Millions of people in China have spent the past year talking to AI companions that remember their moods, greet them by name and play the role of friend, confidant or virtual partner. In a little over a week, many of those characters will go dark. A set of national rules taking effect on 15 July forces the country's largest AI firms to shut down their most humanlike features, and ByteDance and Alibaba have already told users the lights are going out.
The regulation is the Interim Measures for the Administration of AI Anthropomorphic Interactive Services, co-issued in April by the Cyberspace Administration of China alongside four other agencies, including the economic planning commission and the Ministry of Public Security. The name is a mouthful, but the target is specific: AI that is designed to feel like a person.
What switches off
ByteDance has told users that the agent function on Doubao, its flagship chatbot, will go offline on 15 July. Alibaba is moving faster in places, telling users its Qwen app will stop humanlike and user-created agents on 10 July, with wider agent services following five days later, Decrypt reported. Doubao alone counts hundreds of millions of users, so this is not a niche feature quietly being retired.
There is a sting in the tail for anyone attached to a particular character. Doubao is giving users until 15 October to export their data before it becomes unrecoverable inside the app. Qwen, by several accounts, has announced no migration path at all. People who poured months of conversation into a companion may simply lose it.
Who the rules are really for
The clearest thread running through the measures is child protection. Providers are barred outright from offering virtual companion or virtual family-member services to minors, and must get guardian consent before serving anyone under 14, a legal analysis of the rules notes. Beyond age limits, the measures require companion services to run anti-addiction systems, push mandatory usage reminders, offer an instant way out of a conversation and actively watch for signs that a user is forming an unhealthy dependence on the software.
That last requirement is the interesting one. Beijing is not just worried about what these systems say. It is worried about how much people come to need them. The concern that AI companions can foster real emotional dependency, particularly in young or lonely users, has been building in several countries, and China is the first major government to write it into enforceable rules rather than leave it to the companies.
A test other regulators are watching
Whatever one makes of China's broader approach to controlling online speech, the companion question it is trying to answer is genuinely hard, and Western regulators are circling the same issue. The United States has seen lawsuits and hearings over chatbots and teenage users, and Europe's rules on AI and minors are still taking shape. None of them has yet gone as far as ordering the features switched off.
China has, and it did so to some of its own national champions. The next few weeks will show what happens when a state decides that a popular, profitable and emotionally sticky product is too risky to leave running, and simply pulls the plug.
Commentarii · 0