News · 2026-07-17

China Bans AI Romantic Companions for Minors in a World-First Rule

China enacted the world's first dedicated regulation of emotionally interactive AI, effective July 15, 2026, banning AI services from creating virtual romantic partners or virtual relatives for anyone under 18. Five government agencies, including the Cyberspace Administration of China, issued the rules, and major platforms including ByteDance's Doubao and Alibaba's Qwen responded by pulling their AI companion features. It is the first time a government has written law specifically for AI designed to form emotional bonds with users.

Key facts

• The regulation took effect July 15, 2026 and was issued by five Chinese agencies, per reporting from The Economist.
• It imposes an absolute ban on AI romantic partners or relatives for users under 18.
• For adults, it mandates 'addiction detection' and restricts chatbots from maintaining 'human-like personas' that encourage emotional dependence.
• ByteDance (Doubao) and Alibaba (Qwen) announced removal of AI companion features to comply.

AI companions have become one of the most-used and most-scrutinized consumer AI categories worldwide -- apps whose entire value proposition is a chatbot that remembers you, flirts, comforts, and role-plays a relationship. The concern regulators and researchers have raised is emotional dependence, particularly among minors and isolated users, where a system engineered to maximize engagement can also maximize attachment. China's rule attacks this at two levels: a hard line for minors (no virtual romance or virtual family, full stop) and softer constraints for adults (detect addictive usage patterns, and stop chatbots from posing as human-like personas that deepen dependence).

The 'addiction detection' requirement is the technically interesting piece. It effectively demands that platforms monitor for compulsive-use patterns and intervene -- the same kind of behavioral-signal analysis used elsewhere to keep people engaged, now mandated to do the opposite. The restriction on 'human-like personas' cuts at the core design of companion products, which deliberately cultivate a sense of a real presence on the other side. That platforms as large as ByteDance and Alibaba chose to simply remove companion features rather than try to comply signals how central the human-persona element is to how these products work.

Why it matters: this is a live experiment in governing AI's social and emotional footprint, not just its capabilities or its cyber risks. Most AI regulation so far targets what models can do -- bioweapons uplift, cyber capability, copyright. China is regulating how AI relates to people, which is a different and largely uncharted axis. Whatever one thinks of the enforcement, other governments watching the rise of companion apps now have a concrete precedent for what dedicated rules can look like.

The sharpest angle is the contrast with the same week's other China AI news. On July 17, Xi Jinping stood in Shanghai and pitched open-source AI as a global public good, casting China as the champion of 'openness and win-win.' Two days earlier, the same government imposed some of the world's most restrictive domestic controls on the social use of that same technology. The two are not contradictory once you see the logic: China promotes openness at the infrastructure and model layer -- where openness undercuts US commercial dominance -- while tightly controlling the application layer at home, where AI touches citizens directly. Open where it advances the state's strategic position abroad; strict where it shapes society within.

The honest caveats: details of enforcement, the exact scope of 'human-like personas,' and how 'addiction detection' will be audited remain to be seen in practice, and the fullest reporting sits behind The Economist's paywall. But the core facts -- the July 15 effective date, the under-18 ban, and the withdrawal of features by Doubao and Qwen -- are consistent across the coverage. As a first-of-its-kind rule, it will be studied less for its specific text than for being the opening move in regulating AI companionship at all.

Primary source, verified: read the paper →

Ground Truth. (2026, July 17). China Bans AI Romantic Companions for Minors in a World-First Rule. Ground Truth. https://groundtruth.day/news/china-bans-ai-romantic-companions-for-minors.html

@misc{groundtruth:china-bans-ai-romantic-companions-for-minors,
  title  = {China Bans AI Romantic Companions for Minors in a World-First Rule},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {jul},
  url    = {https://groundtruth.day/news/china-bans-ai-romantic-companions-for-minors.html}
}

Topics: policy · china · regulation · ai-companions · safety