China tightens AI rules on ethics, agents and anthropomorphic chatbots

China is quietly building one of the most detailed regulatory frameworks for AI agents and anthropomorphic systems, and July 15 is a key inflection point. The Interim Measures on anthropomorphic interactive services clamp down on AI companions and therapists that mimic human personalities, adding labelling, content, and anti‑addiction rules with special protections for minors. Parallel standards for AI agents require unified identities, logging, and security controls for autonomous systems that act across services. Together, these rules codify the idea that 'always‑on' and 'emotionally human' AIs warrant tighter guardrails than generic chatbots.([iapp.org](https://iapp.org/news/a/china-s-new-ai-rules-ethics-ai-agents-and-anthropomorphic-ai?utm_source=openai))

For the global AGI race, the immediate effect is that Chinese consumer AI products are pruning or disabling agent features rather than retrofit them—ByteDance’s Doubao and Alibaba’s Qwen are already notifying users that custom companions and smart agents will go dark on July 15. That may slow down certain high‑engagement use cases, but it also pushes Chinese labs to design agents with compliance baked in from day one. Over time, that could give them an edge in exporting 'regulation‑ready' agent architectures to jurisdictions that follow China’s lead.([global.chinadaily.com.cn](https://global.chinadaily.com.cn/a/202606/26/WS6a3e5e3da310986e2b462288.html?utm_source=openai))

Outside China, these measures serve as a concrete template for policymakers who want to move beyond principles into operational rules. The more influential that template becomes, the more AGI builders everywhere will have to treat emotional dependency, identity, and autonomy as first‑class design constraints rather than afterthoughts.