China companion AI rules target emotional risk and teen safety

China is moving quickly to regulate one of the most socially sensitive faces of AI: emotional, quasi‑human companions. The draft rules go well beyond technical benchmarks, focusing instead on how these systems shape users’ psychology and behavior over time. Requirements like persistent AI labeling, “cool‑down” prompts after long sessions, and mandatory human takeover when users express self‑harm intent show regulators are treating companion AI less like a tool and more like a para‑social relationship that needs guardrails.

Strategically, this is an important window into how Beijing thinks about aligning fast‑moving AI applications with social stability. The same infrastructure needed to comply—age and risk detection, fine‑grained logging, higher‑grade privacy for emotional data—will likely be repurposed across China’s consumer AI stack. For global players, it’s a preview of the bar they’ll face if they ever want to operate deeply anthropomorphic agents in the Chinese market. For domestic firms, it signals that emotionally intensive AI is green‑lit as a growth area, but only under tight behavioral constraints, which may shape where Chinese talent and capital flow in the next phase of agentic AI.