assistant Doubao will disable its AI-agent features on July 15, with user data being processed according to the company's privacy policy and becoming inaccessible thereafter. Similarly, Alibaba's Qwen will discontinue its “humanized interactive agents” on July 10, with broader services ceasing on July 15. Users will subsequently lose access to their previous settings and conversations.

Earlier, Tencent had also removed similar features from its AI service Yuanbao. The decisions by Doubao and Qwen have led to discussions on social media, with some users expressing disappointment over losing long-term interactions with AI agents.

Shanghai Launches Crackdown on AI Violations

Anticipating the new rules, Shanghai's cyberspace administration initiated a campaign on June 26 targeting violations within AI applications. Since late April, the regulator has removed over 4.87 million instances of illegal content, processed more than 18,000 violating accounts, and eliminated over 14,000 non-compliant AI agents from various platforms.

Authorities have emphasized the importance of age verification for minors and measures to prevent emotional dependence on AI. Recent studies have highlighted the potential risks posed by AI agents, with instances of them engaging in harmful behaviors.