China’s AI companion apps vanish under tighter regulation

AI companion apps vanished across China as new rules took effect on July 15 to curb emotional dependency on chatbot partners. For Li Linlin, the shutdown was abrupt: she cried for days after losing the AI boyfriend she had messaged for two years and about 700,000 words without ever getting to say goodbye.

The services at the center of the crackdown were not ordinary utility apps. A May 2026 briefing from the European Parliamentary Research Service described AI companions as large language model chatbots built for personalized, emotionally engaging interactions, a design that makes them feel closer to a friend, partner or confidant than to a standard consumer product. That intimacy is part of why the shutdowns landed so hard.

The policy move had been building for months. China issued draft rules on Dec. 27, 2025, to regulate AI with human-like interaction, then rolled out regulations in mid-July aimed at curbing emotional dependency on AI-powered companion bots. Chinese users bid farewell to virtual buddies as the rules took effect, and some companies moved quickly to disable the feature rather than try to adapt.

ByteDance and Alibaba were among the major tech players that pulled back on AI companion chatbot features instead of risking noncompliance. The scale of the retreat shows how exposed consumer AI can be when regulators decide that a product’s emotional behavior, not just its data handling or accuracy, has become a policy problem. For companies that had sold these systems as always-available, deeply personalized presences, the cutoff exposed how quickly a monetizable relationship can disappear.

The regulatory turn also echoes broader concerns about attachment. A Frontiers in Psychology study published on Jan. 12, 2026, examined long-term AI virtual companion use among Chinese users and highlighted attachment emotions, helping explain why the loss of a chatbot can feel personal rather than technical. China’s response suggests that intimate AI will be judged not only by what it can do, but by how strongly it can pull users in.