China's AI companion rules: Beijing's real target

The concept of an AI companion may seem dystopian, yet it has become a recurring topic in broader discussions about the risks of generative AI. Essentially, it refers to a conversational agent designed to maintain a continuous, personal relationship with a user, equipped with memory and a consistent persona that ensures coherence across sessions.

Emotional attachment naturally arises from such design, and it has become a key selling point. Much of the usage involves casual role play or simply wanting something that remembers you; at the fringes, this category blurs into ordinary assistants. However, as more people in China began treating these bots as a form of emotional companion, Beijing has now decided that the practice requires regulation.

China’s AI companion regulations take effect on July 15, and in the days leading up to the deadline, the country’s two most popular consumer AI apps quietly disabled the core features. ByteDance’s Doubao informed users that its agent function would go offline on July 15, citing “product function adjustments,” while Alibaba’s Qwen stated that its human-like and user-created agents would cease operating on July 10, with its broader agent services following five days later.

At first glance, it appears that China is shutting down AI agents. That is not the case. The regulations draw a distinction between agents that perform tasks and those that provide companionship, and Beijing has only targeted the latter.

The regulation is the Interim Measures for the Administration of AI Anthropomorphic Interactive Services, jointly issued on April 10, 2026, by the Cyberspace Administration of China and four partner agencies: the National Development and Reform Commission, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration for Market Regulation.

These measures cover services that simulate human personality traits, thought patterns, and communication styles to deliver ongoing emotional interaction. Customer service bots, knowledge Q&A, workplace assistants, and educational or research tools are exempt, as long as they avoid sustained emotional engagement. This is the first dedicated national framework of its kind, and it emerged from a public comment draft released late last year.

A design challenge, not a prohibition

Doubao and Qwen did not violate a direct ban. Instead, they ran into a design conflict. The regulations require companion services to implement anti-addiction systems, send mandatory usage notifications, and provide instant-exit mechanisms, along with real-time detection of unhealthy dependence.

These requirements conflict with agents designed to remember users, maintain consistency across sessions, and sustain an ongoing relationship. Rather than retrofit the feature, ByteDance decided to shut it down. Alibaba appears to have made a similar choice. ByteDance is now directing Doubao users to Maoxiang, a separate app where they can create agents again; Alibaba has not announced any equivalent migration path for Qwen. Tencent’s Yuanbao removed a comparable feature back in June.

The impact has fallen on users. Many expressed their grief openly on Weibo, with one user describing the agents as long-term emotional support and lamenting the difficulty of exporting chat histories. Doubao allows users to view their configurations and conversations in read-only mode until October 15 this year, after which the data will be processed according to its privacy policy and become unrecoverable. Qwen users have received no similar grace period, and their agent data is scheduled for permanent deletion.

What China’s AI companion regulations entail

The content of the regulations is more nuanced than a simple crackdown might imply. Providers are prohibited from offering virtual companion or virtual family member services to minors, and must obtain parental consent before serving users under 14. They are required to create dedicated “minor modes” with usage time limits, reminders to engage in real-world interactions, and enhanced parental controls.

They must also detect users in acute distress and intervene if someone shows signs of self-harm, suicidal behavior, or serious financial loss, escalating the situation to designated guardians or emergency contacts. Engineering emotional dependence or addiction, as well as using emotional manipulation to induce unreasonable decisions, are explicitly forbidden.

The compliance requirements are substantial. Services that introduce anthropomorphic features or exceed thresholds of one million registered users or 100,000 monthly active users must conduct security assessments covering eight areas, from training data handling to minor protection, and submit the reports to provincial regulators. App stores must verify compliance and remove non-compliant products.

On paper, this is a more comprehensive set of user protections than the EU, the US Federal Trade Commission, or California’s SB 243 has yet to put into force.

What the regulations leave unresolved

What the measures do not address is equally significant. They set no technical threshold for what constitutes emotional interaction, and this gray area is exactly why platforms removed entire features rather than risk falling on the wrong side. They combine genuine safety responsibilities with content control and national security provisions that serve the state rather than the user—a package that no other regulator would adopt wholesale.

They also leave unresolved how liability is divided between platform operators and upstream model providers when a violation originates from the model’s outputs, and they grant users no right to export their data. The enforcement context highlights the issue. On June 26, Shanghai’s internet regulator stated that it had removed over 14,000 non-compliant AI agents, citing impersonation of official entities, vulgar role-play, and unauthorized collection of personal data.

Whether this is the right path depends on which part of the rulebook you examine. The safety half addresses documented harms that are largely unregulated elsewhere, ranging from teenagers forming attachments to chatbots to companion apps harvesting intimate data. China’s own official interpretation looks abroad for support, citing the Character.AI lawsuits over psychological harm to teenagers, FTC investigations into companionship services, and European action against Replika.

The control half gives Beijing a lever over what these systems may say, wrapped in the same language of user protection. Both aspects are real, and governments observing this experiment will need to decide which parts they are willing to adopt. Pan Helin, a member of the MIIT expert committee, presented the official stance plainly to the South China Morning Post, stating that “current agents are not yet mature” and framing the policy around safety and standardization.

For now, the companies have taken the safest available route, which is to switch the components off and work out what a compliant version looks like later.

See also: Meta revises AI chatbot policies amid child safety concerns

Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo taking place in Amsterdam, California, and London. This comprehensive event is part of TechEx and is co-located with other leading technology events, including the Cyber Security & Cloud Expo. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.