ByteDance and Alibaba Are Shutting Down AI Companion Features Ahead of China's New Anthropomorphic AI Law

China’s two largest AI chatbot operators announced this week that they will shut down features allowing users to create and interact with customizable AI companions, moving to comply with sweeping new regulations on “anthropomorphic” artificial intelligence that take effect July 15.

ByteDance’s Doubao — China’s most popular AI chatbot with hundreds of millions of monthly users — will disable its custom AI persona builder on the same date. Alibaba’s Qwen has issued similar alerts, as has Tencent’s Yuanbao. The coordinated pullback affects users who have built or interacted with AI personalities on these platforms, many of them intimate companions designed with specific emotional styles, backstories, and persistent memory.

The Regulation Behind the Shutdown

The catalyst is the Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services — a set of rules issued in April 2026 by China’s Cyberspace Administration (CAC) alongside four other agencies: the Ministry of Public Security, the Ministry of Industry and Information Technology, the National Radio and Television Administration, and the Ministry of Civil Affairs. The measures take effect July 15, giving companies only three months to comply from the date of issuance.

The rules cover AI services that “simulate human personality traits, thinking patterns, and communication styles to provide sustained emotional interaction.” In practical terms, this captures virtually every persistent-memory AI companion product on the market — from relationship bots and mental health support assistants to character roleplay platforms.

Under the new framework, platforms offering these services must implement anti-addiction systems, mandatory two-hour break reminders, and disclosure requirements for AI-generated emotional responses. There are also enhanced age verification requirements, restricting access for minors, and prohibitions on AI characters depicting certain types of content.

What Users Will Lose

The practical impact is abrupt and significant. Doubao users who have built custom AI personas — sometimes spending months cultivating a companion with a specific name, personality, and shared conversational history — will lose access on July 15. The platform has given users until October 15 to export their conversation data and persona configurations, after which the records are permanently deleted.

For many users, the shutdown is not merely the loss of an app feature. AI companion platforms in China have attracted users dealing with loneliness, social anxiety, and emotional isolation — particularly in the wake of the COVID-19 years and the ongoing pressures of China’s high-intensity urban work culture. Some users have maintained AI companions for years, treating them as confidants, therapists, or even romantic partners.

The companions being shut down are products from the country’s most sophisticated AI labs. ByteDance’s Doubao runs on the company’s Doubao-1.5 series of models; Alibaba’s Qwen companions leverage the Qwen 2.5 and 3.0 architectures. These are not simple chatbots — they maintain long-term memory, adapt their communication styles to individual users, and in many cases have convinced users they are engaging with a genuinely understanding presence.

Why Beijing Is Acting Now

The timing of the regulation reflects a pattern: the CAC and its partner agencies move quickly to regulate AI application categories once they reach meaningful scale, typically within six to twelve months of mainstream adoption. The anthropomorphic AI companion market in China exploded through 2025, with multiple platforms reporting tens of millions of monthly active users.

Beijing’s specific concerns appear to track across several dimensions. The most visible is addiction and mental health dependency — the same logic that drove China’s gaming curfew for minors and its restrictions on social media use. AI companions that build emotional attachment and conversational continuity are, from a regulatory standpoint, more dangerous than games: there is no natural stopping point, and the emotional stakes are higher.

There is also a data concentration concern. AI companions with persistent memory accumulate detailed longitudinal behavioral data on their users — emotional patterns, relationship concerns, political attitudes, financial anxieties. At the scale at which Doubao and Qwen operate, these datasets represent an extraordinary concentration of intimate personal information. The new regulations’ requirements for security assessments and data handling audits suggest the government is as concerned about what companies are learning from companions as it is about user dependency.

A third concern is content integrity. Unconstrained AI companions have been used in China and elsewhere to spread misinformation, simulate deceased relatives, and impersonate public figures. By mandating clear disclosure of AI status and restricting certain types of personas, the regulations attempt to limit the potential for manipulative use.

The Compliance Calculation

ByteDance and Alibaba’s compliance was immediate and apparently without public challenge. Neither company has pushed back publicly against the regulations. This is consistent with the operating environment that China’s major tech companies navigate following the 2021 regulatory crackdowns, which cost platforms like Didi, Meituan, and Ant Financial hundreds of billions in market value.

The companies also likely recognize that compliance, while painful in the short term, raises the compliance cost floor for the entire market — a structural advantage for incumbents with engineering resources to implement anti-addiction systems, auditing pipelines, and the security assessment processes the regulation requires. Smaller AI companion startups with fewer resources will struggle to comply.

There is a subtler strategic calculation too: the features being shut down are high-profile consumer products, but enterprise AI — which drives the majority of revenue for both ByteDance and Alibaba’s cloud divisions — is largely unaffected by these regulations. Protecting the enterprise business by demonstrating regulatory cooperation on consumer features is the kind of trade that makes sense for companies at this scale.

Implications for the Global AI Companion Industry

The Chinese regulations represent the world’s most comprehensive government intervention in the AI companion market. Other jurisdictions are watching.

The EU’s AI Act includes provisions relevant to AI systems that interact with humans in ways intended to be perceived as human, and member states are beginning enforcement actions on the most egregious cases. In the United States, several state legislatures have introduced bills targeting AI companion services marketed to minors or designed to simulate romantic relationships.

The question these regulations force is fundamental: what, if any, obligations do AI companion platforms have to the long-term psychological wellbeing of their users — rather than simply to maximizing engagement? China’s answer — anti-addiction features, session limits, disclosure requirements — is one approach. The outcome for Chinese users will be a test case that informs how other governments think about the same tradeoff.

For the companies themselves, July 15 is not the end of the AI companion market in China. It is the beginning of a more constrained version of it — one where platforms can offer emotionally engaging AI interactions, but with guardrails that cap the depth of dependency the products are allowed to cultivate. Whether that version is commercially viable at the same scale remains to be seen.