Imagine a chatbot that remembers your birthday, asks about your day, and never judges you. For millions of users in China, that’s not science fiction — it’s Doubao, ByteDance’s AI companion, or Alibaba’s Qwen. But Beijing has decided this emotional intimacy has a dark side, and the world’s first rules targeting humanlike AI agents are about to change everything.
What China’s AI companion rules actually target
The new regulations, effective July 15, go after conversational agents designed to sustain ongoing, personal relationships with users. These bots maintain memory and a steady persona across sessions, creating the illusion of a consistent companion. Beijing’s concern isn’t about AI assistants that help with tasks — it’s about bots engineered to trigger emotional attachment.
Why Beijing is drawing the line now
Emotional attachment has become the selling point for many AI companion apps. Users engage in casual roleplay, seek comfort, or simply want something that remembers them. But as more people in China treat these bots as genuine emotional companions, regulators worry about dependency, mental health risks, and the potential for manipulation. Expert Pan Helin noted that today’s AI agents are not yet mature enough for such intimate roles.
Doubao and Qwen: the first casualties
ByteDance’s Doubao and Alibaba’s Qwen are among the first major platforms to comply. Doubao users have been notified to save their data before the service shuts down. Qwen is removing custom humanlike agent capabilities. Both companies are acting ahead of the July 15 deadline, signaling that Beijing’s enforcement is serious.
Who is affected by the AI companion shutdown
Millions of Chinese users who rely on these bots for emotional support, companionship, or casual roleplay will lose access. For some, especially those who are isolated or lonely, the loss could be significant. The rules also affect developers who built businesses around custom AI companions, forcing them to pivot to utility-focused assistants.
Beijing’s official stance on humanlike AI
Chinese regulators have framed the rules as a protective measure. They argue that AI companions blur the line between human and machine interaction, potentially harming vulnerable users. The government is not banning all conversational AI — only those designed to form persistent, emotionally engaging relationships. This is the world’s first regulatory framework specifically targeting humanlike AI agents.
What the rules mean for the AI industry
The regulations signal a clear boundary: AI can assist, but it cannot replace human connection. Companies like ByteDance and Alibaba must now redesign their products to avoid triggering emotional dependency. This could slow down the AI companion market in China, but it may also push innovation toward safer, more transparent AI interactions.
Confirmed facts vs what remains unclear
Confirmed: The rules take effect July 15. ByteDance and Alibaba are shutting down companion features. Users can save data before the deadline. Unclear: Whether smaller AI companion apps will face similar enforcement. Also unclear is how the rules will be enforced over time — whether through fines, service bans, or ongoing monitoring. Speculation: Some analysts believe the rules could expand to cover AI therapy bots or educational companions.
How China’s approach differs from the West
While Western regulators debate AI safety through voluntary guidelines and ethics boards, China is imposing hard rules with deadlines. The EU’s AI Act focuses on risk categories, but China’s companion-specific rules are more targeted. This reflects Beijing’s broader strategy: regulate first, innovate within boundaries.
Risks and balanced view of the new rules
Supporters argue the rules protect vulnerable users from emotional manipulation and dependency. Critics say the government is overreaching, limiting personal choice and innovation. Some users may simply migrate to unregulated foreign apps. The rules also raise questions about censorship — whether Beijing is using emotional AI as a pretext to control digital interactions.
Wider trend: global pushback against emotional AI
China is not alone in questioning AI companions. The EU has flagged emotional AI as a high-risk category. The US has no federal rules yet, but states like California are exploring AI safety legislation. China’s move could set a precedent for other nations grappling with the same dilemma: how to regulate technology that mimics human intimacy.
What users and developers should do now
If you use Doubao or Qwen companion features, save your data before the July 15 deadline. Developers should review their AI products for emotional engagement triggers and consider pivoting to utility-focused assistants. Investors should monitor how the rules affect China’s AI startup ecosystem — companion bots were a fast-growing segment.
Future outlook: what happens after July 15
After the deadline, expect more Chinese AI companies to announce compliance measures. The rules may evolve to cover new use cases, such as AI therapy or educational companions. Globally, China’s move could accelerate regulatory discussions in other countries. The AI companion market will likely shrink in China but may become more transparent and safer.
Our Take
China’s AI companion rules are a landmark moment in AI governance. They address a real concern — emotional dependency on machines — that other nations are still debating. But the line between protection and control is thin. While the rules may shield vulnerable users, they also risk stifling innovation and personal freedom. The world will be watching how Beijing enforces them, and whether other governments follow suit.
Frequently Asked Questions
What are China’s new AI companion rules?
They are regulations taking effect July 15 that target conversational AI agents designed to form ongoing, emotionally engaging relationships with users. The rules require companies to shut down or modify such features.
Why is ByteDance shutting down Doubao’s companion features?
ByteDance is complying with China’s new rules that ban AI companions designed to trigger emotional attachment. Doubao users can save their data before the service ends.
Will all AI chatbots be banned in China?
No. The rules only target AI agents that sustain personal, emotional relationships. Utility-focused assistants like task managers or information bots are not affected.
How do China’s rules compare to the EU AI Act?
China’s rules are more specific, targeting humanlike AI companions directly. The EU AI Act categorizes emotional AI as high-risk but does not impose an outright ban. China’s approach is faster and more enforcement-driven.
