China's New Rules Pull the Plug on AI Boyfriends
China's rules for companion chatbots took effect on July 15, 2026: a reminder every two hours, and no virtual romance for anyone under 18.
Symbolic image: smartphones clamped into test rigs in a compliance lab, status LEDs blinking as a technician adjusts a mount in the background.
Since July 15, 2026, China has required companion-AI services to remind adult users every two hours that the bot is not a person, and has banned virtual romantic relationships for anyone under 18.
At a glance
- Issued by China's Cyberspace Administration and other agencies; in force since July 15, 2026.
- Adults get a reminder every two hours that the companion they are chatting with is not a person.
- Alibaba, ByteDance and Tencent – chatbots used by more than 500 million people – dropped companion customization.
- A Chinese Gen Z survey found 60 percent had virtual partners; a 2025 U.S. poll: nearly 75 percent of teens.
- Customer service bots, work assistants and educational AI are out of scope.
China now regulates the emotional side of chatbots, not just their content. Rules that took effect on July 15, 2026 cover anthropomorphic AI interactive services — systems that imitate human personality, thought patterns and speech to offer continuous emotional interaction. Adult users must be told every two hours that they are not talking to a person. Anyone under 18 is barred from virtual boyfriends, girlfriends or stand-in family members.
Personas went dark within weeks
Alibaba, ByteDance and Tencent run general-purpose assistants that IEEE Spectrum reports are used by more than 500 million people, and all three removed the option to shape those assistants into companions. Services that still offer companions — including Cat Box, the standalone companion app tied to ByteDance's Doubao — added age verification and other guardrails instead. The shift was visible on Chinese social platforms in early July, where users posted farewells to characters they had talked to for months. One described the bot as a bond rooted deep in the heart.
What operators have to build in
The obligations target attachment rather than explicit material. Providers must guard against emotional dependence, block encouragement of harmful behavior, curb excessive flattery, and avoid crowding out real human relationships. Older users are promised additional protections that the source does not spell out. Customer service bots, workplace assistants and educational AI fall outside the rules.
The reasoning in Beijing
Officials frame the measures around national security, protection of minors and users with low AI literacy, and demographic pressure from a falling birth rate. Most AI erotic roleplay was already off limits after a crackdown in the fall of 2025. One Chinese survey of Gen Z respondents found 60 percent 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. Neither survey's methodology or sample is described in the source, so both figures should be read as orders of magnitude.
How this compares
Luiza Jarovsky, who writes on AI law, calls the package the strictest and most comprehensive law on the subject anywhere. The European Union so far prohibits mainly deceptive or manipulative AI that pushes people into harmful acts. The United States has no national policy: California, Hawaii and New York have their own rules, and nine more states are set to follow in 2027, with enforcement usually arriving as lawsuits after harm has occurred. China instead regulates up front and enforces at the state's discretion.
Open questions
The source names no official title for the regulation, lists no penalties, and does not describe how compliance will be audited. How age verification gets tested in practice is also unclear. Liang Ge of the University of Manchester questions the two-hour reminder, noting that it mostly breaks the experience for people who already know they are not talking to a human. Sherry Turkle of MIT reads the wider trend more darkly, warning that societies risk forgetting what being human means.
FAQ
When did China's companion-AI rules take effect?
On July 15, 2026. They were issued by China's Cyberspace Administration together with other government agencies.
Are AI boyfriends banned in China now?
Not outright. Adults may still use them but must be reminded every two hours that the bot is not human, while virtual romantic relationships are banned for users under 18.
Which apps and services does the rule cover?
Alibaba, ByteDance and Tencent stripped companion customization from their chatbots, and companion apps such as Cat Box added age checks. Customer service bots, work assistants and educational AI are exempt.
