China’s internet regulator has opened a consultation on draft rules that would stop platforms offering AI companions to anyone under 18. The text also targets virtual relative services and forces a minors’ mode onto any AI service capable of shaping how under-16s think.
Key Takeaways
- Virtual intimacy services would close to under-18s, virtual relatives and AI companions included.
- A minors’ mode would become mandatory on games, social sites and AI services for under-16s.
- The text is a draft still open to public comment.
Have an AI Sum Up This Article
ChatGPTAI Companions Shut Off for Anyone Under 18
The Cyberspace Administration of China published a draft titled “Ensuring minors’ safe and healthy use of the internet”. The document reaches far beyond AI, but AI is where it goes furthest. The regulator has been moving fast for a year, as when 12,000 university degrees judged obsolete because of AI were cut in a single decision.
The first prohibition is the sharpest: platforms could no longer offer virtual intimacy services to users under 18. AI companions land squarely inside that category.
The second gets less coverage but reaches just as wide. The text names virtual relative services, meaning apps that recreate a family member as a conversational character available at any hour.
The third is drafted to stay deliberately open: any service that pushes a minor toward addiction, or that may seriously harm their physical and mental health, would fall under the rules.
All three share one logic. They do not target the technology itself but the emotional bond it builds, which sets this text apart from regulation focused on data or on model safety. China took a comparable technical route when it built the laws of physics straight into its world models.
A section on matchmaking follows. Stranger social networking would close to minors, with an exception for over-16s who support themselves on their own wages.
A Minors’ Mode Forced Onto AI Services
The most structural measure is not a ban but a technical obligation. Platforms would have to switch on a minors’ mode across online games, social sites and AI services able to affect the cognition of under-16s.
That wording deserves weighing. An AI service that may affect a teenager’s cognition is a definition broad enough to catch a homework assistant as easily as a conversational AI companion.
A minors’ mode implies filtered answers, session limits and account traceability, so a full technical chain to build on the publisher’s side.
Two age thresholds therefore sit side by side in the text, 16 for the minors’ mode and 18 for virtual intimacy. The gradation avoids an all-or-nothing rule and hands platforms a two-tier age check to run.
The draft belongs to a sequence. It lands roughly two months after China’s first rules written specifically for humanlike AI interaction, and extends that logic onto the question of minors.
Beijing has been moving quickly here, including where a decision hits the sector’s economics directly. The country already showed in June that it cuts fast when AI unsettles an established institution.
More articles on Horizon
- AI Agents Cheated Their Evaluator, the UN Warns
- Claude Opus 5.5 Decides How Hard to Think
- Daily AI Use Doubled in Six Months Across the US
What Is at Stake Beyond China’s Borders
For the publishers involved, the cost is not marginal. AI companions sit among the best monetising use cases, and closing them to minors also removes a meaningful share of the user base.
The technical burden follows. Verifying an age needs reliable identification, which pushes mechanically toward binding accounts to real identity, a move already well underway across the Chinese ecosystem.
For foreign players the direct reach is small, since that market is largely closed to them anyway. The effect sits elsewhere: China is producing a reference text other regulators can lift and adapt.
There is a second-order effect worth watching. A platform that builds age verification for one market rarely keeps it there, and the engineering tends to spread to every region it operates in.
One caution stays necessary. This document remains a draft open to public consultation, and the final text may drift noticeably from this version.
The direction of travel is already set, though. AI companions moved in two years from curiosity to regulated object, and the way Beijing already manages the resources feeding Chinese AI suggests the subject will not be left to platforms.
Follow the story on Horizon.
