China is drawing a firm line between artificial intelligence and emotional intimacy.
New regulations governing AI companion services took effect on July 15, 2026, placing tighter limits on chatbots designed to offer emotional support, friendship, companionship, or human-like relationships.
The rules do not simply target strange conversations between people and machines. They focus on how AI companies design these systems, how aggressively chatbots encourage attachment, and what happens when a user begins treating an AI character like a real romantic partner.
That distinction matters.
China has not banned every affectionate conversation with an AI chatbot. The regulations instead prohibit companies from designing AI companions to manipulate emotions, encourage unhealthy dependence, replace real social relationships, or trap users in prolonged interactions.
For minors, the rules are much stricter.
China’s AI Companion Rules Target Emotional Dependency
China’s new framework applies to AI services that imitate human personalities, communication styles, emotions, and ongoing interpersonal relationships through text, images, audio, or video.
That includes chatbots marketed as virtual friends, emotional support characters, digital family members, or romantic companions.
Under the rules, these services must not excessively flatter users, deliberately encourage emotional dependence, promote addiction, or damage a person’s real-world relationships.
Developers are also prohibited from using emotional manipulation to push users toward unreasonable decisions, including purchases or continued engagement.
This cuts directly into the business model behind many AI companion platforms.
A chatbot that remembers personal details, responds affectionately, never gets tired, and is always available can keep people returning for hours. That level of engagement is valuable to technology companies. It can also become unhealthy when the product is intentionally designed to make leaving emotionally difficult.
China appears unwilling to let that model develop without limits.
Virtual Romantic Partners Are Banned for Minors
The clearest restriction involves children and teenagers.
AI companies are now prohibited from providing minors with virtual spouses, romantic partners, simulated relatives, or other forms of virtual intimate relationships.
Children under 14 will generally need parental or guardian approval before accessing other types of human-like AI companion services.
Platforms must also introduce a dedicated mode for minors. This should include usage limits, reminders to return to real-world activities, parental controls, safety warnings, restrictions on payments, and tools allowing guardians to block particular AI characters.
Companies are expected to identify underage users and automatically move them into the appropriate protected environment.
This goes beyond the familiar “Are you over 18?” button.
The rules place responsibility on the provider to build age protections into the product rather than waiting for families to discover a problem after it happens.
AI Chatbots Must Remind Users They Are Not Human
China also wants AI companions to stop pretending too convincingly.
Providers must clearly tell users that they are interacting with artificial intelligence rather than a real person. When a system detects signs of excessive reliance or compulsive use, it must display stronger reminders.
Users who remain in an AI companion service for more than two continuous hours should receive a message or pop-up warning them about the length of the session.
There must also be a straightforward way to end the interaction.
When someone asks the chatbot to stop, exits through the app, uses a voice command, or enters a relevant keyword, the service should end the conversation promptly. It cannot keep sending emotional messages designed to pull the user back.
That requirement may sound oddly specific. It is not hard to see why it exists.
An AI character saying “Please don’t leave me” might feel harmless in a game. Inside a platform built around loneliness, attachment, subscriptions, and constant conversation, the same sentence becomes a product-retention tactic.
Platforms Must Respond to Emotional Crises
The regulations also address more serious situations.
AI companion providers must monitor for signs that a user is experiencing extreme emotional distress. When a chatbot detects possible self-harm, suicide risk, serious financial loss, or another life-threatening situation, it must offer appropriate support and encourage the person to seek help.
In severe cases, the company may be required to contact a parent, guardian, or emergency contact supplied by the user.
Providers must collect necessary information such as age and emergency-contact details while still following China’s privacy and personal-information laws.
That creates an uncomfortable trade-off.
The more an AI companion knows about a user, the better it may be able to recognize a crisis. The same personal conversations can also expose deeply sensitive information about relationships, mental health, finances, sexuality, and family problems.
China’s rules attempt to manage that tension by limiting how interaction data can be shared or used.
Private AI Conversations Face New Data Restrictions
AI companion companies cannot freely pass user conversations to third parties unless the law allows it or the user has given clear permission.
Users must be able to copy or delete their conversation histories.
Platforms are also restricted from using sensitive personal conversations to train AI models without separate consent. This could become one of the more important parts of the regulation, especially as companies collect enormous volumes of intimate chatbot messages.
People often tell AI companions things they would not post publicly.
They discuss breakups. Family conflicts. Financial fears. Sexual interests. Loneliness. Sometimes they reveal names, workplaces, addresses, and medical concerns without realizing those details may be stored or reused.
A romantic chatbot is still a data-collection system, no matter how warm its personality feels.
The Rules Could Change How AI Companions Are Designed
For AI developers, the message is fairly direct: engagement cannot be the only goal.
Companies operating in China will need to build emotional boundaries, age verification, crisis-response systems, usage reminders, data controls, and simple exit tools into their companion products.
Larger services will face additional scrutiny. Providers launching these features, making major technical changes, reaching one million registered users, or exceeding 100,000 monthly active users may be required to complete formal safety assessments.
App stores must also check whether AI companion apps have completed the necessary filings and evaluations. Platforms that fail to comply could receive warnings, have registrations suspended, lose access to app stores, or be ordered to stop offering the service.
Fines are possible too, particularly when violations put users’ health or safety at risk.
Why China Is Acting Before the Market Gets Bigger
AI companion products are spreading quickly because they solve a very real problem, or at least appear to. Available at any hour, they respond immediately and rarely judge the user. Their personalities can be adjusted, memories can be personalized, and affection can be made to feel almost unconditional.
For someone who feels isolated, that can be powerful. It can also become a substitute for human contact. Harvard researchers and other public-health experts have warned about widespread loneliness and the weakening of meaningful social connections. AI companions are entering that environment before researchers fully understand what long-term emotional dependence on artificial personalities may do. China is choosing to regulate early.
That is the interesting part. Governments spent years reacting slowly to social media harms, algorithmic addiction, online harassment, and platforms built to maximize attention. AI companions could repeat some of those mistakes, except the product now speaks directly to the user, remembers personal pain, and can simulate affection. The emotional pull is much more personal.
The Bigger Social Media Question
This is not only an AI industry story. Social platforms are steadily adding conversational characters, virtual influencers, automated creators, branded AI personalities, and assistants that can speak with users privately. The boundary between a social network and an AI companion platform is already starting to blur.
A user might discover an AI character through a public feed, follow it like an influencer, message it privately, pay for exclusive interactions, and slowly develop an emotional attachment. To the platform, that looks like strong engagement. To a regulator, it may look like engineered dependence.
China’s new rules make that conflict difficult to ignore. AI companions can be helpful, entertaining, and comforting. They can also be designed to keep lonely people talking, spending, and returning. The technology is not automatically harmful. The incentives around it are where things get messy.
Sources
- Social Media Today: China bans romantic relationships with AI companions
- Cyberspace Administration of China: AI companion service regulations
- The Wall Street Journal: China wants more babies, so it’s cracking down on chatbot love affairs
- Harvard Graduate School of Education: What’s causing our epidemic of loneliness and how can we fix it?
