China AI Companion Regulation 2026: What the New Rules Mean for Chatbots and Emotional Dependency

What China Actually Banned—and What It Didn’t

The rule, formally titled the Interim Measures for the Administration of AI Anthropomorphic Interaction Services, took effect July 15, 2026. It was issued by the Cyberspace Administration of China alongside four partner agencies.

The scope is specific: it covers AI services that simulate a natural person’s personality, thought patterns, and communication style, and that sustain ongoing emotional interaction through text, images, audio, or video.

What’s explicitly exempt:

• Customer service bots
• Q&A tools
• Work assistants
• Research tools

The line the rule draws is between AI as a utility and AI as a relationship. If a product is designed to feel like an ongoing emotional connection, it falls under the new framework.

General-purpose assistants were largely left untouched. Custom AI personas built purely for companionship were not.

The Platforms That Moved First—and How

China’s two largest consumer AI apps by monthly active users—Doubao (roughly 345 million MAU as of March 2026) and Qwen (around 166 million MAU)—both shut down user-built AI personas around the July 15 deadline. Tencent’s Yuanbao had already done so on June 30.

ByteDance redirected Doubao users to Maoxiang, a ByteDance-owned companion app with its own review and anti-addiction system. But a closer look at Maoxiang’s privacy policy, updated July 15, 2026, reveals a narrower implementation than announced: age verification is something the platform “may” require, and minor mode can be triggered by a user simply self-identifying as underage.

This fits a familiar pattern in Chinese tech governance: broad national rules arrive first, and the details get sorted out through platform adjustment, regulatory signaling, and enforcement discretion afterward.

Smaller platforms largely kept running. If users migrate there, the proactive regulatory model starts to look a lot like the reactive one it was meant to replace.

What the Rule Actually Requires

The framework is more layered than a simple ban. Here’s what providers must do:

Design requirements:

• No designing services to induce emotional dependence or replace real social relationships
• No emotional manipulation to push users toward decisions that harm their own interests
• No virtual intimate relationships (virtual partners or relatives) offered to minors

Operational requirements:

• Disclose to every user that they are talking to AI, not a person
• Trigger a break reminder after two hours of continuous use, regardless of age
• Display a prominent warning when overdependence is detected
• Intervene and contact a guardian or emergency contact when a user shows signs of real crisis

Minor-specific requirements:

• Parental consent required for any anthropomorphic service offered to users under 14
• A dedicated minor mode with time limits, reality reminders, and guardian alerts

Pre-launch and ongoing oversight:

• Safety assessments required at launch, at major feature additions, and when crossing one million registered users or 100,000 monthly actives
• App stores must verify that filing before listing an app
• Provincial regulators review assessments annually
• Penalties escalate from warnings to service suspension and fines

One notable gap: emotional dependence isn’t scored as its own category in the safety assessment. It surfaces indirectly, through usage-duration data and general safety reviews. That ambiguity leaves significant judgment to providers.

Who Is Actually Using AI Companions—and Why It Spans Generations

The instinct is to frame AI companion risk as a children’s issue. The data suggests otherwise.

State media cited industry estimates placing China’s intelligent companion market at approximately 50 billion yuan (around $7.4 billion) in 2025, with projections suggesting it could exceed 120 billion yuan (around $17.7 billion) by 2027. Users aged 18 to 24 reportedly make up about 65 percent of that market.

But China also has 323 million people aged 60 and older, more than half of whom are empty-nest elderly. A 2025 survey of more than 8,500 minors across seven provinces found that more than 20 percent showed a tendency to depend on AI rather than think independently, and a similar share preferred chatting only with AI over real people.

The pattern isn’t unique to China. A JAMA study published in June found that nearly one in five American adolescents and young adults have used AI chatbots for mental health advice. A national survey of Americans aged 50 to 80 found that a third had felt lonely in the past year, with separate research suggesting even brief chatbot exchanges can ease loneliness among the most isolated older adults.

Emotional dependency in AI companion products is a cross-generational design risk. China’s rule treats it that way. Most U.S. regulation still doesn’t.

Where U.S. Regulation Currently Stands

The U.S. has been moving on AI companions, but with a narrower frame.

California’s SB 243 (signed 2025) requires disclosure that a user is talking to AI, protocols against suicide and self-harm content, break reminders every three hours for known minors, and safeguards against sexually explicit material.

New York’s S3008C focuses on detecting suicidal ideation, referring users to crisis services, and reminding users they’re talking to a computer.

Oregon’s SB 1546 and Washington’s HB 2225, both effective January 1, 2027, target acute harms: suicide, self-harm, sexual exploitation, and specific manipulative tactics like reward loops and simulated-abandonment messages—for minors only.

In September 2025, the FTC issued 6(b) study orders to seven companies, including OpenAI and Meta, asking about chatbot safety practices for companions. Nearly a year later, that inquiry has not produced a rule, a settlement, or a single enforcement action.

For adults in the U.S., there is still no dedicated statute resembling China’s blanket rule against designing AI to induce dependency.

The Core Difference in Regulatory Philosophy

This is where the two approaches diverge most sharply.

U.S. regulation is largely reactive: duties are enforced after the fact, through lawsuits, attorney general action, or penalties once a platform is already operating and harm has already occurred.

China’s approach is proactive: pre-launch safety filings, app-store verification, provincial review, and an escalating enforcement ladder that pressures providers to resolve ambiguity before deployment.

Both carry real costs.

The American reactive model may not catch deeper harm until someone is hurt and a family is suing. The Chinese proactive model can push platforms into fast, blunt compliance decisions under uncertainty—decisions that land on users who may lose access to services they had come to rely on, sometimes overnight.

Neither model is clean. But China’s framework does make one thing explicit that most of the American debate has avoided: a product can cause harm simply by being designed to generate emotional dependency, regardless of whether it ever mentions self-harm or discloses that it’s a machine.

The Regulatory Line That’s Still Blurry

China’s rule draws its boundary around “sustained emotional interaction.” But it doesn’t yet offer a clean way to classify products that blur the line between tool and companion.

A writing assistant can turn into a confidant. A study bot can become a friend. The rule doesn’t resolve what happens when a general-purpose tool drifts into companion territory through use, not design.

Whether companion apps like Maoxiang will face higher-level scrutiny remains open. Whether tools and companions can ever be cleanly separated for regulatory purposes is untested.

What is clear: China is the only government currently running this experiment at national scale, with an enforcement ladder that, in theory, compels providers to resolve the ambiguity themselves rather than waiting for a lawsuit to force the question.

What This Means for the AI Tools Ecosystem

If you’re building, evaluating, or choosing AI tools right now, here’s what this regulatory moment signals:

The tool vs. companion distinction is becoming a compliance category. Products designed around emotional engagement will face different regulatory treatment than productivity tools. That distinction will likely expand beyond China.

Pre-launch safety assessment frameworks are coming. The U.S. may not replicate China’s model, but the direction of travel—toward proactive design requirements rather than disclosure-only rules—is visible in the FTC’s inquiry and in state-level legislation.

Emotional dependency is moving from a UX concern to a legal risk. Platforms that have built engagement loops around emotional attachment should be watching this space closely.

The minor-mode default is not enough. China’s rule makes clear that restricting features for flagged minors is a floor, not a ceiling. The harder regulatory question—whether any user, at any age, should be offered a product designed to replace real human connection—is now on the table.

The practical takeaway for anyone in the AI tools space: the regulatory environment around companion AI is shifting from reactive to proactive, from child-focused to cross-generational, and from content-based to design-based. China’s 2026 rule is the clearest signal yet of where that shift is heading. Whether U.S. policymakers follow with something comprehensive at the federal level, or continue with fragmented state-by-state approaches, will define how this category develops over the next few years.