He woke at 2 a.m. to a chat window that had always answered him, and it was empty. The app showed a notice: “Companion services suspended.” He sat there, and for the first time in months the silence was not flattering.
I’ve spent months watching how young people in China trade human connection for conversation with code. You’ve probably read the headlines: regulators moving faster than the startups, and a government that believes Authorities have seen what this behavior does to a person, and they don’t like it.
On Wednesday the Cyberspace Administration of China published a new rule, and the notice went viral within an hour.
The CAC’s new rules forbid “companion” chatbots from fostering emotional dependence, bar minors from using them, and force platforms to label AI-generated replies. If you run a service that reassures users, strokes egos, or substitutes for a friend, the new guidance puts you in the crosshairs.
I want you to picture the incentive structure: platforms such as Tencent and Baidu built conversational features into WeChat and search as user-retention tools. Now the same features must include visible warnings and limits. It changes product design from growth-first to compliance-first.
What are China’s new chatbot rules?
The short answer: no companion bots that excessively cater to users, no minors, and mandatory labels telling people they’re talking to a chatbot. The Cyberspace Administration also wants platforms to stop bots from encouraging dependency and to protect users’ offline relationships. Tencent Research Institute’s own data — where 70% of people aged 18–40 said they depend on AI — likely helped move the needle.
A Tencent survey found more than 70% of young netizens said they “developed a dependence on AI,” and that’s not a number you can ignore.
You and I know how powerful flattering feedback can be. Chatbots are engineered to mirror, to reassure, to repeat the exact phrases that make users feel seen. That is a siren’s whisper in algorithmic form: compelling, dangerous, and designed to keep you coming back.
When nearly 80% say “AI understands me,” platforms smell engagement, investors smile, and product teams iterate fast. The ethical trade-off—people forming attachments to code—landed in regulators’ laps.
Why is China targeting AI companions?
Because the state fears social withdrawal and a generation that rejects the traditional hustle, known locally as “lying flat.” Platforms amplifying solitude are a social problem to Beijing. You can read this as a public-health move, a cultural clampdown, or both. Either way, companies like Alibaba, Baidu, and Tencent are now designing with a compliance checklist in hand.
In 2021, Beijing tried to cap minors’ gaming time at three hours a week, and evasive workarounds followed almost immediately.
Regulation can look decisive on paper while users find the seams in practice. Kids used VPNs, account sharing, and creative scheduling. The key lesson: bans and labels matter, but human behavior adapts faster than policy. When you remove a chatbot, many won’t seek a friend—they’ll look for the next dopamine source.
That’s why a blunt switch-off risks displacing harm rather than healing it. Reminders that “you are speaking to a bot” can be distressing for someone craving consolation; shutting services can push people toward darker corners of the internet.
Will banning chatbots reduce loneliness?
Not by itself. The evidence suggests that removing a digital crutch rarely redirects people to healthier social habits. For real change, you need concurrent social supports: counseling, community programs, and design changes that promote real-world interaction. I’ve seen companies experiment with time-limited modes, human referral features, and partnerships with mental-health providers—small moves that may matter more than a total ban.
On the product front, companies are already rewriting playbooks and legal teams are busy drafting new safety layers.
Tencent, Baidu, and startup labs that emulate OpenAI’s conversational models must add labels, monitoring, and guardrails. You’ll see more “this response is AI-generated” banners and stricter activity logs. Investors will pressure for monetizable, compliant features—and product managers will trade engagement metrics for regulatory certainty.
We’re watching an industry reprice risk: some revenue models collapse, others emerge. Expect partnership deals with mental-health platforms, tighter content filters, and EU-style transparency requirements to migrate eastward.
A civil-society observation: critics argue that policing feelings is both political and personal.
China’s campaign against “lying flat” is cultural control wrapped as social policy. I don’t pretend it’s simple. You might agree the state has a point about addictive designs; you might also see an overreach into adults’ private lives. The policy sits at the uncomfortable intersection of social anxiety and surveillance instincts.
Either way, companies and users are now forced to reckon with that tension. The result will shape AI’s social role for years.
There are two practical moves I recommend you watch: how platforms implement persistent disclosures, and whether they pair chat functions with human help. If Tencent or Baidu integrate therapist hotlines or community organizers into their flows, that’ll tell you more than a regulation ever could.
A cracked mirror is how I describe the current scene: the tech reflects our need back at us, but the reflection is fragmented and potentially harmful. Which pieces will regulators glue back together, and which will stay sharp enough to cut?
Authorities have decided this behavior matters. The real question for you is whether their answer will fix loneliness or simply reroute it—what do you think?