Two of China’s biggest tech companies are racing to shut down some of their most popular AI features, and the deadline is July 15.

ByteDance’s Doubao and Alibaba’s Qwen — both major AI platforms with tens of millions of users — are disabling the humanlike agent features that let users build persistent emotional relationships with AI. The cause is a new Chinese regulation called the “Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services,” and it signals something important for anyone running AI in a business context.

What the New Rules Actually Say

The regulation targets AI services that “simulate human personality traits, thinking patterns and communication styles to provide sustained emotional interaction.” In plain terms: if your AI is designed to feel like a person over time — building rapport, remembering emotional history, becoming a companion — it now falls under strict compliance requirements in China.

The rules demand anti-addiction systems, mandatory usage notifications, and instant-exit mechanisms. The goal is to protect users, particularly younger ones, from forming unhealthy dependencies on AI companions.

Doubao notified its users that the agent feature would go offline on July 15 due to “product function adjustments.” Qwen moved faster, announcing its humanlike interactive agents and user-created agents would be disabled by July 10, with broader agent functions offline by July 15. After October 15, data from those features will be handled according to standard privacy policies and will no longer be viewable or recoverable inside the app.

Why They Shut Down Rather Than Modify

This is the telling detail: rather than adapting their products to comply, both companies chose a full shutdown.

The reason is architectural. A persistent-memory agent built to maintain a consistent emotional relationship with a user over time cannot simultaneously implement the friction that the regulation requires. You cannot design an AI to feel like a caring, continuous presence while also building in mandatory interruptions, addiction warnings, and instant-exit buttons. The two design goals are fundamentally at odds.

That means there is no easy retrofit. It is a rebuild — and neither company judged it worth the effort for these specific features.

What the Rules Do Not Cover

Here is what matters for business owners: the regulation explicitly excludes customer service bots, knowledge Q&A tools, workplace assistants, educational tools, and scientific research tools — provided they do not involve sustained emotional interaction.

That is most enterprise AI. An AI that answers questions about your inventory, drafts reports from your data, routes support tickets, or helps staff find information in your knowledge base does not fall under these rules. The regulation is targeting consumer AI companions, not productivity tools.

The distinction China is drawing — emotional AI versus functional AI — is one that enterprise teams should be paying attention to globally.

The Regulatory Divide Between Consumer and Enterprise AI

This is one of the clearest examples yet of AI regulation getting specific rather than broad. Early AI regulatory conversations tended to treat AI as a single category. What we are seeing in 2026 is regulators learning to draw distinctions: AI that advises on medical decisions, AI that targets ads, AI that forms emotional bonds with users — these are different things, and they are starting to be governed differently.

China’s anthropomorphic AI rules are likely a preview of similar distinctions coming in other markets. The EU AI Act and various US state proposals are already moving in this direction, separating high-risk applications from lower-risk productivity use cases.

For enterprise buyers, this is actually good news. Business AI — the kind that helps teams work faster, makes data accessible, and automates repetitive processes — is sitting in a much safer regulatory category than consumer AI companions.

What This Means for Business

If you operate in China or across Asia-Pacific, you need to review any AI features that could be interpreted as providing sustained emotional interaction. Customer-facing AI that is warm and personable is not the same as a persistent AI companion, but the distinction matters. Get your legal team to look at the specific language in the regulation before July 15.

If you are evaluating enterprise AI platforms, prioritize vendors who are building for business outcomes, not engagement metrics. An AI workforce platform designed to handle operational tasks, surface data insights, and automate workflows operates on different design principles than a consumer AI companion. That separation is now legally meaningful in at least one major global market.

If you are thinking about AI strategy, pay attention to the line between functional AI and emotional AI. Tools that help people work smarter sit on the right side of that line. Tools designed to make users emotionally dependent on them are attracting regulatory scrutiny worldwide, and the pressure is only going to increase.

The platforms being shut down in China were built for consumer engagement. The AI that helps your business run more efficiently was built for different goals entirely — and right now, regulators are starting to recognize that difference.

Enterprise DNA’s Omni services are designed around operational value: AI agents that handle business workflows, voice AI employees that manage knowledge and reporting, and advisory services that help leadership teams build data-informed strategy. That is functional AI, not emotional AI — and that distinction is increasingly where the regulatory lines are being drawn.

For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.