China drafts strict rules for human-like AI companion apps

China’s new draft rules for anthropomorphic AI are one of the clearest attempts yet to regulate emotionally engaging “AI companions” as a distinct class of systems. By forcing providers to detect user overdependence, issue time‑based reminders, and intervene when behavior looks addictive, Beijing is signaling that parasocial AI relationships are a regulatory priority, not a sci‑fi edge case. The lifecycle obligations around algorithm review, data security and content safety effectively treat these services more like quasi‑clinical products than casual chatbots. ([reuters.com](https://www.reuters.com/world/asia-pacific/china-issues-drafts-rules-regulate-ai-with-human-like-interaction-2025-12-27/))

For the race to AGI, this is less about raw capability limits and more about the operating envelope for human‑like agents at scale. Any lab hoping to monetize companion models in China will have to bake in continuous affect monitoring, robust safety pipelines, and behavioral safeguards from day one. That tends to favor well‑capitalized players that can afford compliance engineering—large Chinese platforms and multinationals with strong regulatory teams—over small startups. At the same time, the explicit focus on psychological risk may create de facto global norms: once you’re building dependency detection for China, it’s cheap to deploy similar guardrails elsewhere.

Strategically, this moves the frontier of AI governance beyond content moderation into regulating the human‑AI relationship itself. As generative agents become more persistent, personalized and proactive, whoever defines those relationship norms will quietly shape how near‑AGI systems are allowed to show up in everyday life.