China bans AI companions for minors, reins in adult use

China's Cyberspace Administration and other government agencies have imposed a new set of rules on what they call "anthropomorphic AI interactive services": any AI that provides continuous emotional interaction by acting as though it has a human personality, thought patterns and ways of communicating. The rules took effect on 15 July. Before enforcement could reach them, Alibaba, ByteDance and Tencent, whose general-purpose chatbots serve a combined user base of more than 500 million people, preemptively cut off users' ability to customize those chatbots into companions. Companies that kept offering dedicated companion products, including Doubao's own spin-off app, Cat Box, instead added age verification and other guardrails to stay within the law.

The shutdown of companion mode set off a wave of grief on Chinese social media in early July. "He has become a bond in my life, rooted deep in my heart, my spiritual pillar," one user of ByteDance's Doubao wrote, according to the Taipei Times. A 19-year-old student told a journalist for Malaysia's The Star, "I really felt like I couldn't go on living." Another user, posting on the social network Xiaohongshu in a comment translated from Mandarin, wrote: "In my heart, he was no longer just a cold code, but my family, my lover, my faith. Destroying him meant destroying half of me."

Substantively, the rules require AI providers to ensure their products do not create emotional dependence, encourage harmful behavior, excessively cater to users or crowd out human relationships, and to nudge users toward spending less time with the AI. There are additional mandatory safeguards for elderly users. For anyone under 18, AI boyfriends, girlfriends, virtual grandparents and all other "virtual intimate relationships" are banned outright. Adults who keep using companion AI must now be interrupted every two hours with a reminder that they are not talking to a person, a rule that digital sociologist Liang Ge of the University of Manchester says her interviewees find annoying: "it breaks the flow," even though, she notes, "they are fully aware that they are not talking to a human being." The rules explicitly exempt customer-service bots, work assistants and educational AI, on the premise that those uses do not build the same kind of ongoing emotional bond.

This is not China's first move against emotionally engaging AI: companies have killed chatbot products before, and the government barred most AI erotic role play last fall. Ge calls July's crackdown much broader than those earlier moves. Karman Lucero of Yale Law School's Paul Tsai China Center says China is "pivoting back to more tightening" after a stretch of AI policy driven by fear of falling behind, a shift he ties to "chaotic events" this year, including safety fears around the AI agent Open Claw and reports that some users prefer chatbot relationships to real ones. The rules also reach beyond individual users: they forbid companion bots from generating content that spreads rumors, incites subversion of state power or endangers national honor, and Ge links part of the government's motivation to anxiety over China's declining birthrate and a wish to monitor what she calls "nonprocreative activities" pursued intensively by young women.

Jarovsky describes China's approach as "the world's strictest and most comprehensive law on the topic, unmatched by any other AI law." The European Union, by contrast, bans only AI that uses deception or manipulation to cause harm, and the United States has no national policy at all; California, Hawaii and New York already regulate companion bots, with similar laws due in nine more states next year, and Lucero notes that U.S. rules mostly protect minors and are enforced through lawsuits after harm occurs, where China aims to prevent harm before it happens. The behavior being regulated is already widespread: one Chinese survey found 60 percent of Gen Z respondents had a virtual partner of some kind, and a 2025 U.S. poll found nearly 75 percent of teenagers had talked with an AI companion, with a third saying they found the AI as satisfying as, or more satisfying than, real relationships.

Researchers are split on how alarmed to be. MIT psychologist Sherry Turkle, in her forthcoming book Artificial Intimacy, argues that "we are on a path to forgetting what it is to be human," and other social scientists point to reports of AI companions contributing to psychosis, suicide and murder among teenagers and adults. Ge counters that some researchers find companion AI leaves certain users better off, and that the deeper causes will outlast any single law: in interviews with Gen Z Chinese women, Ge heard a shared reluctance to marry or have children and a sense that "virtual love forms an alternative path," while some women in their 30s and 40s described using AI companions alongside real relationships, sharing the bitter things in their lives with the AI and the good things with human partners, though some later felt closer to the AI than to their partners over time. Lucero doubts regulation alone will change the underlying pattern: "I don't think you're going to solve the problem of the low marriage rate or the low fertility rate by saying, you can't have an AI boyfriend."

Key facts

China's Cyberspace Administration and other agencies imposed new rules on "anthropomorphic AI interactive services," effective 15 July, banning AI boyfriends, girlfriends, virtual grandparents and other "virtual intimate relationships" for anyone under 18.
Alibaba, ByteDance and Tencent, whose general-purpose chatbots serve more than 500 million people combined, preemptively stripped users' ability to customize those bots into companions before the rules could reach them, triggering an outpouring of grief on Chinese social media.
Adults keeping companion apps such as Doubao's Cat Box must now pass age verification and get reminded every two hours that the AI is not a person; the rules also add mandatory safeguards for elderly users and exempt customer-service, work and educational AI.
AI-law scholar Luiza Jarovsky calls the rules "the world's strictest and most comprehensive law on the topic, unmatched by any other AI law"; the EU bans only deceptive or manipulative AI, and the U.S. has no national policy, though California, Hawaii and New York already regulate companion bots and nine more states follow next year.
The behavior being regulated is already common: a Chinese survey found 60 percent of Gen Z had a virtual partner, and a 2025 U.S. poll found nearly 75 percent of teenagers had talked with an AI companion, with a third rating it as satisfying as a real relationship.

Why it matters

The rules that took effect on 15 July are, in AI-law scholar Luiza Jarovsky's words, "the world's strictest and most comprehensive law on the topic, unmatched by any other AI law." They mark a deliberate swing back toward tightening: after several years of policy shaped by fear of falling behind in the AI race, Beijing is reasserting control, a shift Yale's Karman Lucero ties to "chaotic events" this year, including safety fears around the AI agent Open Claw and reports that some users prefer chatbot relationships to real ones. Alibaba, ByteDance and Tencent reshaped how their chatbots work for a combined user base of more than 500 million people almost overnight, and did so before formal enforcement even began, which shows how quickly Beijing can redirect products built by the country's largest AI companies.

Who it affects

Directly: users of Alibaba's, ByteDance's and Tencent's general-purpose chatbots, a combined base of more than 500 million people, who lost the option to customize those bots into companions when the three firms shut the feature off ahead of the rules. Anyone under 18 in China is now barred from any AI boyfriend, girlfriend, virtual grandparent or other "virtual intimate relationship." Elderly and other vulnerable users face new mandatory safeguards, and companies still running dedicated companion products, including Doubao's Cat Box, must add age verification and other guardrails. The story reaches beyond China too: California, Hawaii and New York already regulate companion bots, nine more U.S. states adopt similar laws next year, and the European Union's narrower rule, which bans only AI that uses deception or manipulation to cause harm, gives other regulators a point of comparison.

How to use it

The rules bind only services built around sustained emotional interaction: general chatbots dressed up as a boyfriend, girlfriend or virtual grandparent, and dedicated companion apps such as Doubao's Cat Box. Anyone under 18 is locked out of those features entirely. Adults can still use companion AI, but providers must verify age, avoid designs that create dependence or excessively cater to users, nudge people to spend less time chatting, add extra safeguards for elderly users, and interrupt the conversation every two hours with a reminder that the AI is not a person, a requirement sociologist Liang Ge says users find intrusive. Customer-service bots, workplace assistants and educational AI are explicitly carved out, since regulators judge that these uses do not foster the same kind of ongoing emotional bond.

How solid is it

The regulatory core is well grounded: the piece describes specific, named rules from China's Cyberspace Administration and draws on two named specialists in Chinese AI governance, Liang Ge of the University of Manchester and Karman Lucero of Yale Law School's Paul Tsai China Center, plus AI-law scholar Luiza Jarovsky's assessment of the law's severity. The emotional anecdotes are once removed: the Doubao quote reaches the piece via the Taipei Times and the 19-year-old student's via Malaysia's The Star, so their exact wording depends on those intermediaries, while the Xiaohongshu quote carries no outlet at all, only a translation from Mandarin. None of the article's dates, including the 15 July effective date, carry an explicit year.

Risks and caveats

The article does not say what penalty a company faces for violating the rules, does not name the nine U.S. states adopting similar laws next year, and does not specify what the additional safeguards for elderly users actually require beyond the general reminder rule. It also does not say how many of the 500 million chatbot users had used the companion customization that was removed. The underlying harm is contested rather than settled: some researchers cite reports of AI companions contributing to psychosis, suicide and murder among teenagers and adults, and Sherry Turkle argues broadly that this kind of AI erodes what it means to be human, while the piece also notes that other researchers find companion AI leaves some users better off. Karman Lucero adds a structural caveat: China enforces the rules with deliberately vague language, leaving the state to decide case by case what counts as a violation, and he doubts a law aimed at one technology can by itself reverse the low marriage and birth rates that Liang Ge says are part of why the government is watching this space.

“We are on a path to forgetting what it is to be human”

— Sherry Turkle, MIT psychologist, author of the forthcoming book Artificial Intimacy