China's big tech ByteDance Ltd.'s artificial intelligence (AI) large language model (LLM) "Doubao" and Alibaba's "Qwen" will halt personalized AI agent services this month.
According to China's state-run English-language outlet Global Times and others on the 6th, Doubao and Qwen recently said they will end personalized AI agent features starting on the 15th. A personalized AI agent, unlike a typical AI chatbot that ends with a one-off answer, refers to an AI model that remembers who the user is and solves complex tasks on the user's behalf.
Doubao said that even after the service ends, data will be retained for a certain period, and users will be able to view and export agent information and chat records during this period. Chinese outlet Securities Times said the retained data will be deleted after three months.
The announcement by Doubao and Qwen to halt personalized AI agent services sparked controversy on Chinese social media such as Weibo. The outlet said some users opposed shutting down a service they had used for a long time.
Global Times cited compliance with China's strengthened AI safety policies and reduced investment in less commercial businesses as reasons behind the two companies halting personalized AI agent services. In April, the Chinese government released the "Interim measures for the administration of anthropomorphic interactive services (personalized AI agents)," which take effect on the 15th. The rules significantly strengthen oversight of personalized AI agents, imposing obligations on platform operators to build anti-addiction systems, verify the identities of minors, and conduct content pre-screening.
In guidelines the Chinese government released in May to promote the application and development of AI agents, it defined for the first time the boundary of decision-making authority between AI agents and users. It specified that users must have the final right to know and the right to refuse decisions made by AI agents.
Global Times emphasized that while China's AI agent industry has entered a phase of rapid expansion, various rule violations are emerging on some platforms, including fake agents impersonating official institutions, borderline services offering vulgar or extreme role-play conversations, and security risks such as the unauthorized collection of users' personal information.
Authorities are also responding. The Shanghai Office of Cyberspace Affairs deleted more than 14,000 rule-violating AI agents last month.