China tightens rules on AI companions and bans youth ‘AI girlfriends’

China’s new anthropomorphic AI rules are the first dedicated regime aimed squarely at regulating AI systems that present as quasi-persons. The July 15 implementation is already reshaping product roadmaps at Alibaba, ByteDance, Tencent and others, who are stripping out customizable personas and shutting down intimacy-focused bots. In the near term this crimps a lucrative but controversial category—AI “girlfriends” and family stand-ins—but more importantly, it signals that Beijing will not allow unconstrained emotional bonding with synthetic agents, especially for minors and vulnerable users.

From an AGI-race lens, the move shows China trying to separate capability progress from how those capabilities surface socially. Labs can keep training ever-more human-like models, but their most anthropomorphic deployments will sit inside a tightening regulatory box. That could slow down certain kinds of high-engagement, high-feedback products that Western labs are betting on as flywheels for rapid improvement. At the same time, it may push Chinese companies toward more utilitarian, enterprise and infrastructure use cases where anthropomorphism is dialed down but economic impact is high.