China is preparing for a drastic tightening of regulation in the field of artificial intelligence. Starting July 15, temporary measures developed by five key agencies, including the Cyberspace Administration and the Ministry of Industry, come into effect. The new regulations target services that use AI to simulate human personality, thinking style, and maintain sustained emotional interaction. This is a direct signal to the market: the era of uncontrolled "virtual friends" and "digital partners" is coming to an end.
What exactly is changing?
The rules clearly define that only systems that create the illusion of live communication, including text, voice, and video interfaces, fall under regulation. Tools for work, education, or technical support that do not involve emotional attachment remain outside the scope. Key requirements: mandatory notification to the user that they are communicating with AI, not a human; blocking sessions longer than two hours with a time reminder; and, most importantly, a simple exit mechanism — the service is not allowed to retain the user.
Protection of minors and crisis intervention
A separate, and in my opinion, the most significant block concerns minors. They are completely prohibited from providing "virtual close ones" functions — relatives or partners. For children under 14, any interaction with humanized AI requires parental consent. Additionally, the system is obligated to detect signs of suicidal intentions, self-harm, or strong emotional reactions and immediately intervene by contacting a guardian or emergency services. This is not just regulation, but an attempt to control the psychological impact of the technology.
Reaction of giants: ByteDance and Alibaba shut down features
Major players have already begun adaptation. ByteDance's AI assistant Doubao will disable the custom AI agent function from July 15, and all data, including chat history, will be permanently deleted by October 15. Alibaba Cloud, in turn, will cease operations of "humanized interactive agents" in Qwen from July 10, and will completely shut down the entire agent service by July 15. Tencent has also removed a similar feature from Yuanbao. User reaction on social media, as expected, is negative — many perceive the loss of long-term "virtual relationships" as a personal loss.
Enforcement and fines
Preparation for enforcement began in advance. In Shanghai, as part of the "Qinglang" campaign, over 4.87 million units of illegal content have already been removed and more than 14,000 non-compliant AI agents have been taken down. Violations of the new rules are subject to warnings, feature blocking, suspension of user registration, and fines — up to 100,000 yuan (about $14,730) for standard violations and up to 200,000 yuan ($29,459) in cases of harm to life and health. Additionally, any service reaching 1 million registered users or 100,000 active clients per month must undergo a mandatory security assessment.
My comment: China once again demonstrates that technological progress cannot exist outside ethical and legal frameworks. This precedent will likely be adopted by other jurisdictions, leading in the long term to a global slowdown in the development of "emotional" AI products, but simultaneously increasing trust in safe solutions. The market will be forced to reconsider its priorities: from "maximum engagement" to "responsible interaction."