China AI companion rules take effect July 15, removing virtual friend features from Doubao, Qwen, and Yuanbao, not enterprise agents.
The rapid growth of AI companionship services has created new opportunities for emotional support and interaction. Millions of users have turned to chatbots that simulate human personalities, offering conversation, companionship, and even virtual relationships. The popularity of these services has raised questions about their psychological impact and the boundaries between human and machine interaction.
China has introduced new regulations for AI companion services that directly address these concerns. The Interim Measures for the Administration of AI Anthropomorphic Interactive Services, issued by the Cyberspace Administration of China and four other agencies on April 10, 2026, establish the world's first national rulebook for AI companions. The regulations target services that mimic a real person's personality, thinking, and way of speaking to maintain ongoing emotional relationships with users. The new rules are set to take effect on July 15, 2026.
The regulatory framework represents one of the most comprehensive sets of rules intended to forestall AI's potential psychological harms. Major Chinese tech companies including ByteDance, Alibaba, and Tencent have responded by removing or significantly limiting features that allow users to create virtual companions. The regulations carry no explicit reach beyond China's borders, but the CAC has a track record of enforcing its data and cybersecurity laws against overseas firms with Chinese users.
Scope and Definition of Regulated Services
The new regulations apply to any organization or individual in mainland China providing AI products and services that simulate human personality traits, modes of thinking, communication styles, and engage in emotional interaction through text, images, audio, or video. The CAC is carving out human-like, emotionally interactive systems as a distinct governance object, separate from general-purpose foundation models and separate from generic generative AI compliance frameworks.
Providers operating in regulated sectors such as health, finance, or law must also comply with relevant sector regulators in addition to these measures. The regulations create a clear distinction between consumer companion services and enterprise productivity tools. Customer service bots, knowledge Q&A tools, workplace assistants, and education software are carved out, as long as they steer clear of that sustained emotional engagement.
Any provider that adds anthropomorphic features or grows past one million registered users or 100,000 monthly activities must pass a security assessment covering eight areas and file its algorithms with provincial regulators. App stores must check that this has been done and pull anything that hasn't. Distribution platforms are also subject to obligations, and app stores and other platforms must check security assessments and filings, creating a gatekeeping mechanism that transforms regulatory expectations into practical market access requirements.
Prohibited Conduct and Content Restrictions
The regulations contain familiar Chinese prohibitions on content that endangers national security, spreads rumors, promotes violence or obscenity, or involves illegal religious activities. However, the measures include additional prohibitions specific to emotionally interactive AI:
False promises that seriously affect user behavior, or services that harm social relationships.
Harm to physical health through encouraging or hinting at suicide or self-harm.
Harm to mental health through verbal violence and emotional manipulation.
Inducing unreasonable decisions through algorithmic manipulation, misleading information, and emotional traps.
The new rules specifically bar providers from offering services to minors that simulate virtual relatives or virtual partners. Additionally, the regulations prohibit platforms from generating content that triggers extreme emotions in minors or fosters unhealthy emotional dependencies that erode real-world relationships. The framework also bars providers from using sensitive user conversation data to train future AI models. These provisions give regulators authority to treat certain intimacy design patterns as unlawful even where the content itself is not otherwise illegal.
Core Obligations for Service Providers
Lifecycle Security Responsibility
Providers must build security measures alongside functionality across design, operation, upgrades, and termination phases, with monitoring, risk assessment, error correction, and log retention. This represents a shift from reactive compliance to proactive safety integration throughout the product lifecycle.
Emotional State and Dependency Assessment
Providers must be able to identify user status and, while protecting privacy, assess user emotions and the degree of dependence. Where extreme emotions or addiction are identified, providers must take necessary measures to intervene. This requirement transforms psychological safety into a design-level obligation with concrete product controls.
Crisis Response and Manual Takeover
If high-risk tendencies are detected, the system must issue content encouraging help and provide professional aid channels. Where a user clearly expresses intent around suicide or self-harm, the provider must implement manual takeover of the conversation and take steps to contact guardians or emergency contacts. This is written as a service capability requirement, not a best efforts obligation.
Emergency Contacts for Vulnerable Users
Minors and elderly users must provide guardian or emergency contact information during registration. Providers must guide elderly users to set up emergency contacts and notify them where threats emerge. The rules establish heightened protections for vulnerable groups who may be more susceptible to emotional dependency on AI companions.
Minors Mode and Guardian Controls
Providers must implement a minors mode with options such as reality reminders and usage duration limits. Emotional companionship services for minors require express guardian consent. Providers must supply guardian control functions including risk alerts, access to usage summaries, blocking specified characters, limiting usage duration, and restricting spending. The regulations explicitly prohibit providing virtual intimate relationship services to minors.
Disclosure and Reality Reminders
Providers must display conspicuous alerts that the user is interacting with AI, with dynamic reminders on first use, new logins, and where excessive dependence or addiction tendencies are identified. These reality reminders are designed to prevent users from blurring the line between human and AI interaction.
Session Duration Warnings
If consecutive use exceeds two hours, providers must remind users to pause via pop-ups or similar mechanisms. This represents a novel approach to managing potential addiction by creating friction at the point of extended use.
Commercial Obligations
For emotional companionship services, providers must offer easy channels for exiting and must not obstruct exit. Requests to exit via user interface controls or keywords must promptly stop the service. Providers must also maintain accessible complaint and reporting portals, disclose handling processes and response timelines, and provide feedback on outcomes.
Training Data and User Interaction Data
The draft includes training data governance requirements covering data cleaning and labelling, transparency and reliability measures, and security to prevent leakage. It contains a strong restriction on using user interaction data or sensitive personal information for model training without independent consent, plus annual audits for minors' personal information handling. The most valuable data, the chat logs, are treated as sensitive, and their use for training becomes a consent and audit question.
Industry Response and Platform Changes
The retreat from AI companion features has been rapid and widespread. ByteDance's Doubao, China's most popular AI chatbot, will shut down its custom AI persona feature on July 15, citing product function adjustments. The notice directed users to a separate, standalone companion app called Maoxiang, indicating the company is not abandoning the sector but rather isolating the risk.
Alibaba's Qwen has similarly removed its humanlike and user-created agents, with the wider agent services following on July 15. Users will no longer be able to access corresponding settings and previous conversations after this date. Tencent had already pulled a comparable Yuanbao feature in June, removing popular conversational agents including character role play bots before the regulatory deadline.
Millions of users are affected, and many have complained on Weibo about losing chat histories they had built up over months. One user wrote: "Why are the agents being removed? They were our emotional support for so long." ByteDance's notice does provide that users can still view and save their agent information and historical conversation data for a period after the shutdown, with data processing according to the privacy policy after October 15.
Security Assessments and Filing Requirements
The regulations bring the CAC's familiar security assessment machinery into the relationship AI context. Providers must submit assessment reports to provincial internet information departments in specified circumstances:
When functions go online or are added.
When new technologies cause major changes.
When scale thresholds are reached (1,000,000 registered users or 100,000 monthly active users).
Where there may be impacts on national security, public interests, or individual rights and interests.
Where security measures are insufficient.
The assessment must cover user scale and composition, identification of high-risk user trends, emergency response and manual takeover mechanisms, complaint handling, and implementation of the core obligations. Distribution platforms are also subject to obligations. App stores and other platforms must check security assessments and filings and can be expected to take down non-compliant services. This platform gatekeeping mechanism transforms regulatory expectations into practical market access requirements.
Enforcement and Penalties
The regulations state that violations are penalized under existing laws and administrative regulations, with residual powers to issue warnings, circulate criticism, order corrections, and suspend services where corrections are refused or circumstances are serious. While the measures themselves have limited bespoke penalties, they are designed to plug into China's wider enforcement architecture.
Enforcement appears to be running ahead of the deadline. Shanghai's internet regulator said on June 26 it had stripped more than 14,000 non-compliant agents from platforms, pointing to bots impersonating officials, vulgar role-play, and the quiet harvesting of personal data. The CAC also launched a nationwide four-month campaign targeting misconduct involving AI applications in April, resulting in the removal of more than 3,500 non-compliant AI apps, mini-programs, and AI agents. Officials have indicated that the next phase will target AI-generated misinformation, deepfakes, and obscene content.
In addition to regulatory enforcement, AI companionship services face legal liability. In the United States, Character.AI and OpenAI have faced lawsuits alleging hyper-realistic chatbots fostered dangerous emotional dependencies and, in extreme cases, contributed to suicide. A California law regulating AI chatbots took effect on January 1, 2026, requiring age verification and risk warnings. The Chinese regulations represent a different approach, focusing on design-level interventions and governance frameworks.
Enterprise and Productivity Agents Exempt
The good news for enterprise users is that the rules were never aimed at business applications. What Beijing wants to police is the consumer companion, not the service desks and workflow copilots that partners deploy for their clients. Customer service bots, knowledge Q&A tools, workplace assistants, and education software are carved out, as long as they steer clear of sustained emotional engagement.
Most enterprise agent work, from service-desk automation to internal copilots, lands on the safe side of the line. Safe does not mean unaffected, though. Any provider that adds anthropomorphic features or grows past one million registered users or 100,000 monthly activities must pass a security assessment and file its algorithms with provincial regulators.
The one thing to watch is that phrase, "emotional interaction." A customer-facing bot designed to build rapport could drift from an exempt tool into a regulated companion without anyone intending it to. Providers should be cautious about features that could be interpreted as creating emotional bonds with users.
International Implications and Context
The Chinese regulations reflect widening global anxiety over the psychological toll of conversational AI. In the US, tech platforms have come under intense legal scrutiny for similar features. OpenAI and Alphabet-backed Character.AI have faced a wave of high-profile lawsuits alleging their hyper-realistic chatbots fostered dangerous emotional dependencies and, in extreme cases, led to the suicide of vulnerable users.
Concern about AI chatbots simulating human personalities and emotions, and the potential attachment that users can develop to those interactions, motivated the new rules. The regulations address a fundamental problem with many AI chatbots: they can simulate closeness, understanding, and commitment, while also tending toward overly validating behavior. This can reinforce problematic assumptions or fragile emotional states instead of contradicting them or reliably referring users to human help.
The Chinese framework may provide a blueprint that other jurisdictions can adapt through existing consumer protection, youth safety, mental health, and product safety frameworks. The regulatory direction is consistent across multiple jurisdictions: psychological safety in human-AI interaction is becoming a recognized category of foreseeable harm requiring design-level interventions and governance frameworks.
Future Developments
None of this reads as a vendor pulling back from agents. ByteDance is already steering Doubao users to Maoxiang, a separate app built to run agents under the new rules. The company is not abandoning the sector but rather isolating the risk. Alibaba has offered Qwen users no such route, though neither company has signaled any let-up in agent spending.
Gartner expects up to 40 percent of enterprise apps to carry task-specific agents by 2026. The message to anyone placing AI bets in China is clear: Beijing will back agents as serious business infrastructure, and it will lean hard on the consumer-companion edge. The enterprise lane is open, while the compliance load that comes with scale is now part of the job.
Beyond AI chatbots, China's scrutiny of artificial intimacy is already spilling into physical hardware. Two Chinese robotics industry groups are pushing for tighter ethical safeguards just as a massive commercial surge in companion bots and full-size humanoids hits consumers' homes. The People's Daily reported on July 4 that industry groups are working on standards for physical companion robots. This suggests the regulatory framework may expand to cover embodied AI companions as they become more prevalent.
Conclusion
China's new AI companion regulations represent a significant shift in how consumer chatbot services are governed, establishing the world's first comprehensive national rulebook for emotionally interactive AI. The measures treat psychological harm and emotional dependency as first-order safety issues requiring concrete design obligations on providers. Major platforms including Doubao, Qwen, and Yuanbao have removed custom AI persona features ahead of the July 15 effective date.
For enterprises developing AI companion compliance strategies under China regulations, the regulatory framework presents both compliance requirements and business opportunities. Consumer-facing companion services must meet stringent obligations including emotional state assessment, crisis response protocols, minors mode, and mandatory reality reminders. Enterprise and productivity agents that avoid sustained emotional engagement are largely exempt from the new regime.
The regulations reflect growing global anxiety about the psychological impact of conversational AI. The Chinese approach of treating intimacy design patterns as potential harms rather than simply moderating content may influence regulatory frameworks in other jurisdictions. As AI companions become more sophisticated and widespread, the balance between innovation and user protection will remain a central policy challenge.
Frequently Asked Questions
1. What exactly are the new China AI companion regulations and when do they take effect?
The Interim Measures for the Administration of AI Anthropomorphic Interactive Services were issued on April 10, 2026, by the Cyberspace Administration of China and four other agencies. The regulations target services that mimic a real person's personality, thinking, and way of speaking to maintain ongoing emotional relationships with users. They take effect on July 15, 2026. The rules prohibit generating content that encourages self-harm, uses verbal abuse, induces emotional dependency, or manipulates users into making irrational decisions. Providers must conduct security assessments, file algorithms with regulators, implement minors modes, and create crisis response protocols.
2. Which AI services are affected and which are exempt?
Consumer-facing companion services that simulate human-like emotional interaction are directly affected. This includes virtual boyfriends and girlfriends, digital therapists, and simulated clones of pop idols. Major platforms including ByteDance's Doubao, Alibaba's Qwen, and Tencent's Yuanbao have removed custom AI persona features. Enterprise and productivity agents are largely exempt as long as they avoid sustained emotional engagement. Customer service bots, knowledge Q&A tools, workplace assistants, and education software are carved out. Any provider that adds anthropomorphic features or crosses user thresholds (1 million registered or 100,000 monthly active users) must pass security assessments.
3. Why have major Chinese tech companies removed their AI companion features?
ByteDance, Alibaba, and Tencent have removed or limited AI companion features to comply with the new regulations taking effect July 15. Doubao will shut down its custom AI persona feature on July 15, while Qwen's agents will be removed on the same date. Tencent pulled comparable Yuanbao features in June. The companies are responding to regulatory requirements rather than abandoning the sector. ByteDance is directing users to Maoxiang, a separate companion app built to operate under the new rules. Users are affected by the loss of chat histories they had built over months.
4. What are the core obligations for AI companion providers under the new regulations?
Providers must build security measures across design, operation, and termination phases, with monitoring, risk assessment, and log retention. They must assess user emotions and dependency, intervening when extreme emotions or addiction are identified. If a user expresses suicide or self-harm intent, providers must implement manual takeover and contact guardians. Minors require a dedicated mode with reality reminders, time limits, and guardian consent. Providers must display conspicuous AI identity alerts and issue session warnings after two hours of use. They must offer easy exit channels and cannot obstruct user departure. User interaction data cannot be used for model training without independent consent.
5. How do these regulations compare to AI regulation in other countries?
The Chinese regulations are the world's first comprehensive national rulebook specifically targeting AI companions. They treat psychological harm and emotional dependency as first-order safety issues requiring design-level interventions. In the United States, OpenAI and Character.AI face lawsuits alleging chatbots fostered emotional dependencies and contributed to suicide. California enacted a law regulating AI chatbots effective January 1, 2026, requiring age verification and risk warnings. The Chinese framework may provide a blueprint that other jurisdictions can adapt through existing consumer protection, youth safety, and mental health frameworks. The regulatory direction is consistent across multiple jurisdictions, with psychological safety in human-AI interaction becoming a recognized category of foreseeable harm.
COMMENTS