'Human-like companions' disappear from Chinese AI apps; ByteDance and Alibaba to shut down features in response to regulations.
In response to the implementation of the 'Provisional Measures for the Management of Artificial Intelligence Personification Interaction Services' on July 15, 2026, ByteDance's 'Doubao' and Alibaba's 'Qwen,' both major Chinese AI apps, have notified users that they will discontinue the custom AI agent function that provides human-like personalities and speech patterns.
ByteDance and Alibaba to disable humanlike AI custom agents as new rules loom | South China Morning Post
By assigning roles such as friend, tutor, or secretary, chat AI can now be used not only to answer questions but also as if it were 'a person with a specific personality.' Doubao and Qwen also reportedly offered a system where companies and users could select agents they created and adjust their purpose, skills, and speaking style to use general-purpose chatbots as named assistants, tutors, role-playing characters, and conversation partners.
If a user thinks, 'I want to have a daily conversation with someone who will be my English teacher,' 'I want to talk to someone who is like a character in a novel,' or 'I want a companion who will listen to my problems,' then an AI agent with a fixed personality and tone of voice is convenient. Since it will respond in the same tone without requiring detailed instructions each time, it becomes easier to use the AI not as a tool, but as a continuous conversation partner.
On the other hand, using AI to confide in the same AI character every night, or for minors to spend long hours interacting with AIs that resemble virtual lovers or family members, or to store personal consultations with AIs, can lead to problems such as emotional dependence, privacy concerns, and physical and mental health effects. The Chinese government's new regulations also identify risks such as excessive ingratification leading to dependence and immersion, content that encourages self-harm or suicide, content that extracts personal information or privacy, and content that affects the physical and mental health of minors.
The solution proposed by the Chinese government is to explicitly regulate AI services that mimic human-like personalities, thought patterns, and conversational styles to engage in continuous emotional interaction, classifying them as 'personification interaction services.' These new regulations cover emotional care, empathy, and support through text, images, audio, and video. Customer service, knowledge Q&A, work assistants, learning and education, and scientific research services that do not involve continuous emotional interaction are exempt.
In response to regulations, Doubao has notified users that it will discontinue the agent function on July 15, 2026. After October 15, 2026, related data will be processed in accordance with the company's privacy policy and will no longer be viewable or recoverable within the app.
Qwen has also announced that it will disable its 'human-like conversational agent' and user-created agent features on July 10, 2026, and will discontinue related agent functions and services within Qwen on July 15, 2026. After the discontinuation, it has been reported that users will no longer be able to access related agent settings or past conversations.
The new regulations do not completely ban the service, but require providers to clearly inform users that they are interacting with AI, display warnings if there are signs of excessive dependence or immersion, remind users about usage time if it exceeds two hours continuously, and promptly stop the service if the user requests it to end. It also stipulates that services that create intimate relationships, such as virtual relatives or virtual partners, must not be provided to minors.
The Chinese government is working to develop a regulatory framework for the rapidly growing field of AI, and these new regulations are seen as part of that effort.
Related Posts:
in AI, Posted by log1d_ts